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>
A failed move of /dev, /proc, /sys or /run is logged at dbg() level
and ignored, so the new init boots without its virtual filesystems
and falls over much later in some unrelated way. The steps after,
chdir/mount/chroot and the final execl(), do return -1 on failure,
but by then all services are dead and the API socket is gone, so
there is nobody left to report to: the system hangs with a live but
useless PID 1.
Make a failed move fatal, and try all four moves even if one fails,
so a bad /dev does not also skip /proc, /sys and /run. A plain
directory is not an error though: /run stays a plain directory on a
tmpfs rootfs, so only a path on a different device than / is treated
as a mount point and moved.
Any failure past the point of no return now drops to sulogin(1) for
a maintenance shell that reboots on exit, same as a fatal fsck() at
boot. Signals are unblocked before the moves so the shell does not
inherit finit's blocked signal mask.
Signed-off-by: Paweł Sobczak <github@fixeq.qzz.io>
The precheck accepts a directory as new init: access(path, X_OK)
only checks search permission, which directories almost always have,
so the mistake is not caught until after teardown. The snprintf()
building init_path also never checks for truncation, so an overlong
newroot + newinit validates the wrong path.
Require a regular file, with EISDIR or ENOEXEC to match what is
printed, and fail with ENAMETOOLONG on a truncated init path.
Signed-off-by: Paweł Sobczak <github@fixeq.qzz.io>
initctl switch-root always exits 0, no matter what happens on the
finit side. do_switch_root_api() sends the ACK before switch_root()
has validated anything, so any failure after that point never reaches
the client -- the error only shows up in the log. The runlevel guard
in api_cb() has the same problem: it rejects the request but still
sends an ACK.
Split the validation out of switch_root() into switch_root_precheck()
and run it before the ACK. A failed check now sends a NACK with the
error message, which initctl prints before exiting 1:
initctl switch-root /mnt
switch-root: /mnt is not a mount point
The ACK is only sent once the precheck passes, since after that point
we are committed. On success finit execs the new init and the
connection dies with no reply at all, so initctl treats only an
explicit NACK as failure.
Signed-off-by: Paweł Sobczak <github@fixeq.qzz.io>
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>
The other dbus-*.sh tests drive libink's own client, so the wire format
was only ever checked against the implementation that wrote it, and the
broker path had no coverage at all. Every bug found in it so far was
found by hand on a target.
Let the dbus plugin bring up a real dbus-daemon, wait for Finit to
claim org.finit, then talk to Finit with dbus-send, which shares no
code with us. The privileged call is the interesting one: it can only
be answered by parking the call and asking the broker who sent it.
The bus reads the policy Finit installs, so a malformed org.finit.conf
fails here rather than on a target.
Tests no longer build --with-libsystemd. Our replacement carries the
real soname but only the sd_notify() symbols, so in the test root it
shadowed the libsystemd the host's libdbus-1 wants and dbus-daemon
died on a missing sd_is_socket. Nothing under test needs the shared
library: serv is the only consumer and it compiles sd-daemon.c
straight in, which it now does regardless of the flag so notify.sh
keeps testing notify:systemd either way.
Staged from the host by lib/sysroot.mk like any other binary, and
skipped when the host has neither program.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Finit speaks D-Bus itself now and claims org.finit on the system bus
when it finds one, but nothing in a default build ever brings that bus
up. The plugin that does was opt-in, so the built-in support sat idle
unless the integrator knew to ask for both halves.
Defaulting it on is only reasonable if the result stays the admin's to
change, and a service registered from C through conf_save_service() is
not: it lands in the run path where it cannot be overridden or emptied
out. So the daemon moves to 20-dbus.conf and its directories to
tmpfiles.d/dbus.conf, the same way hotplug and every other daemon we
ship them for. The plugin keeps only what has to look at the running
system, the stale pidfile and the machine UUID.
Those directories are no longer chowned to messagebus. tmpfiles.d
skips a line whose user does not exist rather than falling back, so
the plugin's messagebus/dbus/root ladder has no equivalent there, and
dbus-daemon binds its socket before dropping privileges anyway.
The plugin already bows out where there is no dbus-daemon installed,
so systems that never wanted a bus are unaffected, and
--disable-dbus-plugin is there for those that have one and would still
rather init left it alone.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
A peer can be dropped from inside its own read loop: a handler emits a
signal, the write to that very peer fails, and the drop lands while
link_connection_process() still holds the connection and will touch
its rx buffer on the way out. Freeing there pulls the ground out from
under it. Unlink the peer and let the event loop free it once the
stack has unwound.
The work has to be scheduled with a non-zero delay. A uev timer armed
with zero is a disarmed timer, so the queue would never run and the
connections would leak instead.
Losing a peer is also not a warning. It is what shutdown looks like
from here, and every reboot said so on the console.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The legacy socket logs a line per command under initctl debug; the bus
logged nothing, so the transport that now carries most of initctl was
the one you could not watch.
libink gets a logger hook rather than a dependency on Finit's: it
passes the emitting function and a formatted message, and dbus.c hands
both to logit() so the two sources read alike. Trace points cover the
connection lifecycle, every inbound call, and why a call was refused.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
On the local bus SO_PEERCRED says who is calling and the kernel is the
one saying it. Behind a broker one connection carries every caller,
so that credential describes dbus-daemon and nothing else, and every
privileged method was refused there, root included.
Ask the bus driver instead. libink parks the call and hands us the
sender; we ask GetConnectionUnixUser and answer when the reply lands,
through the same event loop as everything else. Nothing blocks:
blocking in PID 1 is why libuEv exists. That needs calls libink can
make on a connection it already has, so it gained those too.
Answers are cached, since a bus never reuses a unique name while it
runs. Not across a restart though: a new dbus-daemon numbers from
scratch and :1.7 becomes somebody else, so the cache goes when the
broker does. A sender name too long to key on is refused rather than
truncated, two callers sharing a truncated key would share an
identity.
Privilege is no longer uid 0 alone. The socket is already owned by
the --with-group group, so refusing its members every method that
changes anything left a wheel user able to open the bus and unable to
reboot. Both gates now say the same thing.
Group membership needs NSS, which the C library loads with dlopen(),
so the lookup is compiled out where Finit is built to link statically.
That leaves such a build root-only, which is worth saying out loud
rather than leaving to be discovered.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The D-Bus socket was bound world read/write, on the reasoning that
SO_PEERCRED authorizes each method anyway. That leaves the read-only
surface open to every local user, and it quietly ignores --with-group:
a system that restricts initctl to the wheel group still handed the
same service state to anyone who asked over the bus.
Bind it 0660 and chown it to the configured group, the same gate the
fallback socket has always had. libink takes the mode as an argument
rather than assuming one, since who may connect is the embedder's
policy, not the library's.
The mode is applied at bind(), so there is no window where the socket
is more permissive than intended.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
libink was written peer to peer, where one connection is one client
and one principal. Attaching to a message bus breaks both halves of
that, and two things followed from it.
Signals never reached the system bus. Fan-out is gated on the peer
having sent AddMatch, but a broker subscribes for its own clients and
never sends us one, so every ServiceStateChanged was dropped on the
floor. A connection attached with LINK_ATTACH_BROKER gets them all.
Hello, AddMatch and RemoveMatch write per-connection state. Shared by
every caller behind a broker, that lets one sender exhaust the match
cap or drop another's rule, so we leave all three to the bus, whose
job they are.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The last assertion sampled bar's state one step after asking Finit to
stop foo, but bar passes through stopped on its way to waiting, so the
test failed roughly one run in eight.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The attach is best effort and its diagnostics were tuned for the case
where no broker exists, so a broker that answers but refuses us was
reported as a bare rc=1 at debug level. Chasing that meant reading
the header builder to find out what the number meant.
Failures now quote the error name the broker sent, and the one for a
name we could not claim says which of the three ways it went wrong.
Repeats stay quiet. The probe runs on every service and condition
change, and before syslog is up each line is an open, write and close
on /dev/kmsg, so a broker that keeps refusing would otherwise flood
the console during boot.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
libink was written against the only bus it had, its own, where the
peer on the other end is the client. A broker is not: it routes for
senders it names itself, expects a DESTINATION on anything addressed
through it, and answers on its own schedule rather than next.
Runlevels go on the wire as S and N rather than the digits Finit
keeps internally, since that is what a caller outside Finit means by
one.
The library stays a convenience library, linked into finit and
initctl and installed nowhere: the ABI promise waits until libink is
its own project.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The summary table, the per-service detail, JSON and the quiet and
ident forms all read state Finit already publishes, so they read it
from the bus like everything else rather than through a second path
that has to be kept in step.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Runlevel and version are state, not actions, so they belong behind
org.freedesktop.DBus.Properties rather than another method each.
Finit also claims org.finit on the system bus when it finds one, so
ordinary D-Bus clients can reach it without knowing about
/run/finit/bus. Opportunistic on purpose: no dbus-daemon is a normal
state for the systems Finit runs on, not an error to report.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The bus can already answer questions and change services, so give
initctl the two things it still did another way: watching signals as
they happen, and getting or setting user conditions.
The dbus tests move with it, split by area rather than one file that
grew every time the library did.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Finit had no way to answer the question every service manager gets
asked: what is running, and change it. D-Bus is how the rest of
userspace asks, but linking libdbus, sd-bus or GIO into PID 1 buys a
dependency, an allocator and a main loop we do not control.
So libink: the wire format, an object tree, and a bus of Finit's own
at /run/finit/bus, gated like INIT_SOCKET. It speaks the standard
org.freedesktop.DBus, .Peer, .Introspectable interfaces, and Finit's
own Manager1, Service1 and Cond1 on top. Methods that change
something are marked privileged and answered only for a caller the
kernel vouched for, via SO_PEERCRED.
Server and client both, since initctl is the first thing that needs
to talk to it, and its Start/Stop/Restart/Reload now go over the bus
rather than the legacy socket.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
A cgroup holding processes cannot enable controllers for its children,
so init/ had to stay a leaf. The hotplug helpers 10-hotplug.conf.in
places there ended up in groups where cpu.weight and friends could
never be set.
Keeping PID 1 in the root cgroup makes init/ a domain like the others.
It also unbreaks lxc-based runtimes: liblxc bases the container tree on
PID 1's cgroup and only special-cases systemd's init.scope/, so under
Finit it landed containers in init/, with no controllers available.
Issue #497
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The kernel delegation docs require write access on the directory so
the delegatee can mkdir() children.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The line-based format has had the flag since v4.4 (issue #286), where
it prepends -p to the built-in getty, which turns it into login -p and
passes the environment on. The block format was written from the three
documented tty variants and the flags listed in the tty documentation,
and passenv was in neither, so it was left out. Converting a tty line
that used it therefore lost it, with nothing said.
It only reaches the built-in getty. An external getty is handed its
arguments through command, so there is nowhere to put a -p, and the
setting is refused with a warning rather than quietly ignored.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The migration guide told anyone holding the repeated-stanza idiom for a
per-platform service to split the variants across files or stay on the
line-based format, because a block title is an identity and the
variants have to share one barrier. provides is the answer, so the
guide converts that shape now instead of routing around it, and the
header of 10-hotplug.conf.in no longer points at the workaround.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
A block title is a service identity, so two variants of one service
gated differently per platform cannot share a title. They do need to
share the barrier condition downstream services wait for, which until
now was spelled by the identity alone and so could not be shared:
service syslogd:udev {
if = "udevd"
conditions = { "run/udevadm:5/success" }
provides = "pid/syslogd"
command = "-syslogd -F"
}
Any namespace is allowed, since the point is publishing a name that
existing configurations already wait on. A claim on a condition that
is already owned is dropped with a warning naming the owner, and the
service still registers: the overlap is a configuration bug, and an
init system is more useful degraded than refusing to boot. A real
identity outranks a claim, pid/<ident> is how Finit tracks its own
services, so it is not up for grabs.
Claims are dropped before each reload re-reads the .conf files. Doing
it per service as it re-registers is not enough, since services are
read in file order and one re-registering would lose to a claim
another had not dropped yet, flipping the owner on every reload.
initctl cond dump asked who owned a condition only for the pid/
namespace and printed 'static' for usr/, which now hides a provider.
It asks first and falls back to what the namespace implies.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
A service's condition was computed with mkcond() at each of the six
sites that assert or clear it, and svc_find_by_cond() reimplemented
the reverse lookup a seventh time. maybe_clear_cond() had its own
scan for another service supplying the same condition.
svc_cond_owner() answers who owns a condition, svc_cond_nth() walks
the conditions a service owns, and svc_cond_set()/svc_cond_clear()
apply to all of them. svc_find_by_cond() becomes a wrapper, and
maybe_clear_cond() keeps its rule per condition rather than for the
one it used to compute.
The provides[] storage lands here unused, since svc_cond_nth() reads
num_provides. Nothing sets it yet, so a service still owns exactly
its own pid/<ident> and there is no functional change.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The migration guide covered a stanza at a time, which is the wrong
shape for the two idioms that repeated a whole stanza. One of them,
several candidate binaries for one service, is now a command list.
The other, one service gated differently per platform, has no block
equivalent: those blocks share an identity because they share the
barrier condition downstream services wait for, so they cannot be
given separate titles. For that one the guide says to split the
variants across files, or leave that file in the line-based format,
which Finit still reads.
The udevd example in services.md taught the merge-broken form, and
system/10-hotplug.conf.in pointed readers at it for their syslogd.
Also lists libConfuse among the build dependencies. It has been
mandatory since the new .conf format landed, and build.md still said
two libraries. And corrects the note on variable expansion: it is
${VAR} that libconfuse expands when the file is read, with
${VAR:-default} supported. A plain $VAR reaches the service, which is
what makes `command = "syslogd -F $SYSLOGD_ARGS"` work with envfile.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
A daemon known under more than one name had to be declared once per
name, each stanza carrying nowarn so the ones that were not installed
were skipped quietly. That leaned on the line-based format having no
titles. With a title as the service identity the repetition is no
longer available, so the candidates move inside the block:
command = { "/lib/systemd/systemd-udevd", "-udevd" }
Finit starts the first one whose binary resolves. A candidate that is
not installed is expected here, so nothing is logged for the ones that
lose, and only the winner is looked up again by service_register().
The leading '-' keeps its meaning and is read from the candidate that
wins, or from the last one when none resolve.
libconfuse accepts a bare string for a list option, so the common
`command = "prog args"` is unchanged, and whichp() already skips any
arguments to the command, so the candidate goes to it as written.
tty blocks take the same list. Their command is the getty to run, and
it has the same reason to name alternatives. Both block types now
share svc_command(), which also puts the leading '-' rule back in one
place: tty_translate() had its own copy. A tty needs no command at
all, it may name a device or notty instead, so svc_command() returns
NULL for an empty section and each caller decides whether that is an
error.
system/10-hotplug.conf.in returns to one udevd block, which is what it
meant all along.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The block title is the service identity, and libconfuse merges two
sections that share one, without a word. Two blocks titled the same
in one file therefore loaded as a single service holding a mix of both
declarations, with scalars taken from the last block and lists reset
by it.
system/10-hotplug.conf.in is written this way: two udevd blocks, one
per candidate binary, the way the line-based format spelled a
fallback. Only the second survived the merge, so a system that has
/lib/systemd/systemd-udevd but no udevd got no udevd service at all,
and the whole `if = "udevd"` chain behind it went with it.
CFGF_NO_TITLE_DUPES turns the merge into a parse error naming the file
and the title, and the file is then rejected as a whole. The same
title in another file is untouched, that is how an administrator
overrides a system .conf.
Format detection had to stop agreeing with it. is_new_format() probes
by parsing, so a duplicate title made it answer "not block format" and
conf_parse_file() handed the file to the legacy parser, whose errors
buried the real message. The probe now clears the flag on its own
copy of the option array, keeping the verdict syntactic.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The reload-signal translation calls str2sig() and compares against
SIGHUP, but conf.c never included signal.h. glibc pulls it in
transitively, so the omission went unnoticed until a cross-compile
against uClibc-ng in Buildroot failed to build.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The runparts directory and the dbus pidfile and daemon paths are
embedded in double-quoted values of the generated block files. A
literal quote in either ends the value early and libconfuse rejects
the whole file, and a backslash is read as an escape sequence,
silently mangling the path. The legacy one-liners had no quoting, so
neither failure existed before the block conversion.
conf_escape() doubles backslashes and escapes quotes; verified by
round-tripping hostile paths through cfg_parse_buf().
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Finit's own generated services -- watchdogd, keventd, runparts, and
the dbus plugin -- still went through conf_save_service() as legacy
one-liners, so `initctl show keventd` taught the old format on a
system otherwise converted to the new one.
conf_save_service() now takes the block title and a printf-style body
and writes the file itself:
# Generated by finit:conf_save_service()
service keventd {
description = "Finit kernel event daemon"
runlevel = "S12345789"
notify = "none"
cgroup init {}
command = "/libexec/finit/keventd"
}
vfprintf() into the file also removes the fixed-size staging buffers
in the callers, where a long dbus pidfile path could truncate inside
a quoted string and take the whole generated file with it.
Semantics preserved: watch-only pid:! maps to pidfile without
pidfile-create, the watchdog keeps its watchdog:finit identity, and
log:console becomes log { file = "/dev/console" }.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The README.md symlinks exist for GitHub browsing and collide with
index.md when mkdocs renders both; exclude them like TODO.md.
Two links pointed at anchors that never existed: features.md has bold
captions rather than headings, so "Automatic Reload" gets an explicit
attr_list anchor for the link from the front page, and the TTY link
now spells the actual heading, controlling-tty-for-services.
mkdocs build is silent after this.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The new-format reference describes the block format on its own terms,
which is the wrong lookup direction for someone holding a legacy
one-liner. Aaron migrated Finix OS from the PR description, proving
the need for a token-in, key-out mapping in the user guide.
One table per part of a stanza, worked conversions for the shapes
that changed structurally -- cgroup selection and the three tty
variants -- and the dropped tokens listed with their replacements.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The line-based format accepts `service :80 ...`, deriving the name
from the command basename. The block format has no counterpart, the
title carries both name and ID. Implied by the format description,
but anyone converting such a line deserves to find it written down.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The workflows install libuev and libite from source and everything
else from apt, but never libconfuse, so every build job on this
branch dies in configure:
checking for libconfuse >= 3.3... no
Ubuntu ships libconfuse 3.3 with the static library included, which
covers both the static and regular builds. Staying on 3.3 in CI is
deliberate: it exercises the fallback paths marked
"XXX: Workaround for libConfuse <3.4" that a from-source 3.4 would
leave untested.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Aaron Andersen points out in the #492 discussion that the *Directory
settings carry more contract than create-and-chown: per-directory
modes, specific ownership rules, and cleanup toggles. Without them
config-dir was chowned to the service user, which systemd never does,
an existing directory with drifted ownership was left wrong, and the
runtime directory could not survive a restart.
Now matching systemd.exec(5), and where the man page is vague, the
code in setup_exec_directory():
- each directory takes a matching -mode key, octal with the leading
zero, default 0755. The mode of the named directory is locked
down again on every start, also when it already exists
- config-dir is created but never chowned
- the contents of an existing directory are left alone as long as
the owner is right; on drift everything under it is chowned back
- runtime-dir-preserve = no | restart | yes maps
RuntimeDirectoryPreserve=. A service still qualified to run when
the runtime directory would be removed is restarting, not
stopping, which is what svc_enabled() answers
The dir mechanics move to mksubsysd(), taking resolved ids, with
mksubsys() reduced to a name-resolving wrapper for the dbus plugin.
The child resolves uid/gid once for both directory setup and
privilege drop.
The symlink form, RuntimeDirectory=foo:bar, is not adopted.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
A service that drops privileges cannot create its own PID file in
/run, root owns it. Finit can create the file with pidfile-create,
but the daemon still cannot touch it to confirm a SIGHUP.
Five new settings, block format only: runtime-dir, state-dir,
cache-dir, logs-dir, and config-dir. The value is a directory name,
resolved under /run, /var/lib, /var/cache, /var/log, and /etc,
respectively. The directory is created before the service starts,
mode 0755 owned by user/group, and the full path is exported to the
process as RUNTIME_DIRECTORY, STATE_DIRECTORY, CACHE_DIRECTORY,
LOGS_DIRECTORY, and CONFIGURATION_DIRECTORY. Mode and ownership are
asserted at creation only, a daemon may tighten them afterwards.
The runtime directory is removed when the unit stops, after any
exec-stop-post script, like systemd with RuntimeDirectoryPreserve=no.
A completed run/task counts as stopped unless remain-after-exit keeps
it up. The other four persist across restarts.
These are the first settings with no legacy token: they are validated
by service_set_dir() and stored on the svc that service_register()
now returns. systemd accepts a list of directories per setting; this
is a single name for now, widening later is compatible since
libconfuse accepts a bare value for a list option.
The test sysroot gains libnss_files.so.2, which ldd cannot see, glibc
dlopen()s it. Without it getpwnam() fails inside the chroot, so
user/group settings never resolved and directory ownership could not
be tested.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
rmrf() is needed outside tmpfiles.c. The move also deduplicates the
nftw callback: the contents-only removal used by tmpfiles 'D' entries
is now rmcontents(), sharing the callback with rmrf().
mksubsys() did nothing at all when the user could not be resolved, no
directory and no message, and callers had no way to tell. Now the
directory is always created, ownership is best effort, and an unknown
user is warned about.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Settings that exist only in the block format have nowhere to go: the
legacy line cannot carry them, and service_register() returned an errno
that no caller ever read, so conf.c had no handle on the service it just
created. Return the svc instead, NULL with errno set on failure, errno
zero when a block is skipped on purpose.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The block format spells conditions as bare strings everywhere else, so
requiring `if = "<usr/foo>"` left one sigil behind, carried over from
the line-based `if:` token. A namespace separator already tells the two
apart: a value with a '/' is a condition, anything else is a service
name.
svc_ifthen() picks its mode from the start of the statement and applies
it to the whole, so a statement naming both kinds cannot be evaluated.
That is now an error, as are the old angle brackets, and either one
skips the block:
/etc/finit.conf: mixed: if: cannot mix a service name with a
condition in 'anchor,usr/enable-me', a statement must be all of
one kind, skipping
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Reference sections kept pointing at the line-based format they no longer
document. `sysv` and `task` sent the reader to Services for "<COND>",
the cgroups chapter opened by listing three legacy directives and then
explained further down that only two of them exist here, and the logging
chapter still gave "log:prio:facility.level,tag:ident" as the full
syntax.
Some claims were wrong independent of the format:
- a sysv is a supervised daemon, grouped with service in
SVC_TYPE_DAEMON, not a variation on task
- restart-max has no upper bound of 255, or any other
- the built-in rescue fallback runs in 12345789, not 12345
- conditional loading quotes system/10-hotplug.conf, not
system/hotplug.conf
- the key spells conflicts, not conflict
- the built-in getty no longer wants TERM last, it is a key
`if` takes either a service name or, in angle brackets, a condition,
decided in svc_ifthen(). Only the examples showed this, so it is now
said.
Terminology follows the split index.md already draws: a block is the new
format, a stanza the line-based one.
src/rescue.conf was still line-based, missed because it sits in src/
rather than system/ or contrib/.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
An ambient capability only reaches the effective set when euid is
non-zero, so a service that pairs `capabilities = { "^cap_..." }` with a
root user gets none of the restriction it asks for, and keeps the full
root set instead. Finit read the list, applied it, and said nothing. A
build without libcap dropped the list on the floor just as quietly.
Both now warn, naming the service:
nginx: ambient capabilities ('^') have no effect as root, use a
non-root user, or '%' and '!' entries
The ambient entries are read back from the parsed IAB value rather than
matched in the text, so inheritable ('%') and bounding ('!') entries stay
silent -- those work fine as root.
The warning repeats when the .conf files are re-read on runlevel change,
as parse warnings here already do.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The block conversion changed the bodies of the reference sections but
left every "**Syntax:**" header spelling the line-based format, so each
page opened by teaching the format it then stopped using. Six files
were missed entirely: runparts, files, capabilities, requirements,
runlevels, and switchroot.
runparts had no block spelling written down anywhere, though the parser
has read `runparts`, `runparts-progress`, and `runparts-sysv` all along.
tty gains a table per variant. Its three syntax lines carried nine
positional fields between them, which no longer describes anything the
parser accepts.
Fixes#148
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The examples people copy from were still written in the line-based
format, so the block format was documented but nowhere demonstrated.
Two names in contrib were accidents of the old format, where the
service name falls out of the command basename: the Alpine and Void
keymap task was called zcat, and Debian's console/keyboard setup tasks
carried a .sh suffix. They now carry the name their file implies.
Nothing referenced the old names.
The mdevd coldplug path keeps the name it has always had. Its legacy
line spelled the name inside the cgroup argument, where it names the
cgroup leaf and not the service, so the barrier condition really is
<run/mdevd-coldplug/success> and not the <run/coldplug/success> the
comment above it promises. Converted as-is so boot ordering does not
change; the discrepancy is now written down where it happens.
