Having unit(s) removed/not started, even if it solved the issue and allowed
to boot successfully, should still be considered an error, as something
clearly isn't right.
This patch elevates the log message from warning to error, and adds a status
message to make things more obvious.
This should address TODO item "new dependency type to "group" services
in a target". Semantic of new dependency is as follows. Once configured
it creates dependency which will cause that all dependent units get
stopped if unit they all depend on is stopped or restarted. Usual use
case would be configuring PartOf=some.target in template unit file
and WantedBy=some.target in [Install] section and enabling desired
number of instances. In this case starting one instance won't pull in
target but stopping or starting target(in case of WantedBy is properly
configured) will cause stop/start of all instances.
all other dependencies are in 3rd person. Change BindTo= accordingly to
BindsTo=.
Of course, the dependency is widely used, hence we parse the old name
too for compatibility.
Two of our current job types are special:
JOB_TRY_RESTART, JOB_RELOAD_OR_START.
They differ from other job types by being sensitive to the unit active state.
They perform some action when the unit is active and some other action
otherwise. This raises a question: when exactly should the unit state be
checked to make the decision?
Currently the unit state is checked when the job becomes runnable. It's more
sensible to check the state immediately when the job is added by the user.
When the user types "systemctl try-restart foo.service", he really intends
to restart the service if it's running right now. If it isn't running right
now, the restart is pointless.
Consider the example (from Bugzilla[1]):
sleep.service takes some time to start.
hello.service has After=sleep.service.
Both services get started. Two jobs will appear:
hello.service/start waiting
sleep.service/start running
Then someone runs "systemctl try-restart hello.service".
Currently the try-restart operation will block and wait for
sleep.service/start to complete.
The correct result is to complete the try-restart operation immediately
with success, because hello.service is not running. The two original
jobs must not be disturbed by this.
To fix this we introduce two new concepts:
- a new job type: JOB_NOP
A JOB_NOP job does not do anything to the unit. It does not pull in any
dependencies. It is always immediately runnable. When installed to a unit,
it sits in a special slot (u->nop_job) where it never conflicts with
the installed job (u->job) of a different type. It never merges with jobs
of other types, but it can merge into an already installed JOB_NOP job.
- "collapsing" of job types
When a job of one of the two special types is added, the state of the unit
is checked immediately and the job type changes:
JOB_TRY_RESTART -> JOB_RESTART or JOB_NOP
JOB_RELOAD_OR_START -> JOB_RELOAD or JOB_START
Should a job type JOB_RELOAD_OR_START appear later during job merging, it
collapses immediately afterwards.
Collapsing actually makes some things simpler, because there are now fewer
job types that are allowed in the transaction.
[1] Fixes: https://bugzilla.redhat.com/show_bug.cgi?id=753586
Type=idle is much like Type=simple, however between the fork() and the
exec() in the child we wait until PID 1 informs us that no jobs are
left.
This is mostly a cosmetic fix to make gettys appear only after all boot
output is finished and complete.
Note that this does not impact the normal job logic as we do not delay
the completion of any jobs. We just delay the invocation of the actual
binary, and only for services that otherwise would be of Type=simple.
Recursion is unnecessary, because we're deleting all transaction jobs
anyway. And the recursive deletion produces debug messages that are
pointless in transaction abort.
A transaction can be acyclic, but when it's added to installed jobs,
a cycle may result.
transaction_verify_order_one() attempts to detect these cases, but it
fails because the installed jobs often have the exact generation number
that makes them look as if they were walked already.
Fix it by resetting the generation numbers of all installed jobs before
detecting cycles.
An alternative fix could be to add the generation counter to the
Manager and use it instead of starting always from 1 in
transaction_activate(). But I prefer not having to worry about it
wrapping around.
The functions looked complicated with the nested loops with breaks,
continues, and "while (again)".
Here using goto actually makes them easier to understand.
Also correcting the comment about redundant jobs.
While having a Requires= dependency between units, the dependency is started
automatically on "systemctl start", but it's not started on "systemctl
restart".
JOB_RESTART jobs did not pull the dependencies for starting into the
transaction.
https://bugzilla.redhat.com/show_bug.cgi?id=802770
Note that the other bug noted in comment #2 has been fixed already by avoiding
the deletion of anchor jobs.
Previously transactions could reference installed jobs. It made some issues
difficult to fix.
This sets new rules for jobs:
A job cannot be both a member of a transaction and installed. When jobs are
created, they are linked to a transaction. The whole transaction is constructed
(with merging of jobs within, etc.). When it's complete, all the jobs are
unlinked from it one by one and let to install themselves. It is during the
installation when merging with previously installed jobs (from older
transactions) is contemplated.
Merging with installed jobs has different rules than merging within a
transaction:
- An installed conflicting job gets cancelled. It cannot be simply deleted,
because someone might be waiting for its completion on DBus.
- An installed, but still waiting, job can be safely merged into.
- An installed and running job can be tricky. For some job types it is safe to
just merge. For the other types we merge anyway, but put the job back into
JOB_WAITING to allow it to run again. This may be suboptimal, but it is not
currently possible to have more than one installed job for a unit.
Note this also fixes a bug where the anchor job could be deleted during merging
within the transaction.
Do not attempt to optimize away the job creation by refering to installed jobs.
We do not want to disturb installed jobs until commiting the transaction.
(A later patch to job merging will make the separation of transaction jobs and
installed jobs complete.)
manager.c takes care of the main loop, unit management, signal handling, ...
transaction.c computes transactions.
After split:
manager.c: 65 KB
transaction.c: 40 KB
Olivier Brunel