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 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>
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>
Services, like dbus and teamd for instance, may create their PID files
in a subdirectory of /var/run (today often /run). E.g.,
- /var/run/teamd/a1.pid -- For aggregate A1
- /var/run/dbus/pid
- /var/run/lxc/foo.pid -- For container foo
This patch adds support for dynamically adding inotify watchers to any
new subdirectory created in /var/run (discarding too deep directories).
To match services in this directory the run/task/service/sysv stanza
must contain the pid:!/path/to/pidfile.pid syntax. This pid file name
is also used to create the condition this service asserts using the
following formula:
svc/ + <dirname of service> + <subdir and file without .pid>
E.g., the case of teamd (above) gives condition 'svc/usr/bin/teamd/a1'
The special case of dbus is interesting, since it may not be a special
case, but rather the norm for services using a subdirectory. It is
handled as follows; when a new subdirectory is detected, the directory
is scanned for files matching *.pid. Matching files follow the teamd
case. The directory is also scanned for 'pid', which then gives us the
condition 'svc/usr/bin/dbus'
One last example, illustrated by lxc-start, where we want to track the
condition for the LXC container foo. The service stanza:
service pid:!/run/lxc/foo.pid lxc-start -n foo -F -p /run/lxc/foo.pid -- Container foo
This command has no leading path so the condition is composed entirely
from the PID file location:
svc/ + '' + lxc/foo => svc/lxc/foo
Signed-off-by: Joachim Nilsson <troglobit@gmail.com>
This issue may arise in situations where the name of the service being
compared is a sub-string of the service we want to see if it exist.
Example:
- Two different services: teamd, teamdagentd
- They have the same id, 1 being the default if none is specified
- The service with the shortest name is created first
While checking if the "teamdagentd" service exists we eventually compare
it to the "teamd" service. Prior to this fix we only compare the names
names with the length of the existing service. Hence, teamdagentd will
seem to be the same as teamd since the first part of the string is equal
to the existing.
Signed-off-by: Jonas Holmberg <jonas.holmberg@westermo.se>
Signed-off-by: Joachim Nilsson <troglobit@gmail.com>
This is a safer version of strtok() that makes sure we do not step
outside of string bounds. Coverity complained about this, but was
likely more concerned with the lacking sanitize() and trusting data
read from a UNIX domain socket ...
Note: not only is tokstr() not re-entrant, it is also hard-coded
to only check for whitespace as token separator.
Signed-off-by: Joachim Nilsson <troglobit@gmail.com>
Allows the service/task name derived from the binary to be
overridden with a custom name.
Example:
service name:dhcpd [2] /usr/sbin/udhcpd -f -- DHCP Server
Signed-off-by: Robert Andersson <robert.m.andersson@se.atlascopco.com>
When a process, that has been sent SIGTERM by us, takes a full 3 sec to
terminate, the event loop may have both the SIGCHLD event (where we do
the svc_del()) and the 3 sec timeout event to send SIGKILL in its event
cache.
When we call svc_del() it releases the svc_t memory, which can then be
dereferenced by the SIGKILL timer callback and we're doomed.
There are two fixes to this; 1) the event loop, that uses epoll_wait(),
can set maxevents=1 (instead of today's 10). The kernel will then drop
the SIGKILL timer event before it's delivered to the userspace process.
2) we can postpone deleting the svc_t to a "later stage" when all events
in the event cache have been processed.
This patch implements (2). A later patch will use uev_init1(ctx, 1) to
ensure the event cache handles only one event at a time.
Signed-off-by: Joachim Nilsson <troglobit@gmail.com>
The pidfile plugin just listens to changes in files in /run or /var/run,
so when trying to locate an svc_t from a file change we need to compare
using the proper prefix path.
Signed-off-by: Joachim Nilsson <troglobit@gmail.com>
The service_step() function may call svc_del() on an svc from the
service_monitor(), so we need a way to tell the latter that it no
longer can dereference the pointer.
This regression is limited to v3.1, due to the introduction of the
new non-shared-memory store of svc_t.
Signed-off-by: Joachim Nilsson <troglobit@gmail.com>
In f3669c7 the difference between static and dynamic services was
removed. This led to a regression in handling internally created
services for, e.g., dbus-daemon and udevd, mentioned in issue #90.
This patch introduces a "protected" flag for svc_t to prevent the
mark-and-sweep handling of regularly loaded services.
Signed-off-by: Joachim Nilsson <troglobit@gmail.com>
Until now Finit has used mtime to determine if a file has been changed,
or simply touched to request reload, when `initctl reload` is called.
Using mtime for this purpose is a monumentally bad idea since time can
be changed at any point: a user may adjust the time, NTP continously
tunes the clock, and time stamps are stored as the walltime. So if we
compare timestamps against the last time we (booted or) did reload, we
would miss .conf file changes.
This patch replaces the mtime mistake with an inotify watcher for the
following files: /etc/finit.d/*.conf, /etc/finit.d/available/*.conf,
and /etc/finit.conf. All file changes between (bootstraps and) calls
to `initctl reload` are tracked, like the mtime backend it replaces.
Signed-off-by: Joachim Nilsson <troglobit@gmail.com>
This is a prelude to a major behavioral change. The difference between
a static and dynamic service will be removed in Finit v3.1. This means
all .conf files, including /etc/finit.conf, will be re-read on reload.
However, not all configuration directives will be re-read, some will
only be applicable at bootstrap, i.e. runlevel S.
Signed-off-by: Joachim Nilsson <troglobit@gmail.com>
This change removes a long standing kernel requirement and architectural
mishap in the design of Finit v1. It is no longer necessary to include
CONFIG_SYSVIPC in the kernel to use Finit and its command line tool
initctl. It is also no longer possible to connect to the internals of
PID 1 using shmat(), the initctl tool has already been updated to use
the domain socket API.
Signed-off-by: Joachim Nilsson <troglobit@gmail.com>
We should not let PID 1 fail on systems where Linux is built without
CONFIG_SYSVIPC, or the libc does not support shmget()/shmat().
This workaround implements a fallback to using calloc() for PID 1,
which means the initctl tool will not work properly. A better fix
would be to refactor the shm implementation to instead have initctl
call Finit using api.c for each svc_t. This has been added to the
TODO file as future work.
Signed-off-by: Joachim Nilsson <troglobit@gmail.com>
The generic term for a non-running service is blocked, but a service may
be blocked for several different reasons. When a user stops a service
with `initctl stop` the intuitive expected output from `initctl show` is
"stopped" not "blocked".
When the user later restarts the service with `initctl start` it may of
course be blocked due to wrong runlevel or a missing condition. As can
be expected from the service's configuration.
The rest of the patch refactors the internal API names a bit to reflect
this change.
Signed-off-by: Joachim Nilsson <troglobit@gmail.com>
Joachim Nilsson