A list may not contain comments, the lexer sees the entries after the
'#' regardless:
modules = {
# "fbcon",
"softdog"
}
so the commented-out module candidates sit above the list instead.
setup-sysroot.sh removes 10-hotplug.conf from the test sysroot, so that
file is covered by parsing only, not by make check.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The syntax overview no longer describes a line-based format, since that
is not what the rest of the documentation shows. It now covers the
grammar, the two naming conventions, the nine aliases, and the leading
'-' on a path, and it says plainly that both formats are still read and
told apart per file by content. Without that, a reader with an
existing configuration is left wondering what happened to it.
service-opts.md was a list of modifiers to place between a directive
and its command, so it needed rewriting rather than translating: there
are no positions left to describe. It is now grouped by what the
settings do.
conditions.md needed correcting. It presented '!' as a condition
prefix alongside '~'. It is neither a condition nor a negation, it is
a flag on the block that means one thing on a service and another on a
run or task, so it is spelled reload-signal and required here, and the
page maps the old form to both.
Two things the pages claimed are not true. The kill delay range is
1-300, not 1-60, and stop and reload scripts are no longer run without
a timeout.
ChangeLog.md keeps its line-based examples. Those sit in historical
release entries, and rewriting them in a syntax that did not exist at
the time would misdate the format.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
parse_cgroup() takes two arguments that are not cgroupfs files: the
leaf directory to place the service in, and whether to hand the subtree
over to it. The block format could express neither. 'name' happened
to work, because a free-form key is emitted as name:VALUE and that is
what the parser looks for, but 'delegate' came out as delegate:true and
was filed as a cgroup setting, so it silently did nothing.
Declare both, and emit delegate as the bare flag the parser expects.
Neither means anything on a top-level group definition, so say so there
rather than emitting something that would be written to cgroupfs.
service podman {
cgroup containers { name = "podman" delegate = true }
...
}
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This is what 'initctl create' and 'initctl edit -c' put in front of a
user writing their first .conf file, so it is also the whole of the
"initctl emits the new format" work: neither command generates syntax,
they copy this file and open an editor on it.
The ASCII diagram naming eight positional fields goes with it. A
block has no positions to explain.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Both keys prompt the question they should be answering.
'remain' decides whether a finished run or task keeps existing: without
it the entry is pruned, so the work re-runs on every runlevel entry,
initctl cannot see it, and its post script never fires. With it the
entry stays, is not re-run, and gets a teardown when stopped or when it
leaves its runlevels. That is systemd's RemainAfterExit, and 'remain'
is that name with the informative half cut off.
'manual' says how a service is started but not that it is about
starting at all.
remain -> remain-after-exit
manual -> manual-start
Both keep their old spelling as an alias, which they qualify for twice
over, as abbreviations of the canonical name and as the legacy
spellings.
While here, give sec_getbool() the alias argument its string and list
counterparts already take.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
A condition list could be led by '!', which is not a condition and not
a negation. It is a flag on the block, and it means two unrelated
things depending on which block it sits in: a service or sysv does not
handle SIGHUP and must be restarted to reload, while a run or task
must not hold up bootstrap. Writing '<!>' with no condition at all is
legal, which gives away that it was never an operator.
Give each meaning its own key, valid only where it applies:
service foo { reload-signal = "none" } # restart to reload
task bar { required = false } # do not hold up bootstrap
Using either on a block type it does not apply to warns, as does a '!'
left in a conditions list. Both still translate to that same '!',
which is all a legacy line can carry, so reload-signal takes SIGHUP or
none for now; str2sig() already accepts any case and an optional SIG
prefix.
This also clears the way for the conditions list to grow real
operators, '+' and '-' for asserted and deasserted, without '!'
sitting among them meaning something else entirely.
The '~' prefix stays. It belongs to the list: it marks a dependency
whose reload should propagate here.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The initial implementation was done using a naive translator of the
legacy one-liner format key by key, so it inherited encodings that the
block format exists to remove: a timeout packed into a script path, a
small comma-and-colon language inside the log string, sigils standing in
for booleans, and a log key (services) carrying three (!) types.
Settled naming against systemd, OpenRC, FreeBSD rc.subr, s6 and SMF.
Match systemd's semantics, not its naming.
pid -> pidfile, plus pidfile-create for the rare case
where Finit writes the file rather than the daemon
environment -> envfile, since it names a file to source, and the
top-level environment {} block sets variables
pre/post/... -> exec-start-pre, exec-start-ready, exec-stop,
exec-stop-post, exec-reload, exec-cleanup, each
with its own -timeout instead of "SEC,script"
halt, kill -> stop-signal, stop-timeout
restart -> restart for the policy, restart-max for the count
log -> a block with file, priority and identity, where
/dev/null and /dev/console are spelled as paths
group -> group and extra-groups, no longer positional
nowarn -> a leading - on command, as on envfile
List-valued keys take plural names. Aliases are desc, cond, mod,
caps, env, halt and kill; an alias may abbreviate the canonical name
or preserve a legacy spelling, nothing else.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
'make check' refreshes the sysroot through the setup-chroot rule, but
running a test script by hand does not, so the test exercises whichever
finit was installed last and reports on code that is no longer there.
Both a passing and a failing run are then meaningless, and nothing says
so.
Compare the built binary against the installed one at startup and fail
with the command that fixes it.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Both were bounded by killdelay, the delay between the stop signal and
SIGKILL, because service_run_script() had nothing else to reach for.
That conflates two things: how long the daemon may take to die, and
how long its stop script may run.
Give each hook a timeout of its own, defaulting to killdelay when
unset, so the existing behaviour is what you get until you ask for
something else. parse_script() already falls back that way for the
hooks that had one.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
A stop: or reload: script written with a timeout killed Finit at
config load:
service stop:5,/bin/true service.sh -- Boom
parse_script() takes the timeout as a pointer and the caller decides
whether it wants one. However, both stop: and reload: scripts so far
have no timeout, i.e., NULL. Guard the branch that reads a leading
number.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Three private ones had grown: fnread() in util.c, flen() behind
pid_cmdline()/pid_cgroup() in cgutil.c, and conf_read_template() in
conf.c. Two of them were also wrong in ways the others were not.
fnread() formatted the path into a char[256] and stat()ed it before
opening, so a longer path was silently truncated and then read from
whichever file the truncation happened to name, and the size could
change between the look and the read. flen() existed because neither
of those approaches works on procfs at all, where stat() reports zero
and the only way to learn the size is to read to EOF.
Add fslurp() to util.[ch], which every tool already links. It opens
first and sizes the fd it holds, treats st_size as a hint, and reads
until EOF, so procfs and regular files take the same path. Paths are
formatted by libite's vfopenf(), which allocates to fit. Callers that
need the byte count, /proc/PID/cmdline embeds NUL, ask for it.
fnread() keeps its signature and becomes a bounded copy out of the
result, so its one caller is unaffected.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
A template in the block format registered garbage. conf_parse_file()
routed every file with an '@' in its name straight to the legacy
parser, which read the block line by line: the section header became a
service whose command was the section title, and each key = value line
below it became an environment variable.
service serv:%i { ... } -> service 'serv:eth0' with argument '{'
Substitute %i over the whole file before parsing instead, so format
detection and both parsers see finished text. A bare name@.conf is
still skipped, it is the template rather than an instance of one.
The legacy parser no longer opens the file or substitutes per line, it
is handed the instantiated buffer, so the template convention now has
one implementation instead of two. conf_is_template() applies
basenm(), a directory with an '@' in its name is not a template.
libconfuse cannot name a buffer it parses before 3.4, so a typo in a
template would be reported against "[buf]". Parse through fmemopen()
with the file name preset until the floor moves.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Relocate process-wide global variables from legacy parser that ended up
there because it used to be conf.c, but which is now now frozen at the
4.x feature set. Each variable is moved to their respective "owner".
Give cgroup_current[] and cgroup_settings_current[] named bounds. Their
extern declarations were unsized, so sizeof() on them stopped compiling
once the definitions moved to another translation unit.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The one-liner format has grown crowded and very wide, and every new
service option makes it worse.
Add a second, block-based format, parsed with libconfuse:
service sshd {
description = "OpenSSH daemon"
runlevel = "2345"
command = "/usr/sbin/sshd -D $SSHD_OPTS"
}
Both formats keep the .conf extension and are detected per file by
content. Try-parse strictly with libconfuse; on a parse error,
re-parse leniently to tell a block file with a typo from a one-liner
file. Only a one-liner file reaches the legacy parser, a typo is
reported with its file and line.
Each block is translated to the canonical one-liner and registered
through the existing entry points, so the two formats cannot drift.
The one-liner parser is frozen at the 4.x feature set, new options
land only in the block schema. libconfuse 3.3 or later is required,
CFGF_KEYSTRVAL does not exist before it.
Covers service, task, run, sysv and tty blocks, the static directives,
and the cgroup, rlimit, set and log blocks. Templating and the
documentation rewrite are still to come.
The regression test covers translation of a service block to the
one-liner, a block-format /etc/finit.conf booting with set {} applied
at bootstrap, both formats side by side, and rejection of a typo at
block and at root level.
A rejected file must not fall through to the legacy parser, which
registers a bogus unstartable service per line. assert_num_children
cannot see that, the bogus service has no children either, so the
check is assert_num_services.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The one-liner parser is about to be joined by a second, block-based
format. Give it a name that says which of the two it implements,
before any content changes make the diff hard to follow.
No functional change.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Every test left a stray `sleep 300` behind, reparented to PID 1, where
it lingered for up to five minutes after the test had finished.
wdstart() runs the watchdog in a subshell, so $! is the pid of the
subshell, not of the sleep it forks. wdkill() killed the subshell and
orphaned the sleep.
Kill the child first, killing the subshell puts the sleep beyond the
reach of pkill -P. Neither kill is sure to match, and wdkill() runs
from the EXIT trap under set -e, so both must tolerate failure. Also
return early when wdpid is unset, for failures before wdstart() runs.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Finit 5.0 changes the .conf syntax, which has been essentially
unchanged since 1.x. The published docs track master, so when 5.x
lands, 4.x users lose their reference.
Publish the site under a per-major directory, /4.x/ for now, with
the Material version selector to switch between them. The selector
only needs mike's file layout -- a versions.json at the site root --
which the deploy job now generates from the version directories in
the pages repo, so mike itself is not needed.
The major comes from AC_INIT and the future 4.x maintenance branch
is already in the workflow triggers, so once 5.0 is on master, doc
fixes on the 4.x branch keep /4.x/ updated. A root index.html
redirects to the newest version, and a 404.html rewrites
pre-versioned deep links so old bookmarks and search hits land in
the right place.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Both the Finit project and Infix use the same MkDocs Material setup, and
in the latter the User Guide has picked up a lot of polish that never
made it back here: a single sidebar with section indexes instead of
tabs, footnote tooltips, more pymdownx markup, image zoom tuning, and no
generator advert in the footer.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
* src/pid.c: note the stale-pidfile-cleanup exception to the
documented "Finit does not touch pid:! pidfiles" rule.
* doc/config/services.md: add a user-facing paragraph on the same.
* doc/ChangeLog.md: add Unreleased section covering this PR --
stale pidfile cleanup, restart log with signal name and core
dump flag, and the SIGUNKOWN typo fix.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Cover the scenario fixed in "service: clean stale pidfile after
unclean daemon exit": a daemon with a pid:!/path config dies via
SIGKILL, leaving its pidfile behind, and the next instance must
still come up.
Add a 'serv -x' flag (refuse to start when the pidfile already
exists, dbus-style) so the test actually exercises the cleanup --
without it, plain 'serv' would happily overwrite the file and the
test would pass with or without the fix.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The fallback for unknown signal numbers in sig_name() returned the
misspelled "SIGUNKOWN". Now that this string surfaces in user-
facing logs ("killed by …"), fix the typo.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Replace the bare signal number ("by signal: 9") with the symbolic
name ("killed by SIGKILL") and annotate when the kernel wrote a
core:("killed by SIGSEGV, core dumped"). Makes the restart line
self-explanatory and gives operators a strong breadcrumb when a
daemon dies unexpectedly.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
With `pid:!/path` Finit does not manage the file -- the daemon
creates it on start and removes it on graceful exit. If the daemon
dies before cleanup (SIGKILL, OOM, segfault, exit during startup)
the file lingers and can block the next instance from starting,
e.g. dbus-daemon refuses with EEXIST and the restart loop fails.
Remove the file when it still names the just-reaped PID and that
PID is no longer alive (the liveness check guards against reuse).
Called from service_cleanup(), and from service_monitor()'s
forking+starting branch where cleanup was previously skipped.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
status() returns a pointer to a single static buffer, so calling it twice
in the same cprintf() argument list — status(3) and status(rc) — causes
one to overwrite the other before the format string is rendered. When
status(3) wins, the line shows [ ⋯ ] instead of [ OK ]. Fix by copying
status(rc) into a local buffer before calling status(3).
Also drop the delline() calls added to print() — that macro writes \033[2K
to buffered stdout while cprintf() writes unbuffered to stderr, so the
erase sequences can arrive out of order. The \r\e[K already present in
the cprintf format strings makes them redundant anyway.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Check return value of remove() in delete_cb() and log failures via
dbg(), CID 909395
Replace stat() calls with open(O_DIRECTORY)+ fstat() for newroot and "/"
checks. Eliminates the check-then-use race and lets O_DIRECTORY do the
isdir validation atomically, CID 909394
Drop the explicit close(0/1/2) before opening /dev/console. dup2()
closes the old targets itself, so open() returns a fd > STDERR_FILENO
that can always be closed unconditionally, removing the conditional
guard, CID 909393
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Refactor print() to emit description + final status in a single call to
cprintf(), preventing kernel messages from splitting the two parts.
For two-phase print(-1,...) + print_result() sequences used by, e.g.,
run_interactive, save the last description and re-print it before the
[ OK ] / [FAIL] output so the status is never left stranded on a blank
line when command output or kernel messages have scrolled away the
original description.
Finally, add print_exit() which drains the console output buffer with
tcdrain(2) and resets ANSI SGR attributes + cursor visibility before the
kernel takes back the console on reboot/halt, preventing escape code
leakage into bootloader or early-kernel output.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
When a service without SIGHUP reload support (noreload) is touched and
'initctl reload' is called, service_update_rdeps() correctly identifies
its reverse dependencies but only marks them dirty. It does not clear
the service's condition, so when service_step_all() runs:
- rdeps supporting SIGHUP hit the sm_in_reload() guard and break early,
left running while their dependency is being killed.
- rdeps without SIGHUP support may receive SIGTERM too late, after the
dependency has already died and broken their connection, causing them
to exit from RUNNING state and have their restart counter incremented.
Fix by calling cond_clear() on the service's condition immediately in
service_update_rdeps(), before service_step_all() runs. cond_clear()
calls cond_update() which calls service_step() inline on all affected
services, which see COND_OFF and transition to STOPPING_STATE — all
before SIGTERM is ever sent to the dependency itself.
This mirrors the pattern already used in api.c:do_reload() for direct
'initctl reload <svc>' calls.
Fixes: avahi/mdns stop causing mdns-alias restart counter increment
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Add support for the --exclude-prefix=PATH option to skip rules whose
path starts with the specified prefix. The option can be specified
multiple times to exclude multiple path prefixes.
The -E flag is a shortcut for:
--exclude-prefix=/dev --exclude-prefix=/proc \
--exclude-prefix=/run --exclude-prefix=/sys
This is useful to avoid creating files below virtual or memory-backed
file system mount points.
devmon: assert condition immediately if device already exists
1. udev fires early, creates device nodes in /dev/
2. Config is parsed later, calling devmon_add_cond() for each dev/ condition
3. At this point the device already exists but the inotify event was missed
4. The PR's fexist() check catches this — new node is added to the TAILQ and condition is immediately asserted
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The config parsing happens after udev triggers the initial event,
make sure to assert the condition if the device node exists when adding
it from configuration.
Signed-off-by: Mattias Walström <lazzer@gmail.com>
A service that is reloaded should not trigger dependants to be reloaded
unless the new <~cond> is used. Which is reserve for tightly coupled
services.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
For SysV services with pid:!/path, the pidfile belongs to the service
itself and Finit shouldn't delete it.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This fixes a real bug where `initctl reload syslogd` unconditionally
clears syslogd's pid condition, causing all dependent services (dbus,
dnsmasq, etc.) to be stopped even though syslogd handles SIGHUP
gracefully and its PID/pidfile persist.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
A reload: script, like 'frrinit.sh reload' could potentially take a
while to finish, during which Finit would be blocked. This change
reuses the service_script_add(), used for ready: scripts, to track
these background helpers.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
When the kernel manages to reap a child process before we've moved it to
its proper cgroup it will return ESRCH (No such process), we can safely
ignore such errors for short-lived processes.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Conditions in Finit are dependencies: if A is asserted, service B is
allowed to run. When A goes through FLUX (e.g., upstream reloads),
dependents are PAUSED and then simply resumed when the condition is
reasserted -- this is the correct behavior for barrier-style deps
like <pid/syslogd>.
However, some setups have tightly coupled services where dependents
must be reloaded/restarted when an upstream service reloads, not just
resumed. E.g., the FRR routing stack on Infix OS:
netd <pid/mgmtd> ← zebra <!pid/netd> ← {staticd,ripd} <!pid/zebra>
When netd reloads (SIGHUP), zebra and its dependents must be restarted
to pick up the new configuration.
The new '~' condition prefix marks a dependency as flux-sensitive:
service <!~pid/netd> name:zebra ...
When the upstream condition goes FLUX and returns to ON, the dependent
is reloaded (SIGHUP) or restarted (noreload '!') instead of merely
resumed. Transitivity follows naturally through the condition chain.
Closes#416Closes#476
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
When 'initctl reload' is called after marking a service in a dependency
chain dirty, Finit fails to restart (unfreeze) affected services.
This patch updates the pidfile plugin to watch for IN_ATTRIB changes,
e.g. when a process uses utimensat() to update its pidfile, and adds
service_step_all() at end of reload cycle to guarantee convergence
after conditions are reasserted.
Issue #476
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
In a setup like this, when 'netd' is marked dirty and subsequently is
reloaded, e.g., using 'initctl reload', zebra is properly restarted,
but staticd isn't:
mgmtd <!> ← netd <pid/mgmtd> ← zebra <!pid/netd> ← staticd <!pid/zebra>
Finit must invalidate the condition of zebra to trigger a restart also
of staticd. This to guard against daemons like zebra that may fail to
clean up their pidfiles.
Fixes#475
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Device conditions tracked by devmon were lost on `initctl reload`
because the reconf path did not re-assert them. Add devmon_reconf()
to iterate all tracked device nodes and set or clear their conditions
based on current device presence.
Signed-off-by: Mattias Walström <lazzer@gmail.com>
Verify that 'initctl reload foo' properly triggers dependent
services by checking that bar gets a new PID after the reload.
Also change the second test case from service/foo/running to
service/foo/ready which is the actual condition set by pidfile.so.
Fix a race in slay where the target process could exit between
the PID lookup and kill -9, causing spurious test failures in
tight kill loops (e.g., start-kill-service.sh).
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
When reloading a specific service with 'initctl reload foo', the
pid/foo and service/foo/ready conditions were never cleared, so
dependent services were not notified of the reload.
Clear the service's pid condition and, for pid/none notify types,
the ready condition before reloading. The conditions are then
reasserted by the pidfile inotify handler when the service touches
its PID file after processing SIGHUP.
For s6/systemd services the ready condition is left intact since
their readiness notification may not re-trigger on SIGHUP.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Users starting Finit based systems using U-Boot or Barebox may otherwise
not get a visible cursor at their prompt.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Similar to systemd's RemainAfterExit=yes. Prevents the task from
re-running on runlevel re-entry and ensures the post: script runs
when explicitly stopped or when leaving valid runlevels.
Useful for tasks that set up persistent state like firewall rules:
task [2345] remain:yes \
post:/usr/sbin/teardown-firewall \
/usr/sbin/setup-firewall -- Firewall setup
Not supported for bootstrap-only tasks (runlevel S only) since these
are deleted immediately after completion.
Define __NR_clone3 (435) ourselves when not provided by the toolchain
headers. The syscall number is stable kernel ABI and the same on all
architectures since Linux 5.3.
The existing runtime fallback to fork() handles older kernels that don't
support the syscall.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
In containerized or virtualized environments, standard mount point
directories may not exist at boot. Ensure they are created before
attempting to mount.
When f or F types write content to a file, the mode and ownership
specified in the config should be applied. Previously, ownership was
only applied when create() was used (i.e., when no argument was
specified).
Now we explicitly apply mode and ownership after writing content to
the file.
According to tmpfiles.d(5), the 'e' type adjusts the mode and ownership
of existing paths but should not create them. Previously, mksubsys()
was used which could create directories.
Now we explicitly check if the path is an existing directory before
adjusting its permissions.
Add parse_uid() and parse_gid() helper functions that support both
numeric IDs and name lookups. Update the d/D directory creation
handlers to use these new functions.
This allows config files to specify ownership using numeric UIDs and
GIDs instead of only usernames and group names, matching systemd-tmpfiles
behavior.
The L+ type should replace existing entries with a symlink. Previously,
rmrf() was always called which is only appropriate for directories.
Now we check if the path is a directory first, and use erase() for
files and symlinks.
Allow specifying one or more configuration files as command line
arguments instead of always processing all files in the standard
tmpfiles.d directories.
This enables targeted operations on specific config files:
tmpfiles --create /etc/tmpfiles.d/myapp.conf
tmpfiles --clean /tmp/test.conf /tmp/other.conf
When no config files are specified, the existing behavior of
processing all *.conf files in the standard directories is preserved.
Also refactors file processing into a helper function to reduce
code duplication.
Add support for the --clean (-C) flag to remove files and directories
older than the age specified in tmpfiles.d configuration entries.
The age field (6th column) in tmpfiles.d entries can now be used with
'd', 'D', and 'e' type entries to clean up old files. Supported time
suffixes are: s (seconds), m (minutes), h (hours), d (days), w (weeks).
Example configuration:
d /tmp/cache 0755 root root 10d
When run with --clean, files in /tmp/cache older than 10 days will be
removed. The directory itself is preserved.
Uses a conservative cleanup approach matching systemd-tmpfiles:
- Files: kept if ANY of atime, ctime, mtime is recent
- Directories: kept if ANY of atime, mtime is recent (ctime excluded
because cleanup itself updates directory ctime)
A value of "-" or "0" for age disables cleanup for that entry.
Note: x/X exclusion patterns are recognized but not yet implemented.
Strings like command, description, and environment may contain characters
that need escaping for valid JSON, e.g., embedded quotes in command line
arguments like -V "NanoPi R2S".
Add json_escape() helper to handle quotes, backslashes, and control chars.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The netlink plugin only receives RTM_NEWLINK events for interfaces that
appear after the plugin starts. Interfaces that already exist at boot
(e.g., virtio-net in QEMU) never generate events, so their conditions
like net/eth0/exist were never set.
Moving enumeration to PLUGIN_INIT doesn't work because it runs before
cond_init(), so the condition filesystem isn't ready yet.
Fix by registering an HOOK_SVC_PLUGIN callback that queries existing
interfaces and routes. This hook runs during conf_init(), after the
condition system is initialized.
When a TTY exited with non-zero code (e.g., user with shell=/sbin/false),
it would enter restart state but never recover, requiring manual restart.
The throttling logic from commit f0032ab had two issues:
1. Duplicate exit code check in service_retry() created infinite timer loop
2. TTYs lacked default restart_tmo, causing timer to never start
Fix by removing duplicate check and ensuring TTYs get a 2-second default
restart_tmo for proper throttling.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
When a service is configured to run as a non-root user (@user), finit
correctly drops privileges via setuid() and sets HOME and PATH, but
does not set the USER and LOGNAME environment variables. They remain
set to "root" from boot time.
This causes problems for software that determines its identity from
the environment rather than getuid(). For example, rootless Podman
checks os.Getenv("USER") first when looking up subordinate UID/GID
ranges in /etc/subuid and /etc/subgid.
With USER=root but UID=1000, Podman looks up root's subuid entry
instead of the actual user's, causing applications like newuidmap
to fail. Setting USER and LOGNAME to match the actual user identity
follows POSIX conventions and matches the behavior of su, sudo, and
login.
The do_ prefix is conventionally reserved for local helper functions.
Move switch_root() declaration to private.h alongside iterate_proc()
and remove the now-empty initramfs.h header.
in an initramfs, then transition to the real root filesystem. Useful
for systems requiring early boot tasks like LUKS unlock, LVM activation,
or network boot before mounting the real root.
Adds INIT_CMD_SWITCH_ROOT API command, `initctl switch-root` subcommand,
and HOOK_SWITCH_ROOT plugin hook point. The implementation gracefully
stops services, moves virtual filesystems (/dev, /proc, /sys, /run) to
the new root, deletes initramfs contents to free memory, then execs the
new init as PID 1.
See GitHub Discussion #292 for background.
Add handling for EBUSY when mounting cgroup2 filesystem, which occurs
when cgroups are already mounted. This can happen after switch_root
when cgroups were moved from the initramfs, or in container environments.
