This patch fixes the problem with run/task jobs with conditions, run at
runlevel S (bootstrap), not being given enough time to actually start
and run even once.
The new svc counter 'once' is now inspected before transitioning away
from runlevel S to the configured runlevel. A global (currently hard
coded) timeout of 10 sec is given to all run/tasks.
As a spin-off, this patch starts up the event loop much earlier than
before. Meaning the initial runlevel change at boot is now done in
a finalize() callback when all run/tasks have completed, or when the
global timeout has expired. Starting the event loop earlier gives us
the possibility to rely on the pidfile plugin to provide our run/tasks
with the conditions necessary to start.
TODO: What remains is to add similar handling to all runlevel
transitions, not just the initial one at boot.
Signed-off-by: Joachim Nilsson <troglobit@gmail.com>
This patch implements support for calling run/task jobs on the built-in
`HOOK_*` points in Finit. To map each HOOK point to a condition a new
classification was made, as shown in plugin.h and in the example below.
task <hook/mount/root> /sbin/attachubi.sh # run in parallel with other tasks/services
run <hook/net/up> /sbin/dosomething.sh # run in sequence with other commands
Please note, only rudimentary tested.
Signed-off-by: Joachim Nilsson <troglobit@gmail.com>
Make svc state transition to RUNNING *before* service_start(), because
in that function we may call service_step() as well, which would start
the svc again (since it was still in READY).
Signed-off-by: Joachim Nilsson <troglobit@gmail.com>
Only run HOOK_SVC_START on successful start, and *after* having printed
the result of the started service, otherwise any print-outs from plugins
or tasks we start may overwrite the started service's progress.
Signed-off-by: Joachim Nilsson <troglobit@gmail.com>
Don't really see the point of this code anymore, was possibly an issue
before we refactored the logging subsystem. Comment out for now, and
remove later.
Signed-off-by: Joachim Nilsson <troglobit@gmail.com>
If you have lots of service/run/tasks using the same cmd it's impossible
to tell them apart in debug mode. This patch logs description as well
as cmd to improve on this situation a bit.
Signed-off-by: Joachim Nilsson <troglobit@gmail.com>
This patch adds a 'once' counter to the svc_t for run/task jobs. This
counter is incremented when Finit calls service_start() and is reset on
runlevel changes.
Also, at the end of a runlevel each run/task have their state reset to
SVC_HALTED_STATE so they can be launched again for the new runlevel, if
so declared.
Signed-off-by: Joachim Nilsson <troglobit@gmail.com>
Systems running OverlayFS, or similar, on /etc may have new/modified
.conf files in /etc/finit.d. This patch makes sure to reload them
after `mount -a` has run so any configured bootstrap tasks and the
new hook callback scripts may run as expected.
Signed-off-by: Joachim Nilsson <troglobit@gmail.com>
This patch fixes a rather nasty bug in the SIGTERM+SIGKILL logic
employed by Finit when stopping tasks, e.g. at a runlevel change.
When a process terminates, after Finit having sent it SIGTERM, and the
SIGKILL timer expires *before* the timer has been cancelled, then the
timer callback would cause kill(0, SIGKILL).
Signed-off-by: Joachim Nilsson <troglobit@gmail.com>
This patch extends the existing rlimit implementation to support setting
limits per service, run/task, inetd, and tty.
Use rlimit in /etc/finit.conf to change the global setting, which is
then inherited to each /etc/finit.d/*.conf. For each .conf file the
rlimit is reinitialized to the global finit.conf settings.
Also, add `unlimited` keyword, to replace the now deprecated `infinity`
keyword. The latter is however kept, for compatibility with previous
releases, for the foreseeable future.
Signed-off-by: Joachim Nilsson <troglobit@gmail.com>
This patch not only greatly simplifies the registration of the built-in
watchdog daemon, it also fixes a nasty race condition with an external
watchdog daemon. The INIT_CMD_WDOG_HELLO could reach finit in reversed
order, leading to the built-in daemon shutting down, but finit killing
the just registered external daemon.
Signed-off-by: Joachim Nilsson <troglobit@gmail.com>
Most Linux distributions, embedded or otherwise, do not need this
and only confuses the user of an OS with Finit.
Signed-off-by: Joachim Nilsson <troglobit@gmail.com>
Most Linux distributions, embedded or otherwise, do not need this
and only confuses the user of an OS with Finit.
Signed-off-by: Joachim Nilsson <troglobit@gmail.com>
Some systems use /var/db while others use /var/lib/misc, we need to
support both and cannot rely on _PATH_VARDB, since it too might not
be set properly.
Signed-off-by: Joachim Nilsson <troglobit@gmail.com>
When alibuEv event watcher callback has `UEV_ERROR` in the events mask
we must handle the error. Unfortunately, in most cases there is not
much we can do but ignore it or log the error.