Verify the existing mount is actually cgroup2 before proceeding, and
track whether we mounted to avoid unmounting on error if we didn't.
At least the sysvpart.sh regression test cannot run in parallel yet with
other tests (probably runparts.sh), so we must ensure the tests never
run in parallel, in particular at release.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Finit now requires being able to query at least for the root user and
group before starting any services.
Also, add support for using libraries installed in /usr/local
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Finit has support for "Please press Enter to activate this console."
which means there's no getty yet running. However, when profiling
systems with Finit, and embedded systems in general, a common metric
is the time from power-on to getty has started.
This commit makes sure to rename the process so that BusyBox pidof is
capable of detecting that "getty" has started. This is mostly for the
bootchart2 project's bootchartd, the native BusyBox bootchartd does not
have this issue.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This is a refactor of getuser() and getgroup() so that they always
return a valid user, and group, for all normal use-cases. When an
error occurs we now handle it properly in service_fork() so as to
not attempt to start services with an invalid user/group setting
as root.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
When running Finit under boothcartd (bootchart2 project) the PATH is
lost due to a bug. This was a wakeup, so set critical variables in
main() early, before calling fs_init().
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Try out navigation.tabs, if we don't like it we can revert.
Reorg for stricter sections, what information actually belongs where?
E.g., introduction is now in a Getting Started section.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Some 'respawn' type services, like gettys, may hog the CPU in error
states if the service immediately exits. E.g., due to missing dev.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
During /bin/login phase the TTY device node is chowned and chmodded to
the authenticated user. It will remain in this state until the next
call to getty.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
A user reports inability to re-start getty on the console after logging
out from a serial console. After some digging it was found that the tty
was owned by the last user logged in and 600. Even thougn getty runs as
root, it did not have permission to re-open the device node.
Turns out there was a minor bug in the new capability code that cleared
all capabilities from the root user. A surprising amount of programs
worked just fine, but restarting getty gave it away.
The fix is to only call cap_setuid() when capabilities are set for the
service, otherwise we just fall back to setuid().
Also, refactor service_register() wrt. capabilities a bit so that we can
give users an early warning if the configuration is invalid, by adding a
parse_caps() helper function that calls cap_iab_from_text() to verify.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Implement supplementary group support for services, allowing them to
access resources owned by multiple groups. Uses the @user:group,sup1,sup2
syntax to explicitly specify supplementary groups, in addition to now
reading group membership from /etc/group.
Cgroups v2 limits are hierarchical - a process is constrained by the
most restrictive limit in its ancestor chain, not just its immediate
cgroup. This patch updates cg_conf() to walk up the hierarchy and
report effective limits by comparing values at each level.
This fixes incorrect "max" (unlimited) reporting in 'initctl --json
status', 'initctl cgroup', and 'initctl top' when child cgroups have
no explicit limits but parents do.
For memory.max and cpu.max: take minimum (most restrictive)
For memory.min: take maximum (most protection)
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Comment-out code that makes the cursor "jump" around at boot before
displaying: Please press Enter to activate this console.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
libite v2.6.2 is not yet in Buildroot, so let's relax the dependency a
bit. Load bearing functionality was in v2.6.0, any fixes on top is a
nice-to-have only.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
We now require readsnf() introduced in libite 2.6.0, with bug fixes
this effectively means v2.6.2.
The libuev bump is for 64-bit time_t, with bug fix => v2.4.1
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Fixes an issue where the mouse scroll wheel and Shift+PgUp/PgDn
sometimes would not work properly after login.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Drop clear screen to fix flickering in 'initctl top' output. Also, make
sure to not garble the display if the the terminal is too small.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Instead of /system/10-hotplug/ we now place udevd, mdevd, and others, in
more aptly named groups using the new 'name:' syntax.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This should not be needed, but for some reason we don't get events when
early processes exit, so we end up with lingering cgroups.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The container monitor that podman forks off when starting a container
instance creates subgroups in the cgroup v2 hieararchy that we want to
reuse. This patch adds cgroup_move_svc() which we call from the pidfile
plugins to relocate the conmon process.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This commit activates the use of clone3() for service_fork(), to allow
Linux to create the new process directly in the correct cgroup instead
of later moving it there -- much cheaper and less error prone.
To facilitate this a few new helper cgroup functions have been added and
two new configuration directives introduced: delegate and name:leafname.
The delegate option is for running, e.g., container runtimes that want
to create their own cgroup v2 structur, and the name:leafname allows a
user to change the name of the subgroup under user/system/init.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The PTY approach caused isatty() to return true for services using
the log directive, triggering programs like fprintd (using glib) to
emit ANSI escape codes and other TTY-specific formatting in syslog.
Using a standard pipe ensures isatty() correctly returns false, so
programs produce plain text output suitable for logging.
For services that require line-buffered output, users can wrap the
command with `stdbuf -oL` as documented in doc/config/logging.md.
Fixes#455
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
When setting up a new system with Finit it is very common to make small
logical mistakes that cause "hangs" at boot. This is when Finit waits
for 180 sec. for run/tasks to complete before moving to the configured
runlevel. This patch adds console input monitoring during bootstrap
wait that allows users to press Ctrl-C to skip waiting and proceed to
the configured runlevel.
When Ctrl-C is detected, all incomplete run/task/services are logged
to syslog for later troubleshooting after login.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Relax the constraints introduced in a39ee0b, for issue #342, a bit on
when start/stop/restart/reload service can be called. Also, allow
'initctl reload', but ignore it when the system is in runlevels S/0/6.
This makes it possible to start manual:yes type services at botostrap,
for example, which has been a common feature request.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Add missing memory information to JSON output in json_status_one().
The memory field shows cgroup memory usage in bytes when cgroup
support is available and the service has a valid PID, matching
the behavior of the text status output.
Signed-off-by: Richard Alpe <richard@bit42.se>
Implement Linux capability support for services, allowing them to run
with minimal required privileges instead of running as root. This uses
the modern IAB (Inheritable, Ambient, Bounding) API from libcap.
There are some invalid condition checks in rtc.c, when rtc_file is
missing, it should not consider that as a error and return, otherwise,
it can lead to the case that time_set wont be called at all.
With this fix, when both RTC and file restore fail, it will fall back
to call time_set(NULL) and restore time from rtc_timestamp, this
ensures the OS has a valid time at the very first boot.
Signed-off-by: Ming Liu <liu.ming50@gmail.com>
Allow service IDs up to 64 characters to support SHA-256 hashes,
UUIDs, and other long unique identifiers. This increases memory
usage by ~98 bytes per service instance, which is negligible for
typical deployments.
The IDENT column width in initctl output adapts dynamically based
on actual ID lengths in use, so short IDs remain unaffected.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Both service and sysv stanzas support a cleanup:script (since v4.10), with an
optional timeout. These timers must be stopped when collecting a PID for an
svc_t, or when unregistering an svc_t, otherwise the timer callback will be
continuously called, reporting "spurious problem", and cause 100% CPU usage.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Many packages fail to detect that this replacement libsystemd does not
support the logging API (yet), so let's disable it by default.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This patch fixes an ordering issue where the configuration generation is
stepped before all old/disabled services had been stopped. Finit should
ensure disabled services are stopped before entering the next generation.
If the genation is stepped first, all conditions are put in "flux" state,
so, e.g, a podman container running as a sysv service may lock-up until
it is killed by Finit. Found in the Infix OS project.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
A following compilation error was observed:
| libsystemd/sd-daemon.c:64: undefined reference to `strlcpy'
fix it by include the required libite dependency.
Signed-off-by: Ming Liu <liu.ming50@gmail.com>
Also, added benefit, we don't get compiler warnings for ignoring the
return value of system(), which we don't care about in this case.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
- Allow scripts to take space separated arguments with quoting
- Simplify script-exists check, which already checks for leading /
- Unset current script on parse error
- Allow setting no timeout
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
These variables really belong in conf.c. Relocate and move external
decls. from finit.h -> conf.h to simplify linking of other programs.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
With the relocation of several functions, including inline functions,
from helpers to util, we no longer need to include log.h. Which is a
good thing, since any subsystem that needs logging should explicitly
include log.h
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The file helpers.c should strictly only be used for misc. helper
functions to the main Finit daemon. The functions moved in this
commit are generic enough to be used by any deamon or programs,
and for that we have util.c
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Prevent name clash with upcoming refactor and any confusion with
src/plugin.c functions with the same name.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This refactor should have been done years ago when we first introduced
the big state machine. The old functions service_runvel() and the oddly
named service_reload_dynamic() are actually state machine control fns,
so let relocate them.
Also in this commit, remove the global variable 'sm' and rename a few of
the critical functions to better names.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Historically we've kept these two apart, but their commonalities are far
more than what separates them.
This fixes a bug where ready:scripts were not called for sysv daemons,
and also opens up for adding support for a reload:script for both types.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This commit also switches the stdout/stderr log capture backend for all
notify:systemd services from the sysklogd project's logger tool to the
bundled logit tool.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
clang-tidy complained that parse_prio() always returned OK(0), turns out
it was right. This patch adds validation of the parsing.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This commit introduces a bare-bones replacement for libsystemd:
- Build .so file and add --with-libsystemd to configure
- Add capabilities support to test/src/serv.c
- Update tests to account for a Finit built w/o libsystemd support
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
A systemd service should only use NOTIFY_SOCKET and an s6 style
service reads its notify descript from the command line. For
details, see notify.sh
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This commit introduces a stripped down sd_notify(), taken from the
systemd man page example, which is used by the serv daemon in lieu
of the previous broken implementation.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This patch fixes issue #440 which details how systemd readiness
notification has been broken since its inception in Finit.
In addition to now providing a proper abstract socket, the refactor
also drops the reliance on the Finit client API to fetch a socket.
Now an s6 descriptor is the write end of a pipe() and the systemd
socket is instead created before forking off the service.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
When building with global CFLAGS=-Wunused-parameter the output from the
build looks terrible. This commit explicitly declares unused variables
to prevent triggering -Wunused-parameter
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Fix the following warnings:
| cgroup.c:296:13: warning: implicit declaration of function 'conf_any_change' [-Wimplicit-function-declaration]
| cgroup.c:297:17: warning: implicit declaration of function 'service_reload_dynamic' [-Wimplicit-function-declaration]
Signed-off-by: Ming Liu <liu.ming50@gmail.com>
Fix the following linking issues of tmpfiles:
| exec.c:308: warning: stty is not implemented and will always fail
| plugin.c:237: undefined reference to `run_parts'
Signed-off-by: Ming Liu <liu.ming50@gmail.com>
The existing implementation seemed useful enough to warrant a new
helper function. Care was taken to ensure the revised code no longer
suffers any memory leaks.
The tmpfiles.d spec contains entry types that should be acted upon in different modes
of operation: create, clean, remove, and purge - split up create & remove logic to
clarify the modes of operation finit supports.
The tmpfiles.d spec has proven useful in systemd, enough so that some developers
are starting to ship tmpfiles.d configuration files with their software.
By splitting the tmpfiles functionality in finit out into a separate executable
we are providing a useful piece of software that package managers can hook into.
An svc_t (service/sysv) that is in setup, teardown, or cleanup state
must be ignored by the pidfile plugin for any events regarding PID
files. E.g., a setup script for a sysv service may create a PID file
to alert the system that a setup process for the service is running.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The service_fork() function blocks all signals for the new process, it
is up to the callee to call sig_unblock() before calling exec().
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
All run/tasks must complete before Finit moves to runlevel 2. This
critical piece of information has mysteriously been missing from the
documentation since the start.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Add explicit argument separator before username to ensure correct
parsing of the login command regardless of username format. This
follows best practices for command execution and prevents unexpected
behavior when processing special characters in usernames.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
When a service is stopped, but does not respond to the default SIGTERM,
a fallback kicks in after 3 seconds (default) to send SIGKILL. Since
91a9c83 (v3.2-rc2) this has however not worked since the signal was
only sent to the process group.
It is the job of service_monitor(), when it reaps the process leader,
to forward the signal to the process group.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Don't try to source an optional env:file if it is missing. Otherwise
execution of pre/post/ready/cleanup script will fail.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Introduced to work-around systems with broken RTCs, it also introduced a
dormant bug in Finit, triggered at shutdown on some systems. The call
to dlclose() removed the memory to the rtc_timer which libuev later then
referenced.
Fixes#429
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Initiating a shutdown/reboot from `initctl` should not cause `initctl`
to be killed in do_shutdown().
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
When a user sets up a TTY without a device, triggering the tty->notty
code path, the tty->dev will be NULL and tty_parse_args() still return
OK result.
Found by Coverity Scan
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Enhance fallback entropy generation by mixing in multiple sources:
runtime statistics, PID, and high-resolution timestamps.
This provides better initialization for the PRNG when neither saved
entropy nor hardware RNG are available during early boot.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
When using Finit with an external getty that supports alternative login
program argument, the tty_parse_args() function did not check if a getty
command had already been registered:
tty [12345] /sbin/agetty -L -l /bin/login console noclear
This caused Finit to try /bin/login as the getty program.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This change introduces two new states for the big Finit state machine:
runlevel-clean and reload-clean. Here Finit now waits for any post or
cleanup script to finish before returning OK to the initctl command.
Additionally, the service state machine has been updated to ensure a
run/task/sysv/service calls any post or cleanup script before they are
removed. A new 'dead' state for svc_t is introduced which any removed
svc_t ends up in now instead of becing collected from 'halted' state.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Add cleanup:script support, runs at service removal
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Some subsystems call iwatch_add() without first calling iwatch_del() on
the same path. E.g., cgroup_config().
Issue #417 but unclear atm. if this is the root cause.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
A long running pre/post/ready script must be killed properly, not by
targeting its process group. Process group cleanup is handled by the
service_monitor() when reaping the script's PID.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
New test adds /bin/fail.sh to verify that a failing pre:script (that
also takes too long to run) is detected: exit code and timeout.
Ensure existing test pass full path to /sbin/fail.sh script.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Check exit status of `pre:` scripts, on failure drive service/sysv to
`crashed` state. The exit code of `post:` scripts remain ignored for
now.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Syntax:
[pre|post|ready]:[0-3600,]/path/to/script
Description:
Before this patch all pre/post/ready scripts used the global kill
delay as timeout. After this patch it is possible to disable the
timeout as well as set a timeout >60, which is max kill delay.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Follow-up to 465bc17, which addressed unintended restart of siblings.
This patch fixes a problem where template instantiated services are not
properly reloaded/restarted when marked as "dirty" with `initctl touch`.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
add support for 'nofail' fstab option. if present, finit's fsck invocation ignores errors for this device.
this is useful if you have an fstab entry for a device with UUID= or LABEL= which isn't always present and which you'd like to not
bail on (so you set nofail). in this case finit leaves the presence-or-not decision to fsck, which exits nonzero.
for block devices that are directly listed in fstab this change isn't important, because for such finit looks for the blockdev's existence and skips the fsck if n/a.
Consider the case where container@.conf is an available template. When
creating a container@foo.conf it will share the same base .conf as an
existing container@bar.conf, but we do not expect to restart bar just
because foo is instantiated.
Up until this change, all template siblings were considered "dirty" if a
new one was created or updated. Skipping realpath() for all files that
have a '@' works around the problem.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
These plugins signal success and failure directly to the console, the
user should inspect syslog for more information.
This change is a follow-up to 340cae4, where kernel logs of LOG_ERR and
higher are allowed to log directly to the console. Since syslogd has
not been started before these plugins, the log messages would otherwise
leak to the console.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
In 42ef3d3c, for v4.4-rc1, support for setting a custom RTC restore date
was introduced. Unfortunately the configure script was wrong and caused
config.h to contain
#define RTC_TIMESTAMP_CUSTOM "$rtc_date"
instead of
#define RTC_TIMESTAMP_CUSTOM "2023-04-10 14:35:42"
Furthermore, the error handling for strptime() was wrong, so the restore
date was always reverted to the default.
This patch fixes both issues and extends the DATE of --with-rtc-date to
also include seconds.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This patch introduces a new configure option --with-rtc-file=FILE. When
enabled the RTC plugin detects missing RTC device and falls back to save
and restore system time from a file instead. When --with-rtc-file is
used without an argument the default file is /var/lib/misc/rtc, but the
feature itself is disabled by default.
The usefulness of this feature may not be obvious at first, but some
systems are equipped with an RTC that resets to a random date at power
on. This can be really bad in the case the date is far in the future,
because an NTP sync would then cause time skips backwards, which shows
up in logs and causes a whole lot of pain in alarm systems.
The solution is to disable the RTC driver or device tree node, and when
Finit starts up, the RTC plugin detects a the device node and instead
restores time from the last save game. Meaning time will always only
move forwards.
NOTE: when Finit is built --with-rtc-file we always save to disk, but
only restore from the "save game" if restoring from RTC fails.
If the system has no RTC we always restore from disk.
As an added bonus, this change also makes sure to periodically
sync also the RTC with the system clock. Useful for systems
that do not run an NTP client.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Some boot loaders, like GRUB, leave background color artifacts from
their boot menu. This patch resets the foreground and background
color attributes, and then clears the screen, without clearing the
scrollback buffer.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This patch changes a behavior that's been default since Finit 4.0,
introduced in 4d05bf9 with 4.0-rc2.
If service B depends on A and A needs to be reloaded, then B may be
affected. If A is declared as NOHUP <!>, then A will be stopped and
restarted, during which time the condition it provides is removed,
and B will also be stopped.
However, and as of this patch, if A is declared supporting HUP, then the
condition A provides will only go into flux, during which time B will be
SIGSTOPed instead of needing to be reloaded.
Fix#415
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
When dealing with "L+" and "R", the function call 'nftw' should not
follow symbolic links, otherwise, it would also delete the targets
which is wrong.
For instance, if there is already a symbolic link:
```
/path/to/the/link -> /path/to/some/folder
```
if we set the following in a tmpfile conf:
```
L+ /path/to/the/link - - - - /path/to/the/target
```
the result would be /path/to/some/folder also get deleted, which it
should not.
it could be even worse, when the symbolic link already is pointing to:
/path/to/the/target, the whole /path/to/the/target would be deleted on
next system boot.
Signed-off-by: Mathias Thore <mathias.thore@atlascopco.com>
Signed-off-by: Ming Liu <liu.ming50@gmail.com>
Adds the function fistmpfs to determine if a new tmpfs mount should be
performed on /run and /tmp. The function supports cases where more
complex mount hierarchies are in use, including overlayfs backed mounts.
Signed-off-by: Mathias Thore <mathias.thore@atlascopco.com>
A Linux system booted with the kernel command line option 'quiet' only
logs error (and above) severity messages to the console. For embedded
systems, which is the primary target for Finit, this is what you want
to see.
Hence, and after careful consideration, this patch changes the default
behavior of Finit to allow kernel logs to the console. A build-time
configure flags, --disable-kernel-logging, has been added to restore
legacy behavior.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Systems that have the following tty setup and multiple consoles listed
in /sys/class/tty/console/active misbehave:
tty [12345789] @console 0 xterm noclear passenv
Only the first listed console is started properly, the remaining ones
were registered using the wrong :ID and no arguments to getty.
This patch fixes the parsing and re-use of the base paramenters for
all consoles listed in the active file.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Currently no code checks the return code of ifupdown at runlevel
changes, but for future reference this code has been refactored
to match the changes made in cbdb949 to fix issue #400.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This problem affects all calls to run_interactive() that check the
return value of the command. Causing HOOK_MOUNT_ERROR to *not* run
on mount failure, and sulogin() to *not* be started on fsck error.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Silence log warnings for command x/X (ignore clean for path), because
Finit does not do tmpfiles cleaning at runtime.
x /tmp/podman-run-*
x /tmp/containers-user-*
x /tmp/run-*/libpod
D! /var/lib/containers/storage/tmp 0700 root root
D! /run/podman 0700 root root
D! /var/lib/cni/networks
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
After reports from the field, see issue #397, of lockups at reboot,
we've decided to drop this code from PID 1. It was added before the 4.x
series, when the current progress output was introduced. For the older
style progress it served a purpose since the placement of [OK]/[FAIL]
was on the right hand side.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
When the system shuts down, or user changes runlevels, we don't have to
call cond_clear_update(), because this can lead to nested service_stop()
calls, which in turn lead to out of sync progress updates:
[ .. ] Stopping Foo
[ OK ] Stopping Bar
[ OK ]
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The /etc/issue file on Alpine Linux says "Kernel \r on an \m (\l)", so
\r needs to return the uts release rather than os-release VERSION.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The responsibility of the kernel event daemon is to relay kernel events
to Finit. At shutdown and reboot it is too late for more events and the
daemon should just shut down with other services.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Fix copy-paste of error message from cond-w.c
A read-only root filesystem may have /var/lock, while we want to remove
it and add a symlink to ../run/lock. Ignore errors from this since we
cannot do anything about it. It is up to the user to fix their skeleton
or use an overlay.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Inspired by Alpine Linux, add /dev/mqueue if missing. We should check
the /proc/filesystems first, but this is quicker.
The sticky bit ensures only the owner of files in /dev/shm can delete or
rename files. This is also what Alpine Linux use.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
A service with notify:pid is 'ready' when the pidfile has been created,
the converse also holds true -- when a pidfile is removed the service is
no longer 'ready'.
The state transition for the service has probably already been done, in
svc_set_state(), clearing all <service/foo/*> conditions when the PID
was collected. The pidfile event may arrive later, so for completeness
we make sure the 'ready' condition is not recreated at least.
Problem introduced in 912a281 with the original supoport for service
readiness notification.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
There exist two possible basename functions, a xpg compliant one in libgen.h
and a GLIBC exclusive one declared in string.h, that was previously also declared by musl libc.
Both implementations are expecting different parameter types (`const char *` for GLIBC and `char *` for xpg)
With the removal of the basename function from string.h in musl libc, we could only rely on the xpg implementation.
Unfortunately, the xpg implementation of basename does modify the contents of whatever you put in it,
even though that there really is no need for it.
This is an issue in some cases, where we might want to get the basename of a read-only variable, e.g. a `const char *`,
as trying to modify something read-only is undefined behavior.
So in order to keep things consistent for us, we implement our own version of basename called `basenm`,
that does not modify the passed argument.
As of Finit v4.6 we no longer assert the PID condition for services
declaring themselves as notify != pid. We replace D with a forking
service to catch any future regressions in the pidfile plugin.
No need to check reload PID of D, it is enough to check PID of C.
Also, reduce the number of retries at startup. If we haven't gone
up within 10 sec with this tiny config something is really wrong.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Premise, a service declaring itself 'notify:none' should never assert a
pid condition. However, forking services still need to be supported and
the only way to do that is if they create a pid file. Hence, instead of
skipping pidfile_update_conds() completely we need to filter the type.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
2023-11-13T04:02:23 [WRN]: conf_change_act():failed registering /etc/finit.conf/ event mask 00000004: Not a directory
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This fixes an issue when finding a global environment variable with
spaces in the variable name:
set COLORTERM=yes
Literally, 'set COLORTERM' was the name of the variable.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This change expands the readiness notification system in Finit with the
native 'pid' style, which will remain the default readiness in Finit 4.x
For systems that want to transition to Finit 5.x early, a global option
to set 'readiness none' in /etc/finit.conf, has been added. This change
the service default notification mode to 'notify:none', which can also
be set by Finit 4.x ('readiness pid') for select services.
Fixes#386.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This fixes an annying buglet that creeped in just before the v4.5 release
causing none of the internal services to be registered properly:
conf_save_service():Failed creating ${localstatedir}/run/finit/system/dbus.conf: No such file or directory
plugin_run_hook():Calling modules-load hook n:o 5 (arg: 0x55c4c8133621) ...
load():Scanning /etc/modules-load.d for config files ...
cond_set_oneshot():hook/svc/plugin
cond_set_oneshot_noupdate():hook/svc/plugin => /run/finit/cond/hook/svc/plugin
cond_update():hook/svc/plugin
parse_conf():*** Parsing /etc/finit.conf
conf_save_service():Failed creating ${localstatedir}/run/finit/system/runparts.conf: No such file or directory
The fix, like most, is simple when you find it. We must expand $runstatedir
before creating the #define in config.h
For convenience, here's the patch for the generated configure script:
&<---------------------------[cut here]---------------------------
--- a/configure 2023-11-05 23:39:51.907334321 +0100
+++ b/configure 2023-11-05 23:39:58.339299795 +0100
@@ -15352,7 +15352,15 @@
printf "%s\n" "#define FINIT_EXECPATH_ \"$pkglibexecdir\"" >>confdefs.h
-printf "%s\n" "#define FINIT_RUNPATH_ \"$runstatedir/finit/system\"" >>confdefs.h
+ finit_runpath="$runstatedir/finit/system"
+ finit_runpath=`(
+ test "x$prefix" = xNONE && prefix="$ac_default_prefix"
+ test "x$exec_prefix" = xNONE && exec_prefix="${prefix}"
+ eval echo \""$finit_runpath"\"
+ )`
+
+
+printf "%s\n" "#define FINIT_RUNPATH_ \"$finit_runpath\"" >>confdefs.h
&<---------------------------[cut here]---------------------------
Ensure you include the three empty lines for context at the end!
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
We cannot rely on the auto-detection of Bash completion dir during 'make
distcheck' because autotools does not use DESTDIR, only --prefix for the
install check, and pkg-config returns a system path.
Also, show detected path in configure summary for debug.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
In Finit v4.5 we've moved the start of rc.local and runparts to the
transtion from bootstrap to multi-user, so we must give it time to
finish.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
In Finit v4.5 we've moved the start of rc.local and runparts to the
transtion from bootstrap to multi-user, so we must give it time to
finish.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This odd little feature makes it possible to declare run/tasks with a
condition that does not block Finit transitioning from bootstrap to the
next runlevel.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Regression in v4.5 release cycle, no changelog notice needed.
Reported as part of issue #378 by Alexander Zangerl.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This change allows us to support constructs like this:
RUNDIR=/var/run/somesvc
DAEMON_ARGS=--workdir $RUNDIR --other-args...
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Introduced back in v4.3-rc2, 82cc10be8, the support for automatic
service conditions have had a weird and unintended behavior. Any
change in state (see doc/svc-machine.png) caused Finit to clear
out *all* previously acquired service conditions.
However, when moving between RUNNING and PAUSED states, a service
should not have its conditions cleared. The PAUSED state, seen
also by all conditions moving to FLUX, is only temporary while an
`initctl reload` is processed. If a service has no changes to be
applied it will move back to RUNNING.
Also, we cannot clear the service conditions because other run/task
or services may depend on it and clearing them would cause Finit to
SIGTERM these processes (since they are no longer eligible to run).
This patch not only adds this pre-condition to `cond_clearn()`, it
also clarifies which state (before or after) the particular code
is interested in.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Renamed .conf file for udev/mdev etc. caused 'mdev -df' to start in
tests. This fix closes that again since none of that is needed in
our small namespaced world.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Alexander Zangerl reports that <service/foo/STATE> conditions seem to be
removed when calling `initctl reload`, even though no .conf changes have
been made.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The delayed matcehd works when all run/task/services provide their own
unique identity, but for replacements this is sometimes not the case.
Example from system/10-hotplug.conf.in:
run nowarn conflict:udevd,mdev cgroup.init name:coldplug <service/mdevd/ready> \
[S] mdevd-coldplug -- Cold plugging system
vs
run nowarn conflict:udevd,mdevd cgroup.init name:coldplug if:!mdevd <service/mdev/running> \
[S] @pkglibexecdir@/coldplug -- Cold plugging system
Here they both provide the <run/coldplug/*> conditions, and if the first
is loaded, because we found mdevd, then we should not attempt to load
this one.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The updated 10-hotplug.conf now ensures modules are loaded, so we no
longer need the modprobe.so plugin to be enabled by default.
Also, update the hotplug "plugin" description, it is kept entirely
for backwards compatibility reasons after the plugin was converted
to 10-hotplug.conf.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This commit replaces the `mdev -s` call to populate the device tree with
the "new" `mdev -df` daemon mode, introduced in BusyBox 1.31.0, 2019.
In daemon mode mdev listens to kernel uevents, replacing the hotplug use
of mdev. After creating the netlink socket, 'mdev -df' performs the same
initial scan as 'mdev -s' did.
To perform device (re)discovery, module loading and setup, including any
firmware loading, a coldplug operation is typically required. This is
done by the new /libexec/finit/coldplug script, by Alexander Zangerl.
Unlike 'mdevadm settle', mdev does not offer any mechanism to detect when
the discovery operation is done: on slower systems the triggering side,
the coldplug script, often completes quite a bit earlier than mdev's
uevent processing. I.e., depending on <run/coldplug/success> is not an
indicator of all devices having been (re)discovered and fully set up.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
On some systems udev may very well take more than five seconds to complete.
Probing modules, loading firmware, etc. Note, the timeout only sets the
maximum wait time.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The directory may hold various custom generated conditions that are not
managed by Finit. Not being able to remove the directory is not a
critical error.
finit[1]: cond_delpath():Failed removing condition path /var/run/finit/cond/: Directory not empty
Therefore, ignore ENOENT and ENOTEMPTY and log everything/anything else.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This fixes the regression in the tests, which runs in a very stripped
down world without tmpfiles.d etc.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This adds a new function conf_save_service() replacing service_register() for
plugins and bundled services like watchdogd, keventd, runparts, etc.
The benefits to this change are several:
- Plugin/Bundled services no longer risk starting before udev or other
critical services/task have started
- Definitions can be overridden by an administrator (see docs)
- Increases visibility (user: where are all these services coming from?)
Previously the origin (file the service was loaded from) was NULL.
- Adds another level of extensibility to Finit
The most notable change is that dbus is no longer started before udevd.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The system/10-hotplug.conf needs to index its calls to udevadm for all
calls to udevadm, not just the trigger calls. Otherwise the first
udevadm is overloaded with all subseequent (unindexed) udevamd stanzas.
Also, make sure all udevadm calls to settle/reload are hidden.
Fix#372
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The if: statement, introduced in v4.4, allows for discarding run/task/services
that depend on other services, based on that service's name, or condition.
Discarding in this context means unloading from the configuration.
At runtime, however, it has proven quite useful to be able to conditionally
qualify a run/task/service based on a condition. Consider this example from
the Infix operating system:
run name:startup log:prio:user.notice \
[S] <pid/sysrepo> confd -b --load startup-config -- Loading startup-config
run name:failure log:prio:user.critical if:<usr/fail-startup> \
[S] <pid/sysrepo> confd --load failure-config -- Loading failure-config
The two run statements reside in the same .conf file so Finit runs them in true
sequence. If loading the file `startup-config` fails confd sets the condition
usr/fail-startup, thus allowing the next run statement to load `failure-config`.
Notice the difference between <pid/sysrepo> condition and if:<usr/fail-startup>.
The former is a condition for starting and the latter is a condition to check if
a run/task/service is qualified to even be considerered. The best comparison is
with the [runlevel] option, it too is used to qualify.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This change opens up the runlevel change API from bootstrap. The twist
is that the change is only queued, i.e., the call `initctl runlevel 9`
during bootstrap only changes the configured runlevel to go to after
bootstrap has completed.
Effectively, this change allows overriding the `runlevel` directive in
/etc/finit.conf without having to change the file.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
No more changes are expected after -rc3. Will be used for testing in
Infix and br2-ext-finit before the final v4.5 GA.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This patch extens the SysV compatibility support in Finit by adding
support for limiting `runparts` to run only SNNfoo, or KNNfoo, style
scripts from a directory.
Additionally, by default `runparts` now runs entirely in the background
without any progress. To enable progress, an optional argument has been
added to the runparts command line.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
As pointed out in #366, the configure options --enable-fastboot and
--enable-fsckfix should just alter the default values of the two fsck
command line options.
This commit simplifies the code and makes it possible to override using
the command line regardless of the two build options.
Finally, add the two command line options to doc/cmdline.md
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
We use LOG_CONS to ensure log messages reach the operator, but since
watchdogd starts very early this means non-critical messages like the
initial greeting leaks to console because syslogd has not yet started.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
With non-standard paths, e.g., when running `make distcheck`, the
absolute path to some commands become ridiculously long. However,
this has been a recurring issue for some users in the past, so it
is time to increase the capabilibieies of Finit to cover this.
Yes, a better way is probably to allocate all these strings when they
are used, but that would require a redesign of the initctl API and
likely cause a lot of regressions before everything has stabilized.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The rescue mode that can be invoked from the kernel command line is
potentially unsafe. Many systems lock the root user account, or use
another account for managing, e.g. 'admin'. The sulogin program(s)
would on such systems give the user a root prompt.
In #357 we added support for setting a custom sulogin user, but for some
systems that is not enough. On many embedded systems the /etc/passwd
and shadow files are populated at bootstrap and at the time rescue mode
runs, these files will be unpopulated.
The only, truly safe, approach on such systems is to disable rescue mode
completely. Otherwise intricate Finit plugins have to be used that run
before rescue mode is started -- increasing the complexity of the system
as a whole.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
In issue #357 there's a proposal for optionally allowing /bin/login, but
after intense discussions we've agreed that would be opening up for all
sorts of potential (security) issues.
it's better to keep things as-is but with the added twist of supporting
a custom user at configure time, e.g., 'admin'.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Non-trivial systems, like Infix[1], require more than one lap around the
track to warm up. This is a follow-up to issue #362, commit 4701ede, in
fixing an obvious oversight in the new sm_check_bootstrap() work. While
there are still outstanding bootstrap tasks or services that have not
yet started, we must reschedule the worker.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Drop complete() logic, waiting for run tasks to finish, from the
service_start() funciton to the general service_monitor(). This
refactor frees up the main loop and allows us to answer any API
calls from initctl even from the run task itself.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
For some specialized forms of runparts, e.g., start-stop scripts in SysV
init replacements, there is a need to append an argument. Typically
this is 'start' or 'stop'.
This functionality has been broken for quite some time, meaning we
should consider it being unused by current users of Finit.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Bootstrap-only tests that just verify basic functionality are often very
short. This change handles that by retrying SIGUSR2 until Finit has had
the chance to finalize bootstrap and enable signals.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The code refactored in this commit has long been an eyesore. The
hand-over, from the sequential main() function, to the event loop
was clumsy at best and was very difficult to debug.
Instead of using a series of workers we now delegate everything to
the big state machine by introducing a new SM_BOOTSRTAP_WAIT_STATE.
Both the runparts and rc.local scripts now run in the background, no
longer blocking the event loop from responding to initctl calls. To
ensure they run at the same point in time two things have been done:
1. A new <int/bootstrap> condition has been added which triggers
runparts, which now is a regular task created by conf_init()
2. /etc/rc.local is started from SM_RUNLEVEL_WAIT_STATE when we
transition from runlevel S to any other runlevel.
Fixes#356
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This should clarify the messup on the behalf of the mainatiner in not
documenting this properly, including the deviations introduced in v4.4
See issue #359 for a background.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
- iwatch events return filenames in dir field, ignore file for those
- improve warning message on failure
- drop bogus "Out of memory" error log message
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The execvp() call always fails with -1, so only logging exit code when
collecting the PID is not enough. When debugging the user needs to see
the errno from execvp(), e.g., "No such file or directory", when the
command was missing the absolute path to the executable.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Related to 6f0d448 in that we check if the kernel knows about this
process before we try to stop it. If it's already dead we clean up
and return.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This turns out to be the root-cause of #227. Finit is waiting forever
for proceeses to stop at shutdown/reboot, while a subreaper has already
collected the PID, or Finit for some reason did not collect the PID.
When the process timeout calls service_kill() we now check if the kernel
actually knows of this process or not. If it's already been collected,
we can notify Finit of this by calling service_monitor() to clean up the
'svc' and in turn call sm_step() to finalize the state transition.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
When calling `initctl -b create` from a start script at bootstrap you
risk blocking the boot since Finit currently cannot reply to IPC during
that period.
This patch allows -f to override this builtin check for the following
initctl commands:
- touch
- show
- edit
- create
- delete
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
These changes add a new svc_block_t type: SVC_BLOCK_CONFLICT so a user
can more clearly see why a run/task/service has not been started by
Finit. The reason for the block is by default logged, which can be
escaped by using the `nowarn` flag.
Also, when the conflict is resolved, allow the service to start.
With these changes, the system/hotplug.conf should work better and
cause less questions about "strange" log messages.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Apparently the qemu-system-x86_64 in -nographics mode does all sorts of
crazy things to the terminal. Most annoying, it disables linewrap.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
In 91a9c83 a regression was introduced that affects the way Finit stops
a supervised process and its process group.
Instead of sending SIGTERM to the process, delegating the responsibility
to that process to inform any children it may have, Finit as of 91a9c83
sends SIGTERM to the entire process group. For SIGKILL this is fine,
SIGKILL only runs as cleanup and as a last ditch effort if the process
doesn't respond to SIGTERM.
This regression, introduced in v3.2, directly affects services like
avahi-autoipd that have forked off children that it needs to tell to
exit cleanly before it returns. With the patch in question these
children are never allowed to complete, which in turn causes lingering
169.254 link-local addresses on interfaces.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
When started as a monitor in a container, or for the test suite run from
distcheck, we must allow Finit to find its own tools like initctl. This
should also be a more generic solution that cover more cases.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
A simple distcheck revealed that the local path[] was too small to store
something like /home/jocke/src/finit/finit-4.4-rc1/usr/lib/finit/system/
This refactor should cover all cases.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The user may have their own setup of things, by default we should not
cause unnecessary warnings in syslog.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Due to an old design decision runlevel S was encoded internally as '0',
meaning it was the same as halt/poweroff. If you want to run scripts at
system shutdown this was less then ideal since it meant your scripts
also ran at bootstrap.
This change is quite invasive. It introduces INIT_LEVEL (10) as the
value for runlevel S, meaning all code that parses and/or evaluates
anything for runlevel 0/S was affected and had to be reworked.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Some services (openresolv) tries to talk to it very early, so lets get
dbus up and running as soon as possible.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The logger redirection process was sometimes in the init group instead
of the system/<svc>/ group. This was due to the parent PID being moved
after the logger was forked.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
When we've started a run task, regardless of its exit status, we must
mark it as not dirty to prevent it from being endlessly restarted.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
When the inotify callback is called for a file that has been removed,
the call to realpath() obviously fails (ENOENT). This fix only takes
the regular case into account, files removed from a symlinked enabled
directory is not handled -- needs more work for that.
Regression introduced in 0a94f3f, PR #347, issue #346.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Post audit, concensus is to drop if: and always do post-eval of all
ifdef: statements. Also, rename ifdef: to if: for completeness.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Multiple parallel tests writing to the same /var leads to quite
unexpected results. Let's start fresh in each test instead.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This is the first example of the just minted advanced stanza syntax with
if-statements, conflict handling, and nowarn flags.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Conditional loading of stanza depending on ident is already loaded. The
if: checks as in-band while ifdef: checks out-of-band, i.e., post eval.
The optional leading '!' negates the comparison, if NOT foo then ...
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Slowly migrating away from hard-coded services in plugins. This way
it's possible for the user to both inspect and override as needed.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The tmpfilesd() glob sorts files according to name, we could name our
.conf as 00finit.conf to prevent ordering issues with, e.g. dnsmasq,
but this is more elegant and allows for multi-level override.
NOTE: bootmisc depends on the pidfile plugins since the latter need
to set up its iwatches of /run before bootmisc creates /run.
Depending on if it's a system with /var/run or /run we need to
drop /var/run before recreating it.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This changes the semantics of Finit a bit by checking for the command to
run when registering it, skipping commands that cannot be found in the
absolute path provided in the command, or in $PATH.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Then it works to run the applets from outside the unshare. Useful for
testing and test development.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This change adds very basic tmpfiles.d/ support to Finit. Much of the
basic types are supported, but not all, so for now, please check the
code for details on what is working.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Starting a service like this works fine:
service [2345789] env:-/etc/default/nginx nginx -g 'daemon off;'
However, on `initctl reload` the previous argument was lost while
parsing the end of the new one. So the comparison always resulted
in a 'diff' causing the service to be unnecessarily restarted.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
initctl: Follow priciple of least surprise for "cond get"
Fixes regression in 4.4 release cycle, commit c3c662f, when support for FLUX condition state was added.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The accepted standard in Unix is to report successful executions with
exitcode 0. Therefore, map a "initctl cond get" of a condition to the
following exitcodes:
- On: 0
- Off: 1
- Flux: 255
Fixes: c3c662fe64 ("initctl: ensure 'cond get' support flux state")
Signed-off-by: Tobias Waldekranz <tobias@waldekranz.com>
The TTY and log configuration is overloaded using a union in svc_t, so
we cannot "reset" the log attribute on reload. In fact, a TTY should
never be allowed to have a tty:anything set, so change the parser to
ignore log options for ttys.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Unfortunately we cannot use realpath(3) here since the the PID files
usually do not yet exist at this point.
Add and modify my ugly de_dotdot() from Merecat httpd.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
When this code was moved from finit.c we rely on WDT_DEVNODE to be
defined in config.h, which is controlled by configure. Meaning, before
the relocation, Finit did not honor the configure settings.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This issue was found when the test pre-post-serv.sh suddnely started
failing, unclear how it could ever have worked before considering the
nature of this patch.
The trick here is when reloading a service like this:
service env:/etc/env serv -np -e foo:bar
into this:
service pre:/bin/pre.sh serv -np
In the second the env: has been removed and pre: added. But with Finit
prior to this patch, env: was kept leading to unintended behavior. The
same goes for the other options included in this patch.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
In particular watchdogd we want to start in runlevel S to ensure we
start kicking the WDT. (Some HW WDTs are enabled from power on.)
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
All services registered in the system rely on conf_init() having been
set up properly, e.g., global_rlimit. Having conf_init() be responsible
also for registering static services is only logical, and also helps us
clean up main() a bit.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Instead of the revert in c9fe9afa, we restore HOOK_BASFFS_UP to its
proper place at the end of fs_mount_all(). For this to not cause any
regressions we add a new hook, HOOK_SVC_PLUGIN, and update all plugins
that call service_register() to run at the new hook.
This will cause regressions for external 3rd party plugins that rely on
HOOK_BASEFS_UP to be called at its previous postion. Nevertheless, this
is the proper fix to the problem.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
With all the change and revert mess of the HOOK_BASEFS_UP tests started
failing due to usr.so and sys.so failed to set up their conditions. It
then became clear that boomisc.so cleaned up /run ... with the revert
we got service_register() to be called before global_rlimit[] was setup
causing services registered by plugins, before conf_init(), to crash.
This plugin and test is here to ensure we catch this problem in case of
any future refactor.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Not all Linux systems are based on Debian, and even if they are inspired
by Debian (Buildroot), they do not necessarily use the same defaults.
This patch probes the current system for:
- dbus user and group
- dbus PID file
If the user/group cannot be found we fall back to "root", if the PID
file cannot be determined we ignore PID file readiness.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This reverts commit 5b41c6e since it causes regressions in plugins
adding services to the system. The proper fix for early bootmisc
is to add a new hook, which will be added in subsequent commits.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
No need for special handling of s6 services, we now have the new
service_notify_reconf() function that is called on init reload.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The sysklogd logger is capable of logging the PID of the actual service,
rather than just the logger tool itself (which would be a differet PID
every time).
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
When 'initctl reload' is called new "configuration generation" is
started by Finit. This mechanism ensures services reaffirm their
readiness and helps distinguish between their ready before reload
and after.
For native services the pidfile plugin handles this. The s6 style
services are one-shot only, and thus use a static ready condition.
This patch adds support for stepping the configuration generation
of unmodified systemd style services.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
To reset the crash counter of the 'serv' daemon we need to remove it
from /etc/finit.conf (and reload) between each test phase.
The repeated test phases can now be dropped, since the lack of reset
was the root cause for it failing sometimes in extended runs.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Correct an obvious oversight, we need to ensure the native pidfile based
daemons also work as intended in the new readiness notification world.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Ensure the test mimics actual real-world scenario where daemons do not
create a PID file at all.
Issue #343.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The 'initctl cond' commands were initiallý added to only manage usr
conditions. Recently the 'cond get' command was expanded to allow
reading general conditions as well.
However, since general conditions support three states the command
returned 'on' for conditions that were in flux. This patch fixes
that oversight.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The svc_set_state() function not only handles state transitions, it also
makes sure to cancel any outstaing service timeouts. Before the service
readiness support was added, service conditions were simply a single bit
of information, so clearing its conditions before changing states had no
impact. With readiness we have multiple bits of information that are
set as conditions.
When restarting a crashing service we set a timeout to retry restarting,
or ensure the service is now running. This means that the service can
signal READY before RUNNING is reasserted, but the RUNNING state can be
set also by the service state machine, so we need to ensure we do not
re-assert RUNNING if it's already set, and thereby clearing READY.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Issue #343 reports that readiness notification on crash/restart is lost.
This update to notify.sh reproduces that bug.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Principle of least surprise. When no even the maintainer remembers
the correct command, it is time to add an alias.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Over the years there have been multiple cases of invalid and/or unsafe
uses of signals and initctl commands at bootstrap and shutdown. These
cases cannot be safely supported. This commit locks down finit a bit
to avoid the most common cases.
If you run into this, please open a new discussion at GitHub and we'll
talk about it. Maybe I've been overzealous or you have another use-case
that warrants opening up some or parts of the API.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Not until fs_mount_all() has completed do we have /run, or /var/run, and
never after shutdown has been initiated.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
As pointed out by Jack Newman in issue #338, when conditions are not
available (at boot before cond_init(), or shutdown when /var or /var/run
have been unmounted) we should not try to create any conditions.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Slightly different take on issue #334 making it optional, possible to
enable per system.
reboot-delay <0-60> # default: 0 (disabled)
When enabled (non-zero), runs after filesystems have been unmounted,
the root filesystems has been remounted read-only, and sync(2) has
been called, twice.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The bootmisc plugin sets up the base path for finit and figures out if
the system has /run or /var/run -- needed by the other plugins.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This to save time during regular pushes and pull requests, the tests
run from distcheck take a long time, and they're not getting fewer.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
At this hook point the bootmisc.so plugin runs and creates all relevant
system directories. Much of the rest of system bootstrap relies on this
so it should be called together with the other FS hooks as soon as all
the filesystems have been mounted.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Works locally with/without this patch, but I cannot understand why it
would ever be needed in the first place, so removing.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
When the /etc/finit.d/enabled directory is a symlink elsehwere, e.g.,
/var/finit.d/enabled, a relative symlink will not work.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
sig.c was calling plugin_run_hooks for the shutdown hooks, but the
shutdown hooks do not have registered callbacks in hook-scripts.c
Instead directly call plugin_script_run()
This fixes `initctl reload` correctly restarting all daemons that have new
command line arguments.
Previously command line arguments changes were only acted upon if the
service was explicitly reloaded `initctl reload myservice`.
We use strtok() quite extensively in .conf parser. Parsing of a service
line is split in two, with the command arguments being parsed very late.
Adding a new sub-option parser, based on strtok(), risk losing command
arguments. Thus, change to use strtok_r() so that we always know where
the arg parser is.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Last update of plugin before dropping it in favor of an equivalent .conf
snippet. (Pending new finit-conf repository).
Issue #327
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This patch changes the way progress for unmounting filesystems is
displayed. To see each filesystem being unmounted Finit now must
be in debug mode.
A general "Unmounting filesystems ..." have been added instead as well
as progress for last hook script and finall call to reboot().
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This change drops the confusing status progress output, which was always
OK since the actual modprobe operation runs in the background. No need
to show status of the "added a task to finit, found modprobe" command.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Instead of waiting until runlevel 2, 3, 4, or 5 to load modules with
this plugin, we should load modules as soon as possible, i.e., in
runlevel S. This also means that Finit will clean up all the tasks
so we don't have to see them in `initctl status`.
To allow users to revert to the old behavior, we add a new setting
set runlevel 2345
that per modules-load.d/file changes the runlevel for all modprobe
commands.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
GitHub have now stepped their 'ubuntu-latest' image to Ubuntu 22.04.
With that comes the GLIBC bump to >= 2.34 which require aligning the
time bits and file offset bits to 64 bits. This is libuev v2.4.0.
We also make sure to align libite to the same version between the
workflows to avoid any nasty surprises.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
All hook scripts are called with at least one environment variable set,
FINIT_HOOK_NAME, useful when reusing the same hook script for multiple
hook points. It is set to the string name, also used by the path, e.g.,
hook/net/up.
For all hook points from hook/sys/shutdown and later, FINIT_SHUTDOWN is
also set, to one of: halt, poweroff, reboot.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
For all other run_parts() use-cases we just give the script control over
stdout/stderr to prevent clobbering ANSI color escape codes.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This reverts commit 03c08d3970 due to it
breaking handling of bootstrap tasks, causing endless boot loop.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This protects against any hook scripts or services calling initctl when
we're going down. There's no way to service these requests anyway, so
no point in keeping the socket open. This way initctl can check for
the existence of the socket before attempting any IPC command.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Add support for /etC/os-release, with fallback to /usr/lib/os-release,
overriding some traditional modifiers with the os-release variant. The
traditional values for these modifiers are taken from uname, which on
Linux systems are pretty useless since they always return the kernel
name and version instead of the distro/OS values.
E.g., \s becomes PRETTY_NAME instead of 'Linux' and \v becomes the
pretty VERSION, while \r becomes VERSION_ID.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This change makes it possible to use the standard continuation character
in .conf files. For example:
service [23456] \
log:stderr \
<!pid/foo> \
bar \
-- This is the bar service, it depends on foo # comment that's dropped
Notice the additional support for comments on .conf lines. If you use
the comment character today you must change to escape it. Like this:
service :2 foo -- Foo \#2
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
When removing a whole tree of conditions we need to walk through each
and call cond_update() to ensure that all dependant tasks are stopped
and put back in waiting.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This change prevents Finit from attempting to continue restarting
crashing services that've lost their conditions.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Hopefully prevent spurious failures, dunno why atm., parallel checks
should work now. (Works locally with make -j20 check)
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
When a service starts up the pidfile.so plugin automatically detects any
new pidfiles that show up in the system. If it finds a matching service
which does not have a custom configured pidfile, it auto-assigns the
detected pidfile to the service.