Signed-off-by: Joachim Nilsson <troglobit@gmail.com>
If the speed argument to the `tty` stanza is invalid we should not
default to some other random speed. Instead, leave the line speed
as-is, like we do when starting an external getty.
Signed-off-by: Joachim Nilsson <troglobit@gmail.com>
This patch relocates the speed parser so that exec.c can set the TTY
speed properly as soon as possible. For external getty we make sure
to not touch the TTY speed.
Signed-off-by: Joachim Nilsson <troglobit@gmail.com>
This patch makes the Finit built-in getty behave a bit more like other
getty implementations. We now not only disable ISIG but also VEOF and
VINTR when prearing the TTY for the getty (built-in or external).
Like before, it is up to the getty process to leave the TTY in a
usable state to /bin/login, so we also move the stty() call to
just before calling /bin/login in getty.c
Signed-off-by: Joachim Nilsson <troglobit@gmail.com>
This patch moves the stty() function from getty.c to a separate file so
it can be called from exec.c as well. The fix to #84 is simply to call
stty() in prepare_tty(), which is shared by run_getty() and run_getty2()
and called before "Please press Enter ...".
Signed-off-by: Joachim Nilsson <troglobit@gmail.com>
Time comparision should be >= not just >. May seem a bit silly to
nitpick like this since we're comparing nano seconds, but on systems
with no high-res timers this happens more often than you would like.
Signed-off-by: Joachim Nilsson <troglobit@gmail.com>
When performing an `initctl reload` with one (unchanged) service
depending on, e.g. `net/iface/lo`, its condition will not be set
to ON by the pidfile plugin unless the netlink plugin hook runs
first.
Example:
service <net/iface/lo> /sbin/dropbear ...
Which provides the <svc/sbin/dropbear> condition, will not be
set by pidfile.so during `initctl reload` because dropbear is
still SIGSTP:ed waiting for <net/iface/lo>.
Signed-off-by: Joachim Nilsson <troglobit@gmail.com>
All this code did was trigger prctl() to set process name to "console",
which you could use as a simple means of figuring out if a process was
started from the system console.
There are other ways to do this, which Finit should not be involved in.
Signed-off-by: Joachim Nilsson <troglobit@gmail.com>
Before we call screen_init(), which in turn calls libite:initscr(),
which in turn does a lot of tcgetattr() and termios to STDOUT to figure
out the size, we need to check that we're not backgrounded, or don't
actually have a TTY. Problem reported by @wkz
Signed-off-by: Joachim Nilsson <troglobit@gmail.com>
This patch adds support for a new TTY device keyword `@console` which
auto-detects the system console by reading the SysFS file:
/sys/class/tty/console/active
Embedded systems with one of /dev/ttyAMA0, /tty/ttyMXC0 or the original
/dev/ttyS0, as their active serial console can now instead use a single
line to /etc/finit.d/getty.conf to start the built-in getty on the same
TTY as the kernel already runs on:
tty [12345789] @console 115200 noclear
With no TTY defined Finit will either do nothing (default), or start a
fallback shell if enabled in the configure script.
Signed-off-by: Joachim Nilsson <troglobit@gmail.com>
Disable INTR, QUIT, SUSP, and DSUSP while handing over to getty. It is
up to the getty process to allow ISIG again.
This patch is due to the differences in control line setup between ARM,
x86/x86_64, and most notably PPC. PPC in particular seem to have ISIG
set by default from the kernel.
Without this patch pressing Ctrl-C before getty is started will lead to
Finit receiving SIGINT and defaulting to a reboot, as introduced in the
BusyBox init signal adoption in eeb59eb.
Signed-off-by: Joachim Nilsson <troglobit@gmail.com>
In GCC 7 the -Wall infers the new -Wformat-truncation which finds
problems with string truncation in functions like snprintf().
This patch suppresses the warning by allowing only a subset of the
parent inetd service description to be copied to the child connection.
Signed-off-by: Joachim Nilsson <troglobit@gmail.com>
In GCC 7 the -Wall infers the new -Wformat-truncation which finds
problems with string truncation in functions like snprintf().
This patch fixes a possibly problem in path name creation which,
combined with the scandir() function, might cause loss of .conf
files, failure to call run-parts properly, and kill services at
shutdown.
Signed-off-by: Joachim Nilsson <troglobit@gmail.com>
In GCC 7 the -Wall infers the new -Wformat-truncation which finds
problems with string truncation in functions like snprintf().
This patch fixes a problem in condition parsing for very long
PID filenames which might be truncated in internal buffers
causing a mismatch in Finit condition tracking.
Signed-off-by: Joachim Nilsson <troglobit@gmail.com>
In GCC 7 the -Wextra infers the new -Wimplicit-fallthrough which can be
suppressed by adding a simple comment
/* fallthrough */
Which we had, but was too verbose (and misspelled) for GCC to notice.
Signed-off-by: Joachim Nilsson <troglobit@gmail.com>
Joachim Nilsson