However, when the user later calls `initctl reload` to update the
running configuration set, Finit service_register() mistakenly will
set the service to forking since it now suddenly has a pidfile ...
The new style way of configuring a forking service is to set the
type:forking
option. The old compat way was to set `pid:!/path/to/pidfile`, so
in service_register() we must be very careful not to interpret the
registered pidfile as confiugred by the user -- i.e., check the pid
option.
Bug #313
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
When a service crashes Finit launches a restart timer that periodically
will try restart the service. If a user calls `initctl stop foo` finit
must cancel this timer callback, otherwise we may end up with a weird
state where Finit thinkgs the service is running, but pid: 0, i.e., not
started.
Note: this fix has been expanded upon from the proof of concept
submitted by Jack Newman.
Bug #313
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Running tests in massive parallel causes failing startk-kill-service.sh.
This in turn was caused by the shared test tool 'slay' that cached its
"oldpid" file to a shared area of the rootfs (tenv-root). It must use a
test instance-specific ramdisk.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This change adds support for a new command line option finit.config=PATH
which can be used to tell Finit to start with /etc/factory.conf instead
of /etc/finit.conf.
For the complete experience a new top-level configuration file directive
`rcsd PATH` has aslo been added. It in turn can be used by factory.conf
as follows to override /etc/finit.d:
rcsd /etc/factory.d
Manually verified in myLinux
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Since the s6 readiness notification mechanism closes the notification
socket to Finit (as per s6 spec.) the 'ready' condition should then
always be asserted, even across system reconfiguration events.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Services that have not been changed, or otherwise needed to be reloaded,
need to have their 'ready' condition reasserted on 'initctl reload',
otherwise it will remain in 'flux'.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This change adds support for JSON output to `initctl status` using a
`-j` or `--json` command line option.
Example:
root@anarchy:~# initctl status -j mdevd
{
"identity": "mdevd",
"description": "MDEVD Extended Hotplug Daemon",
"status": "crashed",
"exit": { "code": 100 },
"origin": "built-in",
"command": "/bin/mdevd -C -O 4",
"restarts": 10,
"pidfile": "/run/mdevd.pid",
"pid": 0,
"user": "root",
"group": "root",
"uptime": 0,
"runlevels": [ "S", 1, 2, 3, 4, 5, 7, 8, 9 ]
}
The excellent tool `jq` can be used to extract certain parts of the
output for further scripting. E.g. `initctl status -j foo | jq .exit`
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This was added for the benefit of `initctl status foo`, but we have
other users of these functions that don't expect a leading slash.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The condition subsystem starts very late, this should not affect running
callbacks in plugins. In d810ac0 an extra return statement was included
in a new if() statement block that run plugins hooks. This caused a
massive regression in how Finit starts up and runs.
- drop return; statement that causes regression
- move #ifdefs to surround if() statement
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This change adds support for running hook scripts at two new hook points
during the shutdown process:
- HOOK_SVC_DN: after all services and non-reserved processes have been
killed (and collected)
- HOOK_SYS_DN: after all file systems have been unmounted, *just prior*
to Finit calling reboot() to shut down or reboot the system
See doc/plugins.md for details and an example.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Leave the system as it was when we found it. Also, we'll build things
on top of cond_is_available() to handle pre bootstrap and pre shutdown
tasks.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Always use the configured restart delay for crashing services. If no
delay is configured, we default to an initial 2000 msec for forking
daemons and start-stop scripts, and 1 msec for non-forking daemons.
We must track the increasing delay in the svc_t because processes can
fail quickly in differing ways. E.g, the test daemon 'serv' behaves
extra evil by crashing right after having created its pidile (i.e.,
when it's signaled to Finit it is 'ready' and all is dandy ...). If
we don't track the delay per svc_t we would restart 'serv' after only
1 msec every (other) time.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
When there is no log daemon, and we are running in a container, or we
cannot log to the kernel ring buffer, then we log to stderr. This patch
improves the log output by prefixing each message with a timestamp.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
We want to stress test Finit as much as possible. We add a small log
though (which admittedly gives us a little bit of a delay) just to see
what happens when debugging failing tests.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This patch adds the oncrash:script option to call the post:script
action, if defined, for a crashing service. The EXIT_CODE variable
sent to the script is set to `crashed`.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
By definition, the readiness notification is reserved for daemons. May
possibly be extended also to sysv services later. If you arrive hare
with that request, please file an issue with the project.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Follow-up to abcb3ce, calling the service ready:script when readiness
has been signaled to or detected by Finit.
Issue #300
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Always better to have at least one sulogin available, if the system
provided sulogin isn't available we fall back to the built-in one.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This patch adds service readiness notification to support daemons
employing systemd and s6 notification. Complementing the native
Finit readiness support using PID files that exist already.
The two have slightly different ways of implementing readiness:
- https://www.freedesktop.org/software/systemd/man/sd_notify.html
- https://skarnet.org/software/s6/notifywhenup.html
Finit now provides both a NOTIFY_SOCKET environemnt variable, for
systemd, and a way to start s6 daemons with a descriptor argument.
For details on the syntax, see the `service` documentation.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Before we clear all conditions for a service, ensure pid_file_read()
didn't fail due to the PID file not (yet) containing a valid number.
If the process dies while starting up, and subsequently removes its
PID file, then the file shouldn't exist. However, if we get inotify
before the process has finished writing the PID, then we just return
and wait for the next event.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This was a tough nut to crack. Spent 3-4 calendar weeks (getting blurry
now) to try and figure it all out. In the end, the following changes
were necessary:
- Ensure setup-root.sh is guaranteed to run before any of the tests
- Give all tests their unique /etc, /tmp, and /run inside the chroot.
Turns out tmpfs+overlayfs can be used unprivileged inside an unshare
- A static BusyBox binary with support for mount helpers so we can
do the mkdir magic in skel/etc/fstab for the /etc overlay
The last item turned out to be a bit of a roller coaster ride of its
own. First of all, the original binary we used was from the upstream
BusyBox project[1]. I was sure it couldn't be that hard to enable the
CONFIG_FEATURE_MOUNT_HELPERS ... oh boy was I wrong. To make long and
tedious story short; two new projects on GitHub were created for this
task: 1) troglobit/misc[2] to house a mirror of https://musl.cc
toolchains and 2) troglobit/busybox-builder[3] to download, patch, and
build the thing using a .config from the myLinux[4] project. Patching
this .config was necessary, however, since musl libc is strict POSIX
and does not have any of the BSD extensions, e.g. REG_STARTEND that is
in GLIBC regex(3).
[1]: https://busybox.net/downloads/binaries/1.31.0-defconfig-multiarch-musl/
[2]: https://github.com/troglobit/misc/releases/tag/11-20211120
[3]: https://github.com/troglobit/busybox-builder/releases/tag/1_35_0
[4]: https://github.com/troglobit/myLinux
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This delays the start of the .conf monitor and service initialization
slightly to allow the event loop to run earlier to process any events
from the initial setup.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This patch renames the internal states for run/task/services to avoid
any confusion with the introduction of 'ready:scripts'.
* WAITING -> PAUSED
* READY -> WAITING
A service condition that used, e.g., <service/foo/ready> should now
instead use <service/foo/waiting>.
Note: A new condition with the old name <service/foo/ready> will be
introduced shortly to signify service "readiness", a concept
used in other PID 1, like systemd and s6.
For details, see issue #299.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
In Finit a daemon can signal its readiness by creating/touching its PID
file. An ancient UNIX concept -- a daemon creates its PID file when it
has set up signal handlers and is ready to receive IPC (signals), right
before entering its while(1) loop.
Finit took the concept a bit further, adding readiness signalling also
to daemon's by touching their PID file after having processed a SIGHUP.
For both these conditions Finit now supports a ready:script, called when
readiness is detected. In later commits support for s6 and systemd
style readiness notification will be added that will hook the mechanism
implemented in this commit as well.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This support API is for running things like pre:script, post:script and
the upcoming ready:script for services. A ready:script will be called
when a service notifies Finit of its readyness.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This patch adds the mdevd[1] plugin from Andy's fork of Finit, only
minor whitespace adjustment and group fix introduced, along with
updates to configure.ac and the documentation.
Currently the hotplug.sh calls mdevd-coldplug (instad of mdev -s), but
since mdevd is not guaranteed to have started and be ready, it is also
started with mdevd -C. A better solution is to use the s6 notify code
to synchronize the start of mdevd-coldplug. Support for this is coming
soon to Finit.
[1]: https://github.com/hongkongkiwi/finit/blob/master/plugins/mdevd.c
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Service/task/run are today identified by name:id this means the path to
a command is not necessarily the basis for the identity. Hence, an
svc_t can now be updated with new the correct path to a command.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
As a real life example, this can be useful for a mfgtest mode for the
device, wherein the bootloader can be set to finit.cond=test_mode or
finit.cond=user_mode and then specific services can start up in test
mode which wouldn't normally start.
Or even custom firmware update mode such as finit.cond=fw_update_mode
which performs some specific firmware update steps and doesn't launch
normal processes.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
In fbcd7d2 and 96257cc the state machine transitions were changed to
ensure that pre:scripts only run when conditions have been asserted, and
that manual:yes run/tasks end properly in 'DONE' as their siblilngs.
Also include SVG in repo for reference since that's the best export from
the original at draw.io
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This is a follow-up to #274, in which Andy pointed out that ending up in
state 'stopped' was a bit confusing for manual:yes run/tasks.
With the following change we let all run/tasks end in state 'done', yet
still allow manual:yes ones to transition to 'ready' when the user calls
`initctl start foo`.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The whole point of pre: scripts is that they run just before the actual
run/task/service process, hence they need to be locked behind the same
conditions as the process. Otherwise pre: scripts would be nothing more
than plain tasks.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
As it turns out, not all buses in Linux will add `modalias` attributes
to their devices in sysfs. One notable exception are MDIO buses. The
plugin's scan routine would thus not pick them up.
Fortunately, the information is always available in the device's
`uevent` attribute.
Therefore, default to looking for modaliases in the device's `uevent`
attribute instead, falling back to `modalias` only if `uevent` is not
availble (this should never happen).
Since the modprobe plugin (unsurpricingly) may cause new modules to be
loaded, it makes sense to to delay the initial device node scan until
we know that no mode nodes will show up within the next few
microseconds.
This is a fix to a regression introduced in e51587c. The idea is to
have 'initctl status foo' show all instances of foo, provided there are
more than one. E.g.
root@anarchy:~# initctl -t status foo
0 foo:1 stopped [--234-----] hej knekt
0 foo:2 stopped [--234-----] hej snigel
0 foo:3 stopped [--234-----] hej kalas
And not foobar:
0 foo:1 stopped [--234-----] hej knekt
0 foo:2 stopped [--234-----] hej snigel
0 foo:3 stopped [--234-----] hej kalas
0 foobar stopped [--234-----] hej foobar
If you type an exact match you can drill down like so:
root@anarchy:~# initctl status foo:1
Status : stopped (code=exited, status=0/SUCCESS, manual=yes)
Identity : foo:1
Description : hej knekt
...
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Prior to this change only service stanzas respected the manual:yes
option, now it (should) work on any svc_t type. Not tested on ttys.
The initctl tool has been given an extra manual=yes output for these
types of run/task/service entries, shown only when the manual option
is set.
Example:
root@anarchy:~# initctl status foo.sh
Status : stopped (code=exited, status=0/SUCCESS, manual=yes)
Identity : foo.sh ~~~~~~~~~~~~
Description : Hej foo
Origin : /etc/finit.d/enabled/foo.conf
Environment :
Condition(s):
Command : /root/foo.sh
PID file : none
PID : 0
User : root
Group : root
Uptime : N/A
Starts : 1
Restarts : 0 (0/10)
Runlevels : [--2345----]
Notice also the new `Starts : 1` which is a counter for the number of
starts in the current runlevel. On runlevel change it is reset to 0.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
- Prefix plugin defines with DBUS_ to allow overriding at build-time
- Create missing directories for modern d-bus
Signed-off-by: Andy Savage <andy@savage.hk>
This change makes it possible tell the bundled/built-in getty to call
/bin/login with the '-p' flag, to preserve any environment set in the
Finit .conf files.
Issue #286
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Allow filtering of conditions in raw dump, i.e., the possibility to show
only conditions of a certain type. The TYPE argument is not validated
in any way, so any prefix match is allowed, e.g.
initctl cond dump service/s
Shows all conditions for services starting with 's'.
Fixes issue #272
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
When adding a new service option recently, I copy-pasted and forgot to
update the length argument to strncasecmp(), so it was time to fix that
once and for all.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This change allows endless restarts, similar to `respawn` but honors
`restart_sec`. There is no upper limit on the number of times Finit
tries to restart a crashing service. Same as `restart:-1`
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This seemed odd at first, but it turns out we seem to have adjusted the
time scale for the bootstrap worker, so we were off by ... a factor 10.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Move the stop+start from initctl to the state machine by allowing
stopped tasks to restart once it has been collected. This should
prevent finit from blocking, allowing it to handle other requests
while waiting for the service's PID to be collected.
Effectively, this will allow a service to call initctl to restart
itself, as reported in issue #280.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This change diversifies the exit codes used by initctl on fatal errors.
Previously only Std C codes 0 and 1 were used, with the initctl specific
255 for some cases.
Both the LSB init script codes and BSD sysexits.h codes are used. See
util.c exitcodes[] for the full list. The idea is to complement the
change in 9a76f08, issue #266, so that initctl can be used silently for
scripting purposes.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This patch introduces ERR(), ERRX(), WARN(), and WARNX() as replacements
to their lower case counterparts. These log macros respect the 'quiet'
flag to initctl, leaving only the return code as indicator to user.
Fixes#266
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
No need for the verbose output on error, simply state the cause of the
problem. The user can issue `initctl -h` to check the syntax of the
command.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Drop `| tail -10` from 'log foo' command. Let user call tail themselves
and only limit with tail on status command.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Grep in /var/log/syslog, or messages, for both the service identity (not
just svc->cmd) and also try to find [PID] for more relevant information.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
- Drop basename() of svc->cmd from remaining code
- Replace uses of svc->cmd with svc_ident(svc)
- Use svc_ident() as default log tag
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
What a ride, apparently we support restarting run/tasks in the same
runlevel they have already started once in. Still unsure if adding
support for reloading a particular Finit .conf file for a service
should be in the context of this command (it isn't now). Maybe it
is better left as a future expansion when we switch to the new .conf
format?
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
A user with no cgroup support in kernel does not know what they are
missing, and a user running in a container has someone else managing
cgroups.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Add a plugin that will allow the user to trigger the execution of
scripts from plugin hook points. This is particularly useful for early
boot debugging that needs to take place before regular services are
available.
For example, let's say that you want to enable some kernel tracing
before modules are loaded. With hook-scripts, you can just drop in a
shell script in /libexec/finit/hook/mount/all/ that will poke the
right control files in tracefs.
Init script return pretty quickly after launch, so we cannot use that to
detect if they, or the daemon they launched, have crashed. So we start
the service_retry() with a 2 sec delay, which should be ample time to
allow the pidfile plugin to detect a forking service's new PID.
In release-cycle regression, no public issue needed.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
In some conditions, typically when the same command is used for multiple
services, e.g. the modules-load plugin, the svc_find() function returned
an existing "similar" entry instead of NULL, causing loss of config.
When creating, and searching for, a run/task/service we must follow the
new name:id paradigm to the letter. Always create based on name:id and
always search for matching name:id. The name may be derived from the
command, but they cannot be used interchangably.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This change makes it possible to see the new service's resulting name
and id before we create it with svc_new(). We extend the debug output
slightly to see what's going on and why we sometimes lose modules from
the modules-load plugin ...
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
For many usecases, e.g. the modules-load plugin, the same command can be
used to create multple run/task/services makeing it real difficult to
debug such systems when Finit does not show the correct task name. It
also hides nasty bugs, as will be shown in a later commit.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This adds support for a `set` directive in the modules-load plugin.
Currently only `set noindex` and `set index = NUM` are supported and
either disable indexing (:ID) of modules, or reset the index to NUM,
for the given file. For each new file read the index is reset to
the value it would have had. E.g.
01-foo.conf: => name:id
esp4.ko => modprobe.esp4.ko:1
set noindex
deflate.ko => modprobe.deflate.ko
02-bar.conf:
nls_iso8859-15.ko => modprobe.nls_iso8859-15.ko:3
nls_utf8.ko => modprobe.nls_utf8.ko:4
Fixes issue #257
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Drop the config.h include from helpers.h after a report from a colleague
trying to build an external plugin from the latest GIT sources.
Instead, make sure config.h is included, and properly commented, in all
.c files that have configure #ifdefs and other deps. Also, move more
ot the includes from helpers.h to their respective .c file instead to
reduce the amount of headers an external plugin pulls in.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The bundled sulogin could be considered insecure, so leave it up to the
administrator, or system integrator, to decide which sulogin(8) is best
suited.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Finit can now track and set environment variables from finit.conf. The
syntax is pretty much what you'd expect:
foo=bar
baz="qux"
On reload of .conf files, all tracked environment variables are cleared
so if `foo=bar` is removed from finit.conf, or any finit.d/*.conf file,
it will no longer be used by Finit or any new (!) started run/tasks or
services.
The only variables reset to sane defaults on .conf reload are:
PATH=_PATH_STDPATH
SHELL=_PATH_BSHELL
LOGNAME=root
USER=root
It is entirely possible to override these as well from the .conf files,
but be careful. Changing SHELL changes the behavior of system() and a
lot of other commands as well.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
When iterating over the system fstab file to call fsck, Finit calls the
helper function ismnt(), which opens /proc/mounts to make sure mounted
file systems are not fsck'ed. Both the main function and ismnt() used
the same non-reentrant getmntent() API which caused ismnt() to set the
fstab pointer for the first out of whack.
This change replaces getmntent() in the two critical functions with the
getmntent_r() API instead.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Some systems have /dev/root declared in their /etc/fastb as /. This is
as special device node created by the kernel based on the root=, and a
few other variables. This may be a symlink, and if so, and on a mdev
system, the symlink may point to the wrong device node ... so we must
try to figure out the major:minor from / and then traverse /sys/block
to find our root device.
Note: this commit also does a bit of refactoring for readability and
also adds a few debug lines that may be dropped before release.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
A subtle regression was introduced in 5593372f, when moving to get
getmntent() family of APIs. Finding the 'ro' flag (or not) in the
system fstab broke and Finit never tried to remount / read-write.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This change basically add this for pre/post scripts:
sh -ac '. /path/to/env; exec ./pre-post.sh'
Meaning the env file can be a simple foo=bar file, which Finit reads and
call setenv() on all variables, and the user do not need to call export
on all variables -- additionally, any functions defined in an env file
can be used by the pre-post scripts, but are ignored by Finit. I.e.,
the service only gets the variables set, no functions.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The swapon and swapoff commands do not support a custom fstab, they
assume /etc/fstab is the only true source of swap devices and files.
This change adds rudimentary (!) support for figuring out any swap
device or file to use from any finit.fstab. Please note, options are
not supported at the moment.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Since a forking service, e.g. a sysv init script, will exit very early
we cannot log the same as for regular services "Starting foo[123]", so
instead we log "Started bar[321]" when we get the pidfile update and
the same when stopping a forking service, log "Stopped bar[321]" when
we collect the PID.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This change adds the following service conditions, which can be used to
synchronize other stanzas:
- service/foo/running
- service/foo/halted
- service/foo/missing
- service/foo/crashed
- service/foo/stopped
- service/foo/busy
- service/foo/restart
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This reverts commit c63279630b which
clearly was not needed to fix the problem. After this revert the
test(s) still pass, but now we can also start bridge-stp, e.g.
bridge-stp br0 start
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Default debug level for the netlink plugin is just too noisy. Let's
drop debug messages (mostly) intended for developers.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
When logged in from sulogin we cannot send IPC to Finit, not even
signals, to delegate reboot. For usability, attempt to tell reboot that
it's in sulogin mode -- provided the bundled sulogin is used -- to let
reboot trigger over to forced reboot.
We cannot do much about initctl, just document this as a limitation.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This lets any callers of client_connect() decide what to do wrt. error
recovery and fallback mechanisms. E.g., reboot may want to check if
it's in sulogin mode, or fall back to sending signals to PID 1.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This adds support for bringing up the system with an alternate fstab at
boot. E.g., when using a primary/secondary setup for boot partitions.
By default /etc/fstab is read, like before, this can now be changed
using configure --with-fstab=/path/to/fstab.primary, which sets the
default that can be overridden using finit.fstab=/etc/fstab.secondary
If mounting, or fsck, fails in any way, Finit calls its own bundled
sulogin, or the system sulogin(8), to let the user handle the issue.
If there is no sulogin available, Finit will try to start up in its
rescue.conf boot mode.
Please note, in either of these rescue modes, use `reboot -f` to get the
system to reboot. Finit is on pause in the background in rescue mode
and cannot be relied on (since there may not be any writable filesystems
available.).
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The iterator works on any fstab like file, but to keep confusion to a
minimum, rename local variable.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The setfsent() family of APIs was never standardized and on Linux only
GLIBC implement them. We've tried to overcome this limitation, and to
support uClibc/uClibc-ng and musl libc, by providing replacements APIs
in helpers.c.
However, and since we want to support alternative /etc/fstab files, the
setmntent() family of APIs is more widespread and supports reading from
any fstab or mtab file.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Calling `initctl debug` is supposed to toggle Finit debug messages on
the boot console. This broke in ae09272b when improving support for
running Finit in containers.
Part of this change is a slight refactor of who calles log_init() when
starting up, and when to call ttinit(). We now call ttinit() every time
we toggle debug.
Also, toggling back to normal logging had a bug. The new default log
level for Finit is LOG_INFO, but toggling back set it to LOG_NOTICE.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Usability.
Remains backwards compatible for now, but finit.show_status is as of 4.3
deprecated. Likely to be removed int 5.0
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Explicitly set arg to non-NULL value. The call to conf() later will
steer up the value to point to the full path of the user's finit.conf
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
When we added support for using popen() to run(), to log the output, we
forgot to update the error handling path. Found by Coverity Scan.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
We don't want repeats of issue #236, so let's start by tracking down any
bugs hidden in logging functions.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
On regular desktop systems, like Debian, the dbus-daemon forks off a
dbus-launch process to start up things like your desktop for you. This
process is not known to Finit and lingers in the background since it's
been reparented to init, and to top things off it seems to ignore any
SIGTERMs send to it.
Obviously this makes a shutdown on such systems very noisy, so this
patch changes the print() to a _d() so anyone debugging a system can see
it with `-- finit.debug` on the kernel command line, or `initctl debug`.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This split unmount is a bit crude, unfortunately. A user may have set
up a bind or overlayfs mount on top of a tmpfs -- so unmounting tmpfs
first will then always result in EBUSY.
Avoid logging busy errors and fake OK. We'll catch it later in the
second stage unmount.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Turns out that on Debian `ifdown -a` can block for quite a while at
shutdown/reboot. So we'd like to use the --force option. However,
the BusyBox ifdown tool doesn't support --force, only -f, which in
turn the regular ifdown tool doesn't support.
Regardless, we can allow network shutdown to run in the background
like bring-up, at reboot we want to reboot quickly and don't care
so much, and on runlevel change to single-user mode we can allow
for some lagging behind in the background.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
- Mention why configure script is missing from GIT
- Adjust command line for enabling Finit debug
- Plus some minor formatting issues
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Useful to debug system events after the system has started up and you
don't want to rebuild Finit to add the `-d` command line flag.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This is needed since 5ff27c4f when support for debugging Finit and the
kernel independently was added.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
A forking service, e.g. a SysV init script, that returns is no longer
running, even though the monitored service may still be starting. We
must mark the svc->pid as terminated until we know more -- otherwise
we may stall on a shutdown at that exact point.
Also, when shutting down, and stopping all services, ensure we do not
start the carousel for non-existing PIDs. This might also cause our
stalling at shutdown.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Instead of having to create files, and copying them in place for each
test, we move all static test files to a skeleton rootfs. This makes
it a lot easier to get an overview of how things and how they work.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Note: we need to add /usr/bin and /usr/sbin to the standard PATH
for tests and the test environment itself.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
In https://github.com/actions/virtual-environments/commit/12fa229 the
ubuntu-latest runner was updated to Ubuntu 20.04.4, with Linux kernel
version: 5.13.0-1014-azure. This broke the Finit tests completely
and the only possible solution, for now, seems to be reverting back
to the ubuntu-18.04 runner instead.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
In an upside world, much like the Finit test cases, the root may be
relocated. This make /var/run relative to /var, instead of /. Which
hopefully is safer and covers more use-cases.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
All tests except the first two now suddenly started failing. Might be
the new caching support for the busybox binary, but that works on other
clean clones -- working theory now is that there's something subtle
with the setup-root.sh test -- which PASSes ...
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
- Add `type:forking` service option to trigger guessing pidfile to
watch for, instead of `pid:!foo` option, which is not intuitive.
This option may likely also survive into the new file format :)
- Update docs and add examples
- Update start-stop-serv.sh test case with this new variant
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Before this fix, a service declared with respawn could not be changed at
runtime to remove the respawn flag.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
While without a working Internet connection today I ran into the issue
of not being able to run the tests. This adds a basic caching mechanism
to setup-root.sh which saves busybox-x86_64 in ~/.cache, if available.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
serv: add support for foregrounding, running with and without PID file,
including custom PID filename.
test: verify Finit can start & monitor services that:
1) Fork and creates a PID file in a known location
2) Don't fork and don't create a PID file, but Finit does
3) Don't fork but create a PID file
4) Don't fork and create custom named PID file
Note: Finit cannot support a service that forks and doesn't create a PID
file. This combination is impossible to support without tracking
all processes created in /proc -- which Finit does not do atm.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Calling sync is not needed, remount does this for us.
Remount with 'dummydev' causes warnings and is not needed. It appears
sysvinit used this to try and fix a sparc related bug. Instead, use the
rootfs keyword to ensure we don't accidentally remount a bind mounted /.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Coding style for shell scripts:
- Tabs for indent (change in Emacs needed)
- Braces on their own line, like C functions
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
No point waiting for SysV start/stop scripts to finish, PID 1 should not
risk getting blocked forever by broken scripts. Instead we fork them off
and forget about them until they terminate. If they are buggy and don't
finish, it's the problem of the admin.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
I'll have to ask Jacques what this was for, because it doesn't seem to
be needed to run and monitor the test. Also, it lingers at shutdown,
causing some unintended side effects in Finit.
Commenting out for now, as a reminder to myself.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
When a SysV init script starts a daemon, Finit knows nothing of the PID
it should monitor. The PID is written, by start-stop-daemon or the
daemon itself, to the PID file. Finit monitors for new PID files and
can match the PID in such files with an svc_t.
For the regular use-case, we prefer first looking up the matching svc_t
based on the PID -- assuming we start and monitor the service. As a
fallback we resort to mathching the svc_t's declared PID file with the
new file we just discovered.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
We want to find the PID of the daemon the init script starts, so we need
a way to declare this to Finit.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
SysV init scripts should not go to "done" state but "halted" so we can
do: `initctl stop foo; initctl start foo`, like we do for our regular
monitored services.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Install start-stop-daemon in test root. Add S01-service.sh, which uses
start-stop-daemon to start service.sh. Modify service.sh to respect
signals, and not exit immediately when sleep exits due to SIGTERM.
Remove PID file in signal callback and make sure to exit OK
The test itself is basically a copy of the start-stop-service.sh test.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The slay script checks for common errors and gives some logs and status
of Finit when something goes wrong. Helps detect issue #226 when the
start-kill-service.sh test runs at 100000 laps.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
When running in a container we still want to use any syslog daemon
available for our logging needs. However, the time between the first
logit() in Finit and any such daemon having started can be long. In a
normal (non-containerized) setup we log to the kernel ring buffer, but
that's not available in a container scenario. At least not for
unprivileged containers. So we need to detect all these cases and be
prepared to fall back to log to the console, either using these LOG_CONS
flag to openlog(), or by simply calling vfprintf() to stderr.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This trick can also be used by others who want to run Finit in an
unshare. Set the container environment variable to 'unshare',
like lxc and docker do for their products.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Turns out the kill(2) syscall returns ENOENT, not ESRCH, in our test
suite. Don't know why, the man page never mentions ENOENT, only the
ESRCH code. Let's check for both, either way ithe PID is not there.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
A service may have unexpectedly died, and we never got the signal, so
when stopping services we must set the new state after we've tried to
stop the service. Otherwise the svc_set_state() function starts a
background timer for the SIGKILL job, which may block a reboot.
The kill() syscall tells us if the service was there or not, if not we
must clean up and go to HALTED state.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
At startup (and reconf) of systems with lots of services there is a risk
of losing inotify events, e.g., PID file creation/delete events. This
patch increase the receive buffer (doubles it).
On Linux the getsockopt() for SO_RCVBUF returns double the set size, due
to housekeeping in the kernel. So we don't have to do any adjustments
when setting it.
Issue #226
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
In some (error) cases the PID known to Finit may no longer exist, or may
not have been added to a cgroup (yet). Handle this case by skipping the
output of cgroup info in such conditions.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
On, e.g., a container based system logs may not be available so check
also that the messages fallback exists instead of causing confusing
error in the status output for the service.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
To silence warnings at startup/shutdown, check for existance of swapon
and swapoff before calling them.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
When a run task is started with svc->started = 1, it should be
considered started successfully or failed on the other hand.
Signed-off-by: Sergio Morlans <sergio.morlans@atlascopco.com>
Signed-off-by: Ming Liu <liu.ming50@gmail.com>
This patch adds support for optional logging of output from all run()
commands. For run_interactive() we've opted to log instead of just
redirect, meaning output on error is till on console but also in log.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
We've been discussing, over the years, that we'd like to have an easier
way to debug bringup with Finit. Network bringup is one such case where
it's hard to debug if/what you've misspelled in /etc/network/interfaces
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This fixes a seemingly long-running bug; we only brought down networking
in shutdown/halt and for runlevel 1, single-user mode -- not reboot.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
In commit f0f358a13:
[ service.c: set/clear condition 'done' for run tasks ]
a runtask done condition would be set/cleared when entering DONE/HALTED
states, but it did not cover all the user cases, for instance, sometimes
an end user may want to know if a runtask has finished sucessfully or
to decide what to do on its failures.
So we now change the conditions to: tsktype/tskname/success and
tsktype/tskname/failure.
And this change not only applies to run/task types, but also applies to
sysv type, in case it fails, a sysv/tskname/failure condition would be
set.
Signed-off-by: Ming Liu <liu.ming50@gmail.com>
Refactor new kill/shutdown implementation from 3e0063e to fix the
regression in compiler output:
sig.c: In function ‘kill_callback’:
sig.c:222:12: warning: cast from pointer to integer of different size [-Wpointer-to-int-cast]
222 | kill(pid, (int)context);
| ^
Also, some minor renames and simplificactions to match project style.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
- Check all pointers
- Declaratons always at top of func/scope
- Use established variable nomenclature
- Skip useless if() stmt
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
- Comments preferably at beginning of func/sect
- Reorder code slightly, add whitespace for readability
- Drop useless comment
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This commit implements a test for the signal feature of initctl. The
existing service used in tests (`common/service.sh`) has been extended
with a signal trap handler. When SIGUSR1 is received, it will write the
string `'USR1'` to `/tmp/usr1.log`. The test will run `initctl signal
service.sh SIGUSR1` and assert that the contents of the file
`/tmp/usr1.log` really is `USR1`.
The implementation looks up the named service by using
`svc_parse_jobstr`. The callbacks for `svc_parse_jobstr` has been
augmented to accept a user data parameter. For this use case,
a carrier for the actual signal was needed. The address of the
signal parameter is taken and passed on as a `void *`. The
callback then simply deferences it as an int - the signal number.
In the test framework for finit, a set of shell scripts are used to
setup a test harness, run tests, and finalize tests. In the setup, a
file system is created for the virtualized/containerized environment in
which all tests are run, located at `$FINIT_SRC/test/tenv-root`. In the
tear down phase of each test, the file `/var/lock` is made readable
using `chmod +r`. It is not clear _why_ the teardown code does this, as
there are no references in the test framework to this path anywhere
else. It is conceivable that the teardown phase attempted to "reset the
state" for next test.
The code that that this commit removes does not always work. When
`/var/lock` is a symlink, and resolves to an absolute path, the test
framework does not function properly. The teardown code is run in the
context of the host computer, and touching files outside of the
virtualized environment is not ok.
The removal of the offending code does not seem to affect tests: all
tests pass without it, so its existence is questionable.
Signed-off-by: Jörgen Sigvardsson <jorgen.sigvardsson@gmail.com>
Blocking SIGTERM means Finit will wait another two seconds before it
sends SIGKILL, which casues a delayed [WARN] Killing ... before we can
proceed to shutdown/reboot -- not OK. Don't know what I was smoking
back in 2017, but it must've been good.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
When running in a container we might not have the necessary privileges
to mount cgroups. Don't cause error in this case, just log it.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This patch allows disabling (enabled by default) the modprobe and tty
plugins. They are not particularly useful in container use-cases.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
We extend the service.sh to emulate a well behaving service that creates
and removes its own PID file.
The new test emulates a crash by sending SIGKILL to service.sh. We then
verify that Finit restarts it, and eventually registers the new PID when
the services recreates the PID file.
Issue #226
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
- We need a working /var and /run for some features, e.g. PID
monitoring and status files for urandom.so
- Add empty fstab to silence mount errors from finit
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
A regression was introduced by commit f0f358a1:
[ service.c: set/clear condition 'done' for run tasks ]
svc->type is a integer but mistakenly being used as string, which will
cause crash.
Signed-off-by: Ming Liu <liu.ming50@gmail.com>
Instead of unconditionally waiting 2 seconds for processes to die,
check continuously for remaining processes, and break the loop when
none remain.
Turn PID 1 to a RT process with highest priority 99 during shutdown,
this ensures it would not be preempted by other RT processes.
Signed-off-by: Robert Andersson <robert.m.andersson@atlascopco.com>
Signed-off-by: Mathias Thore <mathias.thore@atlascopco.com>
Signed-off-by: Ming Liu <liu.ming50@gmail.com>
We like to know when a runtask (run/task/sysv)is done, so as to start
the services that depending on it, set a oneshot condition 'done' for
that case and clear it when runtask services enter HALTED state.
Signed-off-by: Ming Liu <liu.ming50@gmail.com>
It should not use relaxed comparison in some places, otherwise it mess
up with scenarios like:
```
$ initctl status ntpdpre
$ initctl status ntpd
```
the output might be same for both of the above commands without this
fix.
Signed-off-by: Ming Liu <liu.ming50@gmail.com>
This patch adds the -v,--version option to initctl, which acts as an
alias to the 'version' command. For now they end up in the same old
function, printing the program version. However, in time this will be
changed so that the 'version' command asks a running Finit for its
version instead of assuming the daemon is of the same version.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
A 'module foo args ...' line in finit.conf can be looooooong. This
patch drops the module args from the progress output for brevity.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
With the introduction of GLIBC v2.34, all 32-bit devices running Linux
get the option of having a 64-bit time_t to survice the UNIX 2038
apocalypse.
Support for 64-bit time_t was added to libuev in v2.4.0. The build
flags for linking against libuev are defined in the library's .pc
files, which basically gives you -D_TIME_BITS=64 for your CPPFLAGS.
However, it turns out the $(uev_CFLAGS) from pkg-config were only used
when building the main finit binary -- not for the plugins, causing
some really wacky behavior as a result. All plugins and Finit itself
had some very different opinions on what `struct plugin` looked like,
including offsets and size. This led to all I/O plugins initializing
their sockets at the wrong offset -- little-endian 32-bit targets set
the HOOK_SHUTDOWN callback hook to address 0x1 -- which the reboot
code then tried to execute, of course leading to a segfault.
This is quite possibly the root cause of issue #216
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Previous patches only checked if cgroups v2 support was available in
the kernel. If mounting failed cgroups support was disabled.
However, much of the core logic to cgroups in Finit revolve around the
CPU controller. This patch adds checks to ensure that the kernel has
basic controller support as well.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This patch prevents bogus error messages from Finit for when
directories or symlinks already exist.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Traditionally we've reset the service restart counter on two occasions:
1) when a service reaches its max crash count (default 10), and
2) when a service somehow "stabilizes"
In the second case the service is restored to normal running state and
we "forget" its bad previous behavior. Meaning we cannot catch daemons
that act in an unstable manner outside the "rage quit" scenario.
This patch allows for slowing aging (decrementing) the restart counter
once every five minutes. Meaning we still catch rage quitters but now
are also able catch other types of misbehavior.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This patch adds a recent feature request to keep track of the total
number of restarts (including crashes and initctl restart) of both
service and sysv daemons.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Clearing of the svc->pidfile was introduced in e1b87d70 for a
restriction that has now been removed.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This patch adds support for detecting if the Linux kernel Fint runs on
has support for cgroups v2. The probe is done very early, in runlevel
S, when attempting to mount /sys/fs/cgroup. Meaning, the log message
may be lost on systems that do not run sysklogd[1].
Here's the relevant logs from a system running myLinux on a kernel
that has cgroups v2 support disabled:
Jan 10 20:19:40 anarchy finit[1]: myLinux 2021.11-696-g8fddaf36ed, entering runlevel S
Jan 10 20:19:40 anarchy finit[1]: Kernel does not support cgroups v2, disabling.
Jan 10 20:19:41 anarchy finit[1]: myLinux 2021.11-696-g8fddaf36ed, entering runlevel 2
All code paths that attempt to set or query kernel cgroups config has
been short circuited when this happens. However, this does not apply
to any cgroup related config in the Finit .conf files -- which may of
course be omitted. Yet, if configuration directives exist, they will
be parsed and may cause error.
Also, the initctl tool checks if /sys/fs/cgroup is mounted at runtime
and disables all status output and commands that rely on the feature.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
initctl cannot link with helpers.c, so let's relocate these helper
functions to util.c instead. Need them to probe for cgroup support.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
A timeout is not the same as a restart delay, which this option really
controls, but the name restart_delay is a bit too long. Hence, this
obvious name so there's no ambiguity what unit we're dealing with.
Keeping restarttmo:SEC as a compatibility setting for the benefit of
the original authors, for now.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Minor coding style cleanup. Also, clang complained of a signed vs
unsigned comparison in service_retry(), max(timeout, svc->restart_tmo)
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
As suggested in issue #197, this new command: 'initctl id [NAME]'
lists identities of all, or named, services. E.g,
# initctl id telnetd
telnetd:23
telnetd:2323
The NAME argument is optional, when omitted it lists all enabled
run/task/sysv/service in the system, with their full name, and any
optional identitty. E.g.
# initctl id
dropbear
tty:console
mdnsd
mini-snmpd
smcrouted
sntpd
ssdpd
sysklogd
watchdogd
httpd
telnetd:23
telnetd:2323
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
When the system has multiple services 'foo' the initctl command
'status foo' should give a focused overview of all matching foo:*,
whereas a 'status foo:2' should show the detailed view of that
particular instance.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Since Finit v4.0 kernel console output has been silenced just prior to
the banner being printed, unless finit.debug mode was enabled. This
patch makes sure to check the kernel loglevel, if it's >= 7, finit no
longer disables kernel console ouput.
Consequently, you now need 'quiet' on your kernel command line, or a
loglevel < 7 for a quiet boot, at least until a syslogd takes over.
This means that you now can debug the kernel and finit independently,
which should be a great comfort to anyone doing bringup or any form of
kernel debugging.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This patch switches from "cp /var/lib/misc/random-seed /dev/urandom"
to use the kernel RNDADDENTROPY ioctl, which actually increments the
entropy count.
Also, the amount of random data saved at reboot is increased from the
fairly low 512 to 32768 bytes.
All in all, this should greatly improve the stability of most systems
during, or close to, bootstrap. In particular embedded systems with
limited or no HWRNG.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Since we just want to clear the complete line, and are already a the
beginning of the line, we can use the simpler \e[K which also seems
to fix the display resize bug in #210.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Due to an unfortunate name clash with the DirectFB project LiTE, the
libite (-lite) project had to change its header namespace from
lite/*.h -> libite/*.h
This patch adds support for the new namepace in Finit, triggered by the
define _LIBITE_LITE, from the .pc file read by pkg-config. This should
only be needed on systems that install libite without the compatibility
symlink lite -> libite/ in the staging include directory.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
It would be good if we can control the services more specifically with
the following parameters like:
restart:N - how many times shall the service to be restarted on
failures (<10), the default max is 10.
restart_tmo - the timeout of the restarting.
norestart - dont restart on failures.
oncrash - once all retrying also fail, how to deal with it,
rebooting or ignoring the failures.
Signed-off-by: Robert Andersson <robert.m.andersson@atlascopco.com>
Signed-off-by: Ming Liu <liu.ming50@gmail.com>
Both DBUS user and pidfile is configurable in DBUS compilation, by
passing "--with-dbus-user=" and "--with-system-pid-file=", let's avoid
hard-coding them here, use macros instead.
Signed-off-by: Ming Liu <liu.ming50@gmail.com>
It's configurable in /etc/login.defs as NOLOGINS_FILE, avoid to hard
coding it here, introduce a macro FINIT_NOLOGIN_PATH instead which is
configurable from compiler flags, it's not defined, then back to
_PATH_NOLOGIN from /usr/include/paths.h.
Signed-off-by: Ming Liu <liu.ming50@gmail.com>
fastboot: once enabled, fsck would be skipped on filesystems listed in
/etc/fstab.
fsckfix: once enabled, 'fsck -ys' would be run instead of 'fsck -a',
and check the return value larger than 1 as a failure and goes
to login console.
They are both disabled by default, no functional changes unless an end
user choose to enable them.
The idea derives from Sysvinit's "FASTBOOT" and "FSCKFIX".
Signed-off-by: Ming Liu <liu.ming50@gmail.com>
The iterator function getmntent() is not stable. It may skip
entries if the contents of the iterated file changes, which
is the case with /proc/mounts when filesystems are unmounted.
As a result, some filesystems were never unmounted.
To fix this, iteratation is now restarted after every
sucessful unmount.
Signed-off-by: Robert Andersson <robert.m.andersson@atlascopco.com>
Signed-off-by: Ming Liu <liu.ming50@gmail.com>
"protected" is a C++ key word, replace it with "is_protected" avoid
compiling failures with C++ compiler.
Signed-off-by: Ming Liu <liu.ming50@gmail.com>
There are small typos in:
- doc/conditions.md
- src/mount.c
- src/svc.h
Fixes:
- Should read `satisfied` rather than `satsifed`.
- Should read `process` rather than `prorcess`.
- Should read `measure` rather than `meausre`.
We've had several logit() messages of LOG_INFO level that never
really made it to syslog because we classified it as "debug".
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This patch adds "the missing" first runlevel log entry when Finit has
printed the console banner and begins its bootstrap phase.
At this point we know syslogd hasn't been started yet, so we rely on
logit() to redirect this to /dev/kmsg in the meantime. Later syslogd
will pick this up, see that it's not from the kernel, and log it to
the proper log file.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
If any of these mount() calls fail, not counting EBUSY because that
would mean sth is already mounted, there really isn't much we can do.
So the least we should do is warn on console if it happens.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Note, this is dead code, currently unsused in Finit. Possibly an
external plugin makes use of it, but even that is highly unlikely.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
I'm writing this in X started on a system running Finit v4.1-pre that
does not have elogind installed. It does, however, still have systemd
installed, so systemd-logind is available.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
- Fix bug when calling sys_update_conds() when dir already exists
and we're not being called from the scandir() handler. I.e.,
every time but the first for each new condition sub-family
- Drop sys conditions that don't affect any svc_t. This may seem
counterintuitive, and we should probably not use oneshot conds,
but if we leave these conds asserted they may cause inadvertent
trigger if a finit.conf is loaded which ha this sys cond. E.g.,
if ctrlaltdel is asserted and we enable a task in finit.conf and
call `initctl reload`, the task would start immediately, even
though ctrl-alt-del may have been pressed a week ago
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This patch removes the cognitive overhead of having to manually set your
OS heading, --with-heading="Foo OS vX.YY". As of this patch, Finit by
default extracts PRETTY_NAME from /etc/os-release. It is now possible
to also disable the heading entirely using --without-heading
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Shouldn't be needed for real cases, but we have seen automated scripts send
'\n' instead of '\r', or '\r' + '\n'.
Signed-off-by: Jacques de Laval <Jacques.De.Laval@westermo.com>
Similar to 6224166 (previous commit), this old code is a remnant of a
bygone era and not needed anymore.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Once upon a time, SIGSTOP was used to pause Finit during flashing (MTD)
of a system image. This to prevent Finit from accidentally starting any
programs, i.e., reading from flash disk during or after upgrade.
This was quite intrusive, and has possible nasty side effects, e.g., any
process with root access sends SIGSTOP prevents TTY login. So this patch
now removes the functionality and recommends using a dedcicated runlevel
for such critical tasks instead.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
When a service is started and then stopped before it has created it's pid file,
it could be left forever in the "stopping" state, if we don't reset the
starting flag.
Signed-off-by: Jacques de Laval <Jacques.De.Laval@westermo.com>
The waitpid() function can return -1 due to EINTR (signal), so we should
restart it to make sure we collect all zombies.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Some systems may want to handle RTC and /dev/urandom by themselves, or
not at all as in the case of containers where the host does all this.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
For most use-cases the kernel will give Finit its arguments as proper
command line args in argc + argv[], like any other program. However,
for some users, most notably Alpine Linux, there is a slightly broken
initramfs that cannot forward more than one argument using init_args,
for such systems you can re-enable the old behavior with a configure
switch --enable-kernel-cmdline -- it's not ideal but what can you do.
The main reason for removing this feature by default is to support
use-cases where Finit runs as the init for container apps that can read
/proc -- we do not want them to use the init args from the host.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
It has barely worked and only caused more problems than solved annoying
issues. We have one console for output, /dev/console, which the kernel
sets up for us.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Also, fix default: case in error handling, must always continue back to
poll() on any recv() error.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Progress at startup has been hidden for services w/o -- description, but
for some reason this check was not added to service_stop(). This patch
rectifies the situation, finally.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This patch adds a `respawn` flag for services and ttys, always set for
ttys, that allows bypassing the crash/restart counter and immediately
restart a 'crashing' service.
For tty type services this is the expected behavior, but for regular
services it is not. That is why `respawn` flags is not advertised in
the docs.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Basic security measure, don't bail to shell if we cannot find/exec
login, instead try sulogin before falling back to plain shell.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
By default, Linux gives us /dev/console for output. This is a pseudo
device that uses the same actual device as the last console=foo listed
in /proc/cmdline. The last one listed is the main console, which is
also the *first* one listed in /sys/class/tty/console/active, so we skip
that when we check for system consoles to avoid duplicating output.
The getty code in tty.c currently has its own handling of @console,
which we keep for now. Ideally, however, the code should be merged.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
By moving the built-in getty to a stand-alone bundled getty we can now
refactor the old run_getty() functions into a single one.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
When TTYs became first class citizens in Finit 4.0 much of the
boilerplate setup is now down by service_start(). Also, the calls to
TIOCSCTTY, VINTR, and SIGINT ignore is not necessary to do here, it
should be done by the getty used, if any.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
We must detach from the controlling TTY before enabling signals and
setting up new stdio. TTYs have their own handling of stdio.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This patch moves the built-in getty out of Finit into /libexec/finit/,
reducing the size of the Finit binary and simplifying the code.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
When starting a getty Finit checks first that the configured TTY device
actually exists and is a TTY. This patch, by Tobias Waldekranz, ensures
Finit (PID 1) doesn't acquire the TTY device as a controlling TTY. If
that happens PID 1 will get all signals sent to the process actually
started with the device as its controlling TTY.
Signed-off-by: Tobias Waldekranz <tobias@waldekranz.com>
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
When an external getty is used, and an absolute path is not given, the
parser tried to use access() to determine built-in vs external getty.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
If we don't extract the level (disregarding facility), statements such as
"logit(LOG_CONSOLE | LOG_NOTICE, ...)" would never reach the kernel log.
Signed-off-by: Jacques de Laval <Jacques.De.Laval@westermo.com>
This patch drops the default signal handlers for SIGPWR and SIGINT with
optional tasks in /etc/finit.conf, triggered by two new conditions:
sys/pwr/fail and sys/key/ctrlaltdel, respectively.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Apparently fgetln() is available in musl libc, but not glibc. This is a
funciton that first appeared in 4.4BSD. We take a risk that fgetline()
is free in all libc's, getline() is not.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This patch is a refactor of the prototype sys condition plugin. It
moves most of the logic to monitor kernel events into a keventd that,
currently only, sets and clears the sys/pwr/ac condition. The sys
plugin itself is now only a monitor of conditions and ensures Finit
follows them. This is a lot more secure and moves (at least one piece
of) netlink processing out from PID 1.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The top-level cgroup init is a leaf group and should be treated as such,
even for user setup tasks/services.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The udevd, dbus, bundled watchdog, and others were started without a
valid cgroup. This is a workaround to ensure they are assigned one.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This fixes a long-standanding issue with finit not controlling the
reboot when the user presses ctrl-alt-delete (PC systems).
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This patch does several things related to /run and system reliability.
- Mount /run with MAX 10% of usable RAM
- Create and mount /run/lock as a separate tmpfs with max 5 MiB
As a spin-off, this patch also fixes permisions on /run/lock to 0777
so regular users can create lock files.
Note: none of this code runs if /run is mounted alread in /etc/fstab
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Handle EAGAIN properly, for both regular and resync flow, on any error
in the regular flow (unless ENOBUFS) we want to check for nl_ifdown on
any of the successfully parsed messages.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This patch adds support for calling recv() repeatedly to get the netlink
response from the kernel. As a result, the recv() buffer can be reduced
down to 4k again.
Both the regular flow and the resync flow now follow the exact same code
path, except for the ENOBUFS handling. If we get ENOBUFS in resync, we
are screwed anyway.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
For RTM_GETLINK we need a `struct ifinfomsg`, not `struct rtmsg`,
otherwise the kernel will get 8 extra bytes and complain about it.
This patch makes sure to set the correct iface change mask as well, and
increases the debug logs a bit to get a fix on sizes used. We increase
the recv() buffer 8k -> 64k to make sure we can get all data in one big
swoop. Plan is to refactor this mess in a later commit.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This is a major redesign of the netlink plugin to be able to handle
ENOBUFS¹ properly. Pending verification, this change replaces the patch
to increase socket buffer size, which in real life turned out to be
insufficient.
When nl_callback() calls recv() and it fails with ENOBUFS, we consider
our cache of the kernel state invalid and thus:
1. deassert all net/ conditions
2. open a new (temporary) netlink socket
3. send RTM_GETLINK and re-assert all interfaces using nl_link()
4. send RTM_GETROUTE and re-assert all routes with nl_route()
Like before, the kernel will not send us a RTM_DELROUTE when it removes
the default route, so we still have to track this ourselves. This patch
also refactors that functionality to only resync routes when the ifindex
associated previosly with the default route goes down or is removed.
The previous change that added nl_default() to recheck, has been dropped
to instead reuse the standard nl_route() callback.
___
¹ see netlink(7) for details.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This patch fixes the problem with Linux not sending netlink route change
notifications when interfaces for these routes goes down. When an iface
goes down we now send a route request to the kernel and check the return
message, if no default route is found we deassert net/default/route.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This patch restores parts of 4febd28, but in a more orderly fashion, in
order to get a standard Debian system to boot properly with X and tools.
The patch adds fs_finalize() which checks if /dev/sh/, /dev/pts, /run,
and /tmp have been mounted properly from /etc/fstab. If not, then the
function makes sure to mount tmfps (or devpts) file systems as expected
by most modern systems.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This patch changes the behavior of SIGUSR1. Previously USR1 could be
used to halt a running Finit system, similar to BusyBox init. However,
compatibility with sysvinit and systemd has been deemed more important.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Coverity suggests validating against only a set of known characters.
However, the kernel allows just about all characters in an interface
name. This version of valdiate_ifname() is blatantly stolen, more
or less, from linux/net/core/dev.c
https://code.woboq.org/linux/linux/net/core/dev.c.html#1020
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
sulogin should return OK (exit code 0) when the user presses Ctrl-D,
regardless of the prompt shown. Otherwise Finit may look for another
sulogin to run straight after this one.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Simplify build system.
The logit tool is, as of v4.0, installed into /libexec/finit/logit and
thus not part of the system $PATH. If a sysadmin or distro wants to
remove it from the system it's entirely possible since Finit always
checks for logit availability before attempting to use it.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Before 4e5d06b the only way to go into rescue mode was to skip certain
steps and then attempt to load /lib/finit/rescue.conf. After 4e5d06b
we keep the old handling as a fallback in case early sulogin fails.
The new tty option 'rescue' is added, which recue.conf is updated with.
This option implies notty mode and will try sulogin (again) before it
falls back to start /bin/sh as a login shell.
The reason we keep this is to be able to handle all possible use-cases
and also allow sysadmins to set up the behavior that fits their needs.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This patch does two things:
1. A system with a root account that has no password is now
allowed to login without password, as one expects.
2. A systems with a locked root account, is treated as (1).
The latter of the two may very well turn out to be a really bad idea.
Much angst have been poured into it, yet the end decision is to allow
access.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Most major init systems have a way to start a rescue shell. This is
usally started when the string `rescue` is read from the kernel's
command line.
The traditional Finit way of handling this has been to skip certain
phases of the standard boot, skip /etc/finit.conf et al, and go for a
/lib/finit/rescue.conf instead. This pratice has been imroved over
the last few commits with a sulogin. However, there was no way to
resume boot, like most other init systems offer.
This patch implements a very simple rescue mode, utilizing the new
sulogin, replacing the old (and experimental) emergency shell. If
Ctrl-D is pressed at the maintenance login prompt the boot is now
resumed. The same goes for exit/Ctrl-D from the maintenace shell.
The preferred sulogin is the bundled /libexec/finit/sulogin, but
if that isn't installed, the first one in $PATH is used. If no
sulogin at all is found, Finit proceeds as before this patch, to
skip certains steps and go for a /lib/finit/rescue.conf. Hence,
al steps can be controlled by an administrator.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
- splice out sh() as a separate function
- make sh() session leader and set controlling TTY (TIOCSCTTY)
to make it a proper login shell with Ctrl-C functionality
- simplify call to execl()
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
These plugins should not run in rescue mode, because the system may be
in a very bad state and we do not want to make the situation any worse
than it already is.
Essentially, only services in rescue.conf should run in rescue mode.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
- Check if watchdogd *and* `WDT_DEVNODE` exists before registering the
built-in watchdogd at boot
- Update bootstrap.md with this additional constraint
- Update config.md with references to bootstrap and the new constraint
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This patch fixes an issue where services that support SIGHUP are not
properly stop/started on changes to their command line arguments.
A change to a service's env: file, e.g. /etc/default/foo for service
foo, must also be counted as a change to the foo args.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This changes the default to redirect output from run/task/services to
/dev/null. Use --disable-redirect to get the old pre v4.0 behavior.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
We need at least the softdog watchdog driver loaded for proper watchdog
reset at reboot. This can safely be replaced with a hardware specific
module that provides /dev/watchdog ... without one Finit will try to
restart the built-in finit-watchdog service 10 times to no avail.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This patch changes the configure option to enable the built-in watchdog
from --enable-watchdog to --with-watchdog[=DEV]. This is the convention
for features that take arguments.
Also, improve log messages to aid debugging when finit-watchdogd does
not start properly. This means flushing logs to syslogd with closelog()
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This patch restores the start of the built-in/bundled watchdogd. It is
tracked in the `wdog` variable and handled as an exception at shutdown.
This is also a follow-up to 7b74c99, ensuring that we only kill/stop the
built-in watchdog, not any external. External ones can register to be
the controlling watchdogd in the system -- Finit is not the arbiter for
singletons, this is up to the system engineer.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
If we get a notification and the service dies immediately, and also
removes its pid file, we need to take corrective action.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This patch updates the service state machine with two new states: SETUP
and CLEANUP. If an executable pre:/path/to/script is defined for a
service, it is called every time the task goes to READY state. If an
executable post:/path/to/script is defined for a service, it is called
when the task goes to HALTED state.
Each of these two scripts default to a three (3) second execution time
before they are SIGKILLed. This can be adjusted with the `kill:SEC`
option for the service. There are no execution guarantees, nor are
there any way of detecting if the script was killed before completion or
not -- except for running Finit in debug mode and inspecting the result
printed by system_monitor().
Note: the post:script MUST be idempotent since transitions between READY
and HALTED can take place any number of times before a task goes
to its RUNNING state.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
These two BusyBox daemon's run in concert so it makes sense they also
share the same cgroup. In Finit that means sharing the same .conf file.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Before this patch bootclean() ran first in setup() which caused to to
remove the entire /var/run/finit directory, and other files as well,
created earlier. Only possible fix is to split clean and setup in
two and make sure to call clean as soon as we've mounted everything.
Note: this introduces a new behavior, and anyone hooking into the
same point to do good-stuff(tm) may be affected by this.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
- Avoid using reserved C++ keyword 'new'
- Rename new -> next
- Rename old -> prev
- Add debug for value being set to cond path
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
- Drop vhangup(), how did things ever work *with* this in?!
- setsid() + TIOCSCTTY are best buds, see notty code
- Allow storing any non-NULL string as tty->dev, expand in service_register()
- Reorder parse_cmdline_args(), we need to expand @console to current dev
- Fix tty arg parser, swapparoo for external getty
- Refactor tty_atcon(), iterated over by service_register()
- New arg format, need to translate for old run_getty2()
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This patch re-enables the fallback shell handling after the big TTY
refactor. We do this by allowing the fallback shell to run as a
regular service.
Note: this also adds the "hidden" support for 'notty' option for
tty configurations stanzas. This is just to pick up from
where the kernel left us, reusing stdin + stdout.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
As soon as we collect a PID we should cancel any outstanding timer
actions. Because we may very soon svc_del() it ...
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Some C-lib, like GLIBC, convert NULL to <nil> when printing %p,
which we shouldn't rely on since others just bite the dust.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Initial refactor of the tty implementation to use the service/run/task
general backend. This enables all the features of services also for
ttys, except logging because it makes no sense.
Work in progress:
- plugins/tty.c does not work anymore, could possibly be removed in
favor of usinga (a new) condition instead (if-tty-exists)
- fallback tty does not work anymore, should we remove it, or can we
handle it as an optional built-in with (a new) condition?
- @console does not work anymore, needs to generate N cloned services
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Calling `tree /tmp` causes too long execution time and Travis timing out
the job. So let's install to a sane directory.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
- Show if exited/signaled
- Show status code and the std /NAME
- Show signal value and the std /NAME
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
- List /etc/finit.conf
- Allow edit of /etc/finit.conf, using finit/finit.conf or NULL
- Add helpful question to edit /etc/finit.conf if no edit arg.
- Show available files if delete has no arg, like disable
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The main purpose of this patch is to remove top-level cgroups when they
are remvoed from finit.conf. However, this is tricky, becasue there may
still be supervised services (and children of them) running. We must
postpone removal until a cgroup.events change. To make matters worse,
events may arrive out-of-order, i.e. the last-child event for the parent
cgroup before the sub-group.
A spin-off from this patch is that we can now support arbitrarily long
group names and group config. (Only applies to top-level cgroups, not
run/task/services.)
Limitations:
- Processes in "removed" cgroups are not migrated/orphaned, the reasons
are several: overhead and complexity and the fact that memory cannot
be migrated
- A lingering child of a moved service may prevent an old top-level
cgroup from being removed.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
We have these neat extras for added value, if they should fail, for
whatever reason, we still want to present 'top' to the user, which may
have to use 'reset' if exited improperly and lost cooked/cursor.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
We check it everywhere else, and a log message for failing to move PID 1
to the init cgroup is useful as well.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
No need to zero out whole buffer for strlcat(), or other string check
ops. Also drop a few completely useless memset() calls.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
- Set TTY in raw mode to be able read a single char at a time
- Hide cursor
- Trap 'q', Ctrl-C, Ctrl-\ and SIGTERM to restore TTY and cursor
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This is intended to be used with the 'top' command to aid in setting up
and verifying proper limits for services.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Replacing write() with dprint() without thinking leads to duplicate
output in getty. In particular '\n login: foo login: ".
Must check if length is > 0, or dprint() will calculate it on its own.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This patch corrects a logical glitch, or design flaw, in initctl. The
'restart FOO' command did not stop+start FOO only send SIGHUP (provided
FOO supports SIGHUP). Hence, a new command 'reload FOO' is introduced,
which does exactly that, and 'restart FOO' now stops and restarts FOO.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Fixes the GCC warning for unchecked return value from write():
... ignoring return value of ‘write’, declared with attribute warn_unused_result
The new dprint() function wraps write() in a bounded retry-loop that
catches EINTR. The length argument is optional computed with strlen()
if omitted (zero).
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
By keeping track of when a full restart is needed, Finit can relieve
the user of having to track modifications to service configuration. This
also makes initctl reload easier to explain and reason about as the effect
of a reload should now be that any configuration updates to services will
be fully "commited" by initctl reload.
Signed-off-by: Jacques de Laval <Jacques.De.Laval@westermo.com>
This patch adds support for modifying settings for the default cgroups;
init, user, and system, as well as adding up to a total of eight groups
for the system.
Services can now be assigned to a cgroup, with optional extra settings
for that particular process group. The syntax is slightly contrivied
but follows the overall Finit syntax of prop:value,prop':value', e.g.
cgroup maint cpu.weight:123,mem.max:10000
Starting with the introduction of rlimits, a group of services sharing
the same .conf file can share the same (locally "global") rlimits, and
now also the same cgroup, e.g.
cgroup.maint
service foo
service bar cgroup:mem.max:1000
This puts foo and bar in the same top-level cgroup 'maint', with an
extra memory restriction on bar for max 1000 bytes memory.
NOTE: 'mem.' is a Finit extension, a shorthand for cgroups2 'memory.'
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Since the introduction of the iwatch framework we can now safely free
the memory allocated by realpath() and prevent leaks in a more elegant
way than before.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
When reloading we check if service .conf files have been modified since
the last time we read them. In 3bb1e41 we changed it so we load only
the enabled/*.conf files, which led to our no longer properly detecting
changed files since we monitor changes in what the symlinks point to.
This patch fixes that by checking the target for changes.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Since Linux 2.6.4 the SOCK_SEQPACKET socket type is available for UNIX
domain sockets on Linux. It is intended for; "a sequenced-packet socket
that is connection-oriented, preserves message boundaries, and delivers
messages in the order that they were sent." which sounds like us.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The finalize() function is a bit too long, which means it may block the
API socket initctl (from the final runlevel switch at the beginning) use
to contact os via.
We should perhaps break it up even more, but at least once like this to
give the event loop a chance to interleave with an API callback.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
In particular this can happen from cgroup.c when multiple services map
to the same cgroup.events (due to being in the same cgroup)
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
When calling initctl early in the boot process, e.g. just when Finit is
switching to the final runlevel, it may take a few milliseconds for it
to actually be ready to process our request.
This patch adds poll() to both the write(REQ) and the read(RSP) steps
to prevent from getting weird errors when Finit is busy. The timeout
is currently 2 sec in both directions, which should be enough.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The bootmisc plugin creates lots of required system directories which
udev, and possibly also mdev, need to operate. E.g., a system which
has an empty tmpfs for /var need to populate that before we run.
The plugin loader handled this dependency implicitly before, loading all
plugins in alphabetical order. We should not rely on that for proper
operation.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This plugin should be able to start much earlier than on network UP,
it uses a UNIX domain socket to communicate so loopback should not be
needed.
Also, Finit supports runvels up to 9 (0 and 6 are special), so allow
dbus to run in all these runlevels. It is up to the user/OS to set
any policy for what runlevels to use.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
For commands like `initctl show foo`, initctl looks in the available/
subdirectory for `foo.conf`. This patch adds support for falling back
to return `/etc/finit.d/foo.conf` if the service doesn't exist in the
`/etc/finit.d/available/` subdirectory.
Static services, outside of the `enabled/` handling, live in the base
dirctory and this is what you expect from the tool. "Do what I mean"
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Older kernels (4.19) don't have the summary cpu.stat in the cgroup2
root, so let's ignore that problem and return zero.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
In debug mode it's highly confusing to see Finit load available/*.conf
files when you expect it to only load finit.conf and enabled/*.conf,
so let's keep the dangling symlink check to itself and actually load
and parse from enabled/ as it was intended :)
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Sneaky bug, when switching between different sets of configurations, a
new instance of a previously removed svc can have its condition removed
by svc_gc() of the old instance. This patch prevents such accidents.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Having removal status stored in the dirty status resulted in removal
status being forgotten when a service was marked as dirty, for instance
when a dependency was updated. This side effect of marke_dirty seems a bit
unexpected and instead of adding exceptions to the logic for when marking
a service dirty - let's separate the two things (dirty and removed) from each
other.
Signed-off-by: Jacques de Laval <Jacques.De.Laval@westermo.com>
More often than not, the file to write to in sysfs changes rather than
the value. This patch changes echo() into a fnwrite(), flipping what
is vsnprintf()'ed, and adds a stupid str() function that converts any
value (float/int/double/uint64_t) to a static string buffer.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The strlen() function can easiliy go out of bounds. Use memchr()
instead, we have the max buffer len as argument anyway.
Also fix call to sanitize() which used wrong length.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Replaces previous cgreaper.sh functionality for release notification and
cleanup of service groups.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Since the cgroups support in Finit is not yet officially released, we
decided to change the back-end to use cgroups v2 instead of the aging
and rather clumsy cgroups v1. Even this initial refactor is a lot
easier to read and understand, more can still be done since there's
a lot of concepts, data values and the ilk that can now be shared
between different controllers.
Still ToDo: inotify for cgroup.events to clean up leaf nodes, which
in cgroups v1 was handled by cgreaper.sh. This is fixed
in the next commit in the series.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Even systemd have collapsed a few of the cgroup controllers into a
semi-unified hierarchy and uses this approach. We just take it to
the extreme and have collapsed all of them (like cgroup v2).
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Two major bugs: `if (!cg)` and missing `strdup(path)`.
Also convert to use std hash table for lookup of previous cpu load value
for a given cgroup path. The hash table is sized after the current num.
rows on the screen -- resize currently not supported. The value given
to hcreate() should be 25% greater than the estimated num of entries,
but we take a wild guess just to make sure we don't run out of space at
runtime.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
- Fix obvious refactor mistake in cpu.shares assignment
- For unified memory hierarchy we need to set .use_hierarch=1
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The ps command currently only lists processes in the three main control
groups: init, system, user. Kernel threads are not show at all.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
As of now, the default finit cgroup behavior is to use a unified
hierarchy of controllers under /sys/fs/cgroup/finit.
We mount cpu,cpuacct,cpuset,memory (if available) and gain the
ability to control our three major groups: init, system, user.
The default CPU share setup is ~10% for init and user, and 90% for
system. These are guaranteed CPU shares to ensure we do not starve
PID 1 or user processes. Support for configuring these limits will
be added in a later commit.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
With the new practise of keeping record of process pid files, we no
longer need to log/update when reading back the same PID we already
have on file.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The previous patch just added dynamic tracking of non-declared pid
files, so we no longer need to make stuff up.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Services that create their own PID files usually don't declare one with
Finit. This patch adds support to track those PID files anywayt at
runtime for the purpose of identifying match svc_t when a PID file is
removed, i.e. when a service exits.
- On IN_CREATE the pidfile.so plugin saves the pid file name in svc_t
- On ON_DELETE the pidfile.so plugin finds svc_t based on pid file
Quicker tracking and less dead code, win-win.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Transitioning to runlevel 0 or 6 handles graceful shutdown of any
managed run/task/sysv/services. So we can replace the old 2 sec delay
with a shorter one for any non-managed still lingering process.
Also, some minor cleanup.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
To aid with debugging, highlight if env file, binary, or both are the
missing component(s) and causing status 'missing'.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Protect against corner cases where pid conditions are not cleaned up.
We don't want such conditions to remain asserted when the process has
terminated.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
When we try to start a service/run/task we call whichp() to see if the
binary exists, either tha absolute path given in the .conf file, or in
the $PATH we run with. If binary, or the env: file, doesn't exist we
now set svc_missing() state.
On `initctl reload` we unblock the service to be able to check again.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This patch fixes a few really hard problems wrt PID files:
1. Listening for IN_CREATE events means we get notified immediately by
the kernel when someone calls open()/fopen() on a PID file. Reading
the contents returns 0, thank you atoi() ... so we drop IN_CREATE
and instead look for IN_CLOSE_WRITE, there fixed it! Not quite ...
2. Some programs, like Zebra, and other Quagga/Frr daemons, don't
close() their PID files after creation. Instead they ftruncate()
and keep them open, and locked. Presumably to get a mechanism to
detect already running instances -- messes life up a bit for the
rest of us though. So we need to read files after IN_MODIFY too.
3. We also want to track IN_DELETE so we can deassert conditions when
services exit gracefully and clean up their PID files
The rest fo the commit is debug instrumentation changes.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
In 7447192 we dropped support for 'cond set' and 'cond clear' commands
with the motivation they were unsafe and sent the wrong message to the
user. That was true, but mostly because it was too generic and required
the user to call `initctl reload` to apply the changes.
This patch restores the behavior, albeit in a very reduced and simple
format. All conditions set with this command are constrained to the
'usr/...' namespace. No subdirectories are allowed. The argument to
the 'cond set|clear' command is disallowed if it contains '/' or '.'
but anything else is supported, for example:
initctl cond set foo:2
creates a static/oneshot condition in /run/finit/cond/usr/foo:2
These conditions are static and are fully handled by the user. The
initctl command is the recommended, and only supported, way of setting
and clearing usr conditions.
This patch also includes a new plugin, usr.so, which is a very simple
inotify plugin for the /run/finit/cond/usr/ directory. When files are
created or removed here the plugin tells the Finit condition engine to
update and trigger service changes.
For instance, the following service is not started by default at boot:
service <usr/foo> myservice -- MyService
However, as soon as `initctl cond set foo` is called, myservice starts.
Consequently, it is stopped when `initctl cond clr foo` is called.
Another major difference from the original is that this implementation
doesn't send IPC commands to create/delete the conditions. This makes
calling these new commands non-blocking so they can be used very early
in the bootstrap, by plugins, if needed.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
- Use _PATH_foo for all condition paths, *with* trailing /
- Read condition file first, may not exist, in which case we save time
- Change from libte makepath() to mkpath(), this changes from hard-coded
0777 perms on all cond dirs to 0755 -- possible security fix
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Provided a configuration that looks like this:
ospfd.conf:
service [2345] <!pid/zebra> log ospfd -A 127.0.0.1 -u root -g root -- OSPF daemon
zebra.conf:
service [2345] <!> log zebra -A 127.0.0.1 -u root -g root -- Zebra Routing daemon
If zebra.conf is changed, we restart it when `initctl reload` is issued.
This change ensures that ospfd is also restarted, because ospfd depends
on zebra, we must restart it too.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Check if ifname contains double periods, this should *not* be a valid
name, though eth0.10 is, et0..10 is not. Should protect better against
directory traversal attacks.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Fix ev->len validation; must check against sz read(), not total buffer
size. Also, fix off-by-one in comparison.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Fix ev->len validation; must check against sz read(), not total buffer
size. Also, fix off-by-one in comparison.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Fix ev->len validation; must check against sz read(), not total buffer
size. Also, fix off-by-one in comparison.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Check for spaces and slashes in interface name to prevent path based
attacks. Found by Coverity Scan.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This is a major refactor to use the same construct as in the pidfile.so
plugin to parse kernel inotify events for when TTYs are added removed
from the system.
Found by Coverity Scan.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
- Align the two implementations in logit.c and utmp-api.c
- Use libite APIs to handle common constructs
- If gzip fails, don't remove rotated-to file
- Don't try gzip if rename fails
- Issues found by Coverity Scan
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Ignore EBUSY errors, the kernel has the ability to automount /dev as
soon as the rootfs has been mounted. This may be added to procfs and
sysfs later, so we make this a general change in fs_init().
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
- Decleare some return values with (void)fn(), for cases where
we don't care (dropping table headers), or best effor
- Check return value from fgets() and chdir() in some cases that
are valid, i.e., continuing execution is pointless
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
No point in using rq.data, which could have been modified, instead use
the input argument to the function in the error message.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
While skimming through the results of the latest Coverity Scan, I
discovered that that the reset logic of global rlimits was broken.
This it seems to have been since its first introduction in Finit.
We fix this by reading initial rlimits at bootstrap, then for each
reconf, including the first, we seed global rlimits with the initial
ones -- thus resetting between each reconf.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The utmp_set() function allows id==NULL, so default the line to NULL and
avoid parsing empty lines. Which is better code anyway.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Mostly used during development, doesn't really provide any value to the
user at normal runtime. Leave it commented out though so user can do
debug themselves if needed.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
- Refactor screen_init()
- use native impl. of TTY probing from pimd project
- check if TIOCWINSZ works
- check if we're running in watch(1), for initctl
- check if ANSI goto(999,999) escape seq. works (invasive)
- fallback to 80x24
- Drop screen_exit()
- Rename screen_init() to get_width(), for now, matching pimd
- Relocate call in main() to banner(), first fn to write to TTY
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
With the recent changes to the condition handling it has become more and
more evident that the canonical reference for a task/run/service is the
NAME:ID representation. Up until now we've kept the older JOB:ID for
some sort of compatibility fallback.
This patch removes the support to simplify maintenance going forward.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This is quite possibly the ugliest pieces of code in this project.
It is a quick fix to the problem with strtok() to split up the cmd into
cmd + args[]. Without it wordexp() will later get very cranky about
trying to expand `"Partial` strings from commands like this:
/sbin/udhcpc-wrapper.sh -f -S -V "myOS v9.99" -t 8 -T 5 -A 10 -R -p \
/var/run/udhcpc-vlan1.pid -i vlan1 -x hostname:myos-12-34-50 -o \
-O 1 -O 3 -O 6 -O 7 -O 12 -O 15 -O 42 -O 121 -a
Yes really.
It's an obviously far from perfect solution. One might even start to
question why we have an args[64[64] per svc_t and not just a cmd + arg?
Well, there's a lot of improvement potential here, let's talk!
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
If mg doesn't exist, command doesn't evaluate the second argument (vi),
so we need to use `... || command -v vi` instead. Found during demo,
because why not.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This patch fixes an issue with rescue mode on Alpine Linux where the
rescue shell was never properly started.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Turns out that on some systems, that has none of the first two, and
neither of the two environment variables set in ${VISUAL:-${EDITOR}},
they evaluate to 'sh /etc/finit.d/file.conf', which is true, so we
wouldn't even try mg or vi ...
This patch reverts to the spirit of Andreas Egebergs initial commit,
which turns out to be a better solution anyway :)
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Both Debian and Builroot based systems use /etc/default and
Alpine Linux use /etc/conf.d, so let's just put both in by
default. If the directories exist we monitor them.
The --with-syconfig configure option is now disabled by default.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The module setting is deprecated, other mechanisms exist. This change
is to ensure we don't try to (re)load any modules at runtime, only at
bootstrap.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This patch collapses the two similar functions parse_conf() and
parse_conf_dynamic() into a single function. It also lifts old
restrictions on what config stanzas are allowed in /etc/finit.d
The only remaining restriction is the global rlimits, they remain
as the last setting only possible to define /etc/finit.conf which
then applies to all other services.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The enable/disable commands require either both the available/ and
enabled/ subdirectories in /etc/finit.d, or at least available/
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
A system may, or may not, have the available/ and enabled/
subdirectories in /etc/finit.d. This is up to distros but
we support the following, example uses /etc/finit.d as the
FINIT_RCSD base directory:
1. No subdirectories
- enable: nop
- disable: nop
2. Only available/
- enable: checks if there's a service in available/
then symlinks to /etc/finit.d/
- disable: checks if service in /etc/finit.d is a
symlink, then removes it
3. Both available/ and enabled/
- enable: relative symlink in enabled/ is created if
service is available in available/
- disable: symlink in enabled/ is removed if exists
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
- Gamble a bit by redirecting stderr to /dev/null instead of command -v
- Drop vi from VISUAL/EDITOR expansion alternatives, tested later anyway
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Services that are instantiated by plugins, and thus don't have a .conf
file, tried to use uninitialized rlimits. This caused the kernel to
immediately kill those services for violating their own rlimits.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
- Check for env in /etc/conf.d on Alpine
- Update all services with their respective env: file
- Disable default modules, noew modprobe plugin should handle this
- Add watchdog service
- Rename syslogd -> syslog, in case we want to integrate sysklogd
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Command Edit <CONF> opens file /etc/finit.d/available/<CONF>.conf in
the editor referenced by Linux environment variable $EDITOR
If $EDITOR is not set it will try to open emacs and if emacs is not
found it will try to open the file in vim.
Signed-off-by: Andreas Egeberg <andreas.egeberg@gmail.com>
Add support for monitoring environment files for services declared with
the `env:[-]/path/to/file` option. The default path to such files has
been chosen to follow Debian and Buildroot /etc/default/*
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Services that are instantiated by plugins, and thus don't have a .conf
file, tried to use uninitialized rlimits. This caused the kernel to
immediately kill those services for violating their own rlimits.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
- Check for env in /etc/conf.d on Alpine
- Update all services with their respective env: file
- Disable default modules, noew modprobe plugin should handle this
- Add watchdog service
- Rename syslogd -> syslog, in case we want to integrate sysklogd
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Add support for monitoring environment files for services declared with
the `env:[-]/path/to/file` option. The default path to such files has
been chosen to follow Debian and Buildroot /etc/default/*
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
- Use iwatch, modeled after pidfile plugin
- Use one .conf watcher for all *.conf paths/files
- Use full path of conf file for changes, from realpath().
This fixes a long-standing limitation on unique filenames
for services, that absolutely nobody knew about
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
We don't want to show progress (starting/stopping/restaring) at
runtime, since Finit v3. Many systems hooked up to a console
get confused by sudden output, or even ansi escape sequences.
This patch drops code added recently which caused a regression
in this policy. The resulting code is even more readable.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This patch handles a corner case when Finit may not have detected a
supervised process has died. When a user calls `initctl restart foo`
we now send such lost PIDs to the service_monitor() for restart.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Really kill them, our monitored process may be about to restart, so we
don't want any unintended side effects from lingering children in prior
instances.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This patch fixes a long standing issue where removing a service from
/etc/finit.conf does not stop and unregister it. The issue was caused
by the Finit support for "protected" services, e.g. services created
by plugins like hotplug.so
To reproduce issue before this fix:
cat /etc/finit.d/available/ntpd.conf >> /etc/finit.conf
initctl reload
The NTP service now runs smoothly, as expected. Later on, we decide to
drop it from our system:
sed -i 1,2d /etc/finit.conf
initctl reload
... and the NTP service continues to run unaffected. Not what most sane
users expect.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The cond_reload() fn is called on `initctl reload`, thus we don't need
to call service_step(), since we call service_step_all() later in that
cycle. So we can drop cond_update(NULL), which calls service_step().
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
With the redesign from <svc/path/to/pidfile> to <pid/name:id> in d1fac6f
we moved to matching svc_t only against their PID, which could pop up in
any *.pid or */pid in /var/run. This patch drops the (hopefully) last
remnants of the old <svc/> legacy.
To ensure we don't try reading the PID value from socket files, like
/var/run/initctl, we add simple fnmatch() of the inotified file. Two
calls to fnmatch(), for portability reasons, not every system has GNU
libc extensions like FNM_EXTMATCH.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
* test_start_stop_service
Test that we can start/stop (via initctl) a service configured in
finit.conf
* test_start_stop_service_sub_config
Test that we can start/stop (via initctl) a service configured in
finit.d/ files
Signed-off-by: Magnus Malm <magnusmalm@gmail.com>
For now, do not mount /etc as tmpfs, but instead create the dir when
building the rootfs via test/test_root/Makefile.
Signed-off-by: Magnus Malm <magnusmalm@gmail.com>
The script testenv_enter.sh, by default, runs /bin/sh in the environment
currently setup.
A bit easier than copy pasting the old hint text printed before test run.
Signed-off-by: Magnus Malm <magnusmalm@gmail.com>
If a test defines a function 'test_teardown(), that function will be
called before the test env teardown is run.
Signed-off-by: Magnus Malm <magnusmalm@gmail.com>
This patch makes it possible to create a service with optional
arguments, similar to the systemd EnvironmentFile= setting.
The major difference to systemd, is that Finit does not attempt
to start a service if it cannot find the env. file, unless it
is prefixed with '-'.
Example:
/etc/default/syslogd:
SYSLOGD_OPTS="-m0 -r 128k:10 -s"
/etc/finit.d/enabled/syslogd.conf
service env:-/etc/default/syslogd syslogd -F $SYSLOGD_OPTS
Only downside, right now, is that changing the contents of the file
/etc/default/syslogd does not mark the syslogd.conf as modified and
a subsequent `initctl reload` will *not* restart syslogd.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This patch drops the stand-alone reboot binary and folds it into the
initctl tool, which now becomes a multi-call binary.
With this change the reboot/shutdown/poweroff/halt/suspend commands also
no longer default to sending signals to PID 1, instead it now uses the
initctl <--> finit UNIX domain socket API. Signals are only used as a
fallback in case of non-working domain socket. The -f,--force mode of
operation still works, where `reboot(2)` is called directly.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
On systems with udevd, or systemd-udevd, we must wait for udevd to start
before calling udevadm. This patch updates service and runtask stanzas
to the new condition system naming and adds 'log' to the udevadm cmds.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The refactor in 13893b9 caused a regression in composing service
arguments. This patch should rectify that.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Add a new test that verifies that Finit will:
- Start a service that's been added to the config
- Stop the service once it's removed from the config
The second step is currently failing.
Signed-off-by: Jacques de Laval <jacques@de-laval.se>
Simplify log redirection, probe for logit early and fall back to
use a while-loop of syslog() instead.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
By not putting logit in the same process group as the service whose
output is consumed by logit, we can send signals to the service,
and it's group, without affecting logit. This means that logit will
continue to log away until it's stdin is closed, which will happen
when the service is reaped.
Signed-off-by: Jacques de Laval <jacques@de-laval.se>
Delay duplication of stdout/stderr until after forking a logit
process, otherwise we'd have a ligering reference to stdout/stderr.
Signed-off-by: Jacques de Laval <jacques@de-laval.se>
The tzset() function sets the global variables 'daylight', 'timezone' and 'tzname'
according to the value of the TZ environment variable, or to the value of the
_CS_TIMEZONE configuration string if TZ isn't set, or to UTC0 if neither is set.
Signed-off-by: Andreas Egeberg <andreas.egeberg@gmail.com>
Modeled after systemd.show_status, currently without the 'auto' setting.
- finit.debug[=bool]
- finit.show_status[=bool]
- finit.status_style[=old,classic,new,modern]
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This patch reverts back to the progress/status behavior of Finit < v3.0
because this is what most users expect, including the maintainer. The
resulting code and configure script is a lot simpler to understand and
maintain:
- No more --enable-progress or --enable-progress-classic configure
flags. Instead a progress_style variable in helpers.c that can
be changed at compile time for those that really need it.
- No more 'splash' kernel commnand line option. This turned out to
be *very* confusing to many users who believed it was some sort of
graphical splash screen à la Plymouth.
Also, when Finit debug is enabled we now have a global 'debug' flag
which now alo controls if klogctl() should be called to prevent the
kernel logs to the console or not.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
- Default to power-off the system, even for shutdown command
- Fix -h switch for SysV init shutdown command compatibility
- Add -P switch for systemd shutdown command compatibility
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This reverts commit 8a71f2d323 because
if no other application has a working utmp then there's no poit for
Finit to have one. Instead we'll be disabling all utmp support when
running on a musl based system.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Instead of just walking /run/finit/cond, also show the source of each
condition. The hook and net conditions are hard-coded to init[1] atm.
and unknown conditions are shown as unknown[0].
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
In the v4 bump we removed tar.xz archives in favor of bringing back
tar.gz. This was a bit premature and could easily cause unnecessary
headaches downstream.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Printing the progress of run-parts script just like for `rc.local`
seems reasonable. Also log a warning if any script exits abnormally.
Signed-off-by: Jacques de Laval <jacques@de-laval.se>
Purpose of script is to restore an Alpine Linux install to its default
state. Some trace elements of Finit may remain, but critical parts like
reverting back to /bin/busybox as /sbin/init are in place at least.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The SysV API compatibility layer gives too little value to be worth
the effort of maintaining it. Users are encouraged to use the
`initctl` tool instead.
Signed-off-by: Jacques de Laval <jacques@de-laval.se>
Slight reshuffle of columns so status and condition listings are more
similar. Also, calculate and adjust necessary column widths.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
With the recent change in the condition handling we've decided to put
the (name,id) tuple in focus for everything Finit.
This patch changes from job:id in the status listing to name:id instead.
In verbose mode we change to show the job:id and the command + args.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This patch removes the built-in inetd support from Finit. We recommend
using an external inetd instead, e.g. xinetd.
If you liked the feature set our inetd provided; filtering per interface
and port redirection, then please let us know or use the code in this
patch (MIT licensed) to recreate it. We are open to reintroducing it,
but then as a stand-alone daemon like the bundled watchdogd and getty.
So long for now, old friend.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Change service conditions from the non-obvious <svc/path/to/foo> to
<pid/foo:id>, utilizing the unique name of the service instead of the
weird composition of paths from /run and PID file names. Hopefully
making it more clear that one service's pidfile is another service's
`<pid/foo>` condition.
Note: this is an incompatible change to the condition system!
The Finit major version will be stepped to indicate this.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
From, e.g.,
Starting foo:, PID: 123
to
Starting foo[123]
or, when an ID has been given
Starting foo:id[123]
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This fixes the problem with lingering Makefile's in /etc/finit.d/ on the
target systems. We also get the new enabled services properly set up.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The initctl tool's enable/disable command can then be used to manage
these services after install. This is what most users expect.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Planned changes will directly change the behavior of conditions, and
instead of having to implement compatibility wrappers and further
complicate the code, we've taken the decision to go for v4 directly.
This allows us to also rip out the complex inetd support.
Other changes in this commit:
- Revert back to .tar.gz distribution archives, by popular demand
- Add .sha256 checksum for .tar.gz archive
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
We have our own subdirectory alread in /run, for conditions, and the FHS
recommends services maintain their own subdirectory if they have >1 file
for runtime storage.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The <svc/foo> condition was created to synchronize starting services,
hence the abbreviation. Example: the Quagga ripd needs to start after
the zebra daemon to ensure its UNIX domain socket is active, otherwise
events may be lost.
However, considering that synchronization was implemented with UNIX PID
files, e.g. waiting for /var/run/quagga/zebra.pid to be created, the
condition abbreviation name <svc/foo> was hard to understand by most
newcomers to Finit. To make matters worse, a new feature to track or
even create PID files for services that don't create one themselves,
using the syntax 'pid:/path/to/foo.pid' was added.
Connecting the dots between these wasn't obvious.
This patch renames service conditions pid conditions and also adds
a compatibility wrapper to the Finit .conf parser. Any condition
given in old .conf files with 'svc/' prefix are internally renamed
'pid/', along with a LOG_INFO notice in syslog.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This to ensure that services, and conditions they may create, actually
are recorded in the condition subsystem.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The recently updated pidfile plugin now also watches the subdirectories
in /var/run, but for that to work it must witness the creation of these
subdirectories. The bootmisc plugin creates several, e.g., /run/quagga/
in which PID files like zebra.pid are created.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
- Absolut paths are only necessary in case of ambiguities
- Without absulute paths conditions are shorter, e.g. <svc/syslogd>
- Drop <net/lo/up> condition for SSH daemons, not necessary
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
As documented in the kernel docs and systemd[1], the default system
console is the first listed in /sys/class/tty/console/active, which
is the reverse order of console= given on the kernel cmdline.
Some distros don't have console= on their default command line in
their bootloader, e.g. Alpine Linux, some multiple, e.g Buildroot.
Instead of relying on the value given at configure time we probe
sysfs and open all (at most three) consoles listed. This way we
at least get the default console for our progress output.
- drop old --with-console from configure script
- drop /proc/cmdline fishing in favor of sysfs
- replace CONSOLE from configure with new console()
[1]: http://0pointer.de/blog/projects/serial-console.html
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
- Changing the GNU defaults is messy. The best way is to leave them
alone and let the user set their install locations themselves.
- Show --exec-prefix in config Summary, that's where the binaries go
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
From the start, the output from the 'list' command has been different
from the output of the 'show' command. No longer.
Also, drop the 'AVAILABLE', 'ENABLED', and 'STATIC' prefixes. They were
a bit screamy and superfluous. 'STATIC' wasn't even a good name for the
locally administered ones.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
All Debian based distros are moving towards a /etc/network/interfaces.d/
world. Adding support for that is too complex, so we're reverting back
to calling 'ifup -a' or 'ifdown -a'. We still guard this by checking
both for 'ifup' and the existence of /etc/network/interfaces (which in
the new world only contains a line to source per-interface files).
By popular demand, this patch also changes to bring up/down networking
in the background. Useful for embedded system bringup and containers.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Even musl libc supports %m today, as well as most syslog daemons, but
that's no excuse to use this.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Some targets, like Raspberry Pi, have both serial and HDMI console
enabled. Depending on the order of arguments on the kernel cmdline
either one were selected the default 'stderr', which finit used.
This patch adds support for parsing the console= string(s) and
opening multiple outputs for system startup/progress. In the
end, however, a getty still needs to be started on each of the
consoles to use.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
We need /proc for rs_remount_root() and conf_parse_cmdline(), /dev for
early multi-console support, and /sys for the cgroups support. All of
these may be in the system /etc/fstab, which fs_mount() later remounts.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Many systems have a very verbose kernel by default. This patch calls
`klogctl(SYSLOG_ACTION_CONSOLE_OFF, NULL, 0)`, which is sysklogd also
calls but later when it starts emptying /dev/kmsg.
We do this to get an undisturbed console while starting up.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This seems to be inline with Busybox init, SysV init and, according
to the comments in code, the intended behaviour of Finit.
The effect is that `SIGPWR` will lead to Finit issuing the syscall `reboot`
with `cmd` parameter set to `RB_POWER_OFF` (rather than `RB_AUTOBOOT`), which
in turn will lead to a halt rather than a reboot.
The effect of the above inside a PID namespace other than the initial PID
namespace is that `wait` inside the parent of Finit will report that Finit was
killed with a `SIGINT` signal rather than `SIGHUP`.
The contents of /etc/finit.conf and friends could potentially change
after mounting /etc/fstab. If, for example, /etc was to be re-mounted
from ro to rw using an overlay filesystem, we want any user changes to
the configuration to take effect.
Instead of heuristically trying to determine the correct hotplug
daemon to use (mdev/udev), defer to the user to specify it in the
configuration. Same goes for the built-in watchdog.
Traditionally, Finit has setup _some_ filesystems that are Nice To
Have (tm), using options that do the Right Thing (tm).
On the one hand, this is very convenient for users that don't
necessarily know what a reasonable mode= value is for /tmp etc. (pun
intended). It also means that mountpoints can be created if they
don't exist.
On the other hand it means that all users are forced to submit to the
choices made by Finit. Also, different filesystems where spread out
at many different points in `main()`, meaning that the fact that
`/proc` was available at a given hook point did not imply that `/sys`
was, for example.
Instead, defer _all_ filesystem setup to the user by simply calling
`mount -a`. We then ship an example `fstab` that shows how to
re-create the old behavior.
This patch to the banner "hello" is hopefully the last in this new style
progress. Previous efforts have used unicode symbols not available in
all fonts. This time we standardize on one that looks good both the
default Ubuntu Mono and Source Code Pro fonts. It's a bit small in my
preferred font, Envy Code R, but that might be a mistake in the font.
Not tested with PuTTY or Windows Hyper Terminal. I think they use the
Courier font by default, both of them, which I have tested.
Signed-off-by: Joachim Nilsson <troglobit@gmail.com>
Some ppl have their own '.' in service descriptions, adding anther just
messes things up and is confusing. The old style progress still adds
dots, but thats for padding up to [ OK ].
Signed-off-by: Joachim Nilsson <troglobit@gmail.com>
Because we want to reserve @console for going dumpster diving in
/sys/class/tty/console/active and 'console' to mean the literal
/dev/console
This reverts commit 217df432d6.
Signed-off-by: Joachim Nilsson <troglobit@gmail.com>
Up until now Finit has only read /proc/cmdline to determine debug level,
single user mode, etc. This patch is one in a series to make enable it
to run also in containers.
Signed-off-by: Joachim Nilsson <troglobit@gmail.com>
Define MODULES_LOAD_PATH only when it's unset, this allows it to be
overridden, for instance, by passing CFLAGS.
Signed-off-by: Ming Liu <liu.ming50@gmail.com>
This patch makes the RTC plugin a bit more stand-alone, it is now also
back in the hands of the user to set a default timezone.
Signed-off-by: Joachim Nilsson <troglobit@gmail.com>
Most of the helper functions around TTYs assumed that STDIN_FILENO was
already set up and unique for the give TTY they operated on. It turns
out it wasn't. This patch refactors the open() of TTYs to ensure we
do not clobber our TTY settings with multiple calls to tcsetattr() for
the same TTY.
Took a while to track this one down. Usually I don't run systems with
more than one TTY, but the Alpine system on my home server did, and boy
I've never been this close to giving up on computers, ever.
Signed-off-by: Joachim Nilsson <troglobit@gmail.com>
On Alpine Linux 3.6 the tcsetattr() calls failed. This turned out to be
a deeply rooted issue not limited to just musl libc. The workaround is
to call tcsetattr() in several rounds. In particular the modem signals,
c_cflag bits, always fail, so we take a bet on Linux having set up the
terminals (hope as a strategy) and ignore that problem, for now.
Also, use TCSAFLUSH with tcsetattr(), that seems to work better and is
what BusyBox stty use.
The local CTL() macro is replaced with CTL() from ttydefaults.h, which
exists also in musl libc.
Signed-off-by: Joachim Nilsson <troglobit@gmail.com>
The svc_find_by_pidfile() function is called async from the pidfile
plugin, so by the time the queue is processed the origin may have
moved or been removed even. Safe to ignore.
Signed-off-by: Joachim Nilsson <troglobit@gmail.com>
This patch sets the TZ variable to a sane default, UTC. Fixing, at
least, the time restore problems of the RTC plugin.
Also, let the C library figure out DST, as the docs say -1 does ...
Signed-off-by: Joachim Nilsson <troglobit@gmail.com>
When starting up a system booted from initramfs / may already be
mounted (read-only), so we can safely fsck it before proceeding
to remount it as read-write later (possibly).
Signed-off-by: Joachim Nilsson <troglobit@gmail.com>
In case we run into problems mounting /sys or /proc we should have
signals and the optional emergency shell available for recovery.
Signed-off-by: Joachim Nilsson <troglobit@gmail.com>
Unlike find(1), nftw(3) defaults to follow symlinks. For some kernel
configurations this turned out to be detrimental and caused the plugin
to loop forever scanning modalias files.
Also, check if /sys is mounted before even calling nftw().
Signed-off-by: Joachim Nilsson <troglobit@gmail.com>
This patch is a follow-up to the idea introduced in 1d6737e, but this
time we make sure to always append the svc ID to conditions if their
.conf has been declared with :ID
Meaning, the following .conf:
service :0 /sbin/teamd --config-file /etc/teamd-lag0.conf
service :1 /sbin/teamd --config-file /etc/teamd-lag0.conf
results in the following two conditions:
svc/sbin/teamd/0
svc/sbin/teamd/1
With a custom 'name:foo':
service name:lag :0 /sbin/teamd
service name:lag :1 /sbin/teamd
the resulting conditions look even prettier:
svc/lag/0
svc/lag/1
Signed-off-by: Joachim Nilsson <troglobit@gmail.com>
This patch makes the ':ID' field optional when registering non-inetd
run/taks/service etc. It has always been optional in the .conf files,
but internally we defaulted to ':1'.
To open up for more useful conditions we now default to zero/NULL :ID,
meaning the user now *must* keep track of the :ID field if they have
multiple services with the same name. This should already have been
an issue since such services would likely have conflicted with their
PID files (same name). So it's unlikely to affect any user severely.
Signed-off-by: Joachim Nilsson <troglobit@gmail.com>
When registering a service there is the option to set a custom PID file,
and even tell Finit to maintain that PID file in case the daemon doesn't
Until now there hasn't been a way to create custom conditions though ...
This patch leverages the new 'name:foo' argument to tas/run/services to
trigger condition 'svc/foo' (notice lack of path!) when a custom name is
set and either the default, or the custom, PID file is created.
Signed-off-by: Joachim Nilsson <troglobit@gmail.com>
This patch makes it possible to declare tty's using the magic keyword 'console',
an alias for the already existing '@console' keyword, which automatically detect
valid console(s) from the console=foo kernel command line.
Signed-off-by: Joachim Nilsson <troglobit@gmail.com>
The default runlevels are otherwise 2-5. Usually, however, when
debugging a system you may want to have runlevel 1 (single user
level) available to log in.
Signed-off-by: Joachim Nilsson <troglobit@gmail.com>
This is a more sane default, and also what systems expect, and what
other getty implementations support. I.e., unless specified, or set
to 0, the tty baud rate is zero -- which means do not change speed
at all.
Signed-off-by: Joachim Nilsson <troglobit@gmail.com>
The svc_find_by_pidfile() function may be called with "/run/./foo.pid",
or similar, which must be sanitized before being compared to the PID
file of each svc_t.
This patch introduce realpath() to untangle the paths and check if the
file actually exists.
Signed-off-by: Joachim Nilsson <troglobit@gmail.com>
Services may exist in multiple runlevels, i.e. one behavior at bootstrap
and another at runtime. If the user does not set a specific PID file to
be used as condition finit can match the wrong svc.
This patch extends svc_find_by_pidfile() to also compare the contents of
the PID file with the PID of the registered svc_t.
Signed-off-by: Joachim Nilsson <troglobit@gmail.com>
This patch removes the --enable-rw-rootfs build flag. Like OpenRC we
now check /etc/fstab instad; if `/` is listed *and* doesn't have the
`ro` flag set, we remount / read-write.
Signed-off-by: Joachim Nilsson <troglobit@gmail.com>
AC_PLUGIN([netlink], [yes], [Basic netlink plugin for IFUP/IFDN and GW events. Can be replaced with externally built plugin that links with libnl or similar.])
AC_PLUGIN([hook-scripts], [no], [Trigger script execution from hook points])
AC_PLUGIN([hotplug], [no], [Start udevd or mdev kernel event datamon])
AC_PLUGIN([rtc], [yes], [Save and restore RTC using hwclock])
AC_PLUGIN([tty], [yes], [Automatically activate new TTYs, e.g. USB-to-serial])
AC_PLUGIN([urandom], [yes], [Setup and save random seed at boot/shutdown])
AS_HELP_STRING([--without-udev-rules], [Skip install of curated udev rules under /lib/udev/rules.d, default: yes when keventd enabled]),, [with_udev_rules=yes])
AC_ARG_WITH(sulogin,
AS_HELP_STRING([--with-sulogin@<:@=USER@:>@], [Built-in sulogin, optional USER to request password for, default: yes (root)]),[sulogin=$withval],[with_sulogin=yes sulogin=yes])
Some files were not shown because too many files have changed in this diff
Show More
Reference in New Issue
Block a user
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
Joachim Wiberg