| 6466c1cc5efc4ff05fabdd670cf78a6a7ca45afb |
|
13-Jun-2017 |
Dianne Hackborn <hackbod@google.com> |
Fix issue #62390590: SecurityException in JobIntentService$...
...JobServiceEngineImpl$WrapperWorkItem.complete
When a job is in the process of stopping a job, we should no
longer allow new work to be dispatched for it.
Also there was an issue with how we determine wether a caller
is valid for the current job -- this was only based on the uid
of the currently running job, but of course that context could
be re-used for a new job of the same uid. Instead, we now create
a different callback binder for each client, so we can identify
the exact client each time it calls back. (This also allows us
to hang information about why the job last stopped on that
client state, so we can always report it.)
Finally make a bunch of classes final that should have been.
Test: bit CtsJobSchedulerTestCases:*
Change-Id: I1b00dd2da710414dd2898c4d39a5c528d54b95ea
/frameworks/base/services/core/java/com/android/server/job/controllers/ContentObserverController.java
|
| f9bac16d61db0fceb15484587ecf876c2b802c37 |
|
21-Apr-2017 |
Dianne Hackborn <hackbod@google.com> |
Fix issue #32180780: Sync adapters inappropriately being run...
...during full-data backup/restore
The activity manager now tells job scheduler service about the
current backup that is running (only if it is a full backup), it
there is a new condition where we won't consider jobs associated
with the current backup to be ready to run.
Also... just a little optimization here. :) The focus is on
scheduling jobs with a 0 deadline, meaning they should run right
away. Now the timing controller does a quick check for a new
job to see if its constraints are already satisifed, and doesn't
do anything further if that is the case (doesn't add to the list,
doesn't re-evaluate alarms, etc). And in the path to scheduling
a job, we do a check to see if the new job is already ready and if
so then just directly add it to the pending list and schedule it.
Doing this required removing what I think is the last bit of code
relying on handler serializing for thread safety, so now everything
in the job scheduler is protected by our global lock and we can
do whatever we want with the lock held and be assured the state
remains consistent.
Also did some small optimizations to many of the other controllers,
mostly switching from an ArrayList to an ArraySet for their tracked
jobs, since one of the things we do frequently is add/remove jobs.
Finally, added some nullability annotations to the JobScheduler
APIs.
Test: bit CtsJobSchedulerTestCases:*
Change-Id: I533fad94ba59468a52fe3d077a0ceab3427f0012
/frameworks/base/services/core/java/com/android/server/job/controllers/ContentObserverController.java
|
| 7da13d7c3e5b48c0410ae869c5679652de97e5aa |
|
05-Apr-2017 |
Dianne Hackborn <hackbod@google.com> |
Add new "work queue" feature to JobScheduler.
This gives semantics similar to the start command
queue of services.
The implementation is currently lacking in URI permission
grant handling of the work intents; that will be coming
in a follow-up change.
This includes a first step of adjusting/fixing locking
within JobSchedulerService. The JobServiceContext class
has a bunch of stuff it does that assumes it doesn't need
locking because it schedules the work on a handler. However,
to be able to correctly implement the work finish flow (that
takes care of stopping the job when there is no more work),
we can't dispatch these asynchronously so need to get rid of
that and just do explicit locking.
The switch to explicit locking is half-way there (again the
remaining part will be a follow-on CL). Right now we have
the locking, but still also the handler. But it turns out
there were a number of things we were doing without a lock
held where we actually should have been holding a lock, so
this is better anyway.
Test: new tests added
Change-Id: Iebd098046209b28e60fd2f4d855d7f91cd3a8b03
/frameworks/base/services/core/java/com/android/server/job/controllers/ContentObserverController.java
|
| e46393e04b94b7c583042b259ac935d00af9b640 |
|
29-Jul-2016 |
Tony Mak <tonymak@google.com> |
We should registerContentObserver with app's user id
Change-Id: Icef7a71f1773c102b766625271ed0156d6c59d77
Fix: 30464189
/frameworks/base/services/core/java/com/android/server/job/controllers/ContentObserverController.java
|
| 4b425712c6a777148af767cb30240d632c51f55b |
|
28-Jun-2016 |
Christopher Tate <ctate@google.com> |
Fix #29737133 : content-trigger jobs not running properly
When removing the ContentObserver wrapper from our internal
bookkeeping we were using the wrong key. That led to future
re-registrations thinking they were reusing a currently-
registered observers, but instead never getting onChange()
notifications.
Change-Id: Id3111db057ae63194049d7d48d45b75be6bb0000
/frameworks/base/services/core/java/com/android/server/job/controllers/ContentObserverController.java
|
| 121e1645d1396b47f0fefa978f8c77d2c4c6c968 |
|
22-Jun-2016 |
Dianne Hackborn <hackbod@google.com> |
Fix issue #29545190: JobScheduler ContentUri NOTIFY_FOR_DESCENDANTS flag...
...isn't respected if previous job was scheduled with same URI
Change-Id: Ib04f6d7cf72e7efc8e84c1a759e91ba520eda139
/frameworks/base/services/core/java/com/android/server/job/controllers/ContentObserverController.java
|
| e9a988caca733d2f292991a52a0047685a69812f |
|
28-May-2016 |
Dianne Hackborn <hackbod@google.com> |
Work on issue #28942589: Tune job scheduler
We now have a new settings key that provides all of the existing
tuning parameters, plus some newly redone ones for dealing with
different memory levels.
Changed the minimum batching for overall jobs from 2 to 1, so
we will never get in the way of immediately scheduling jobs
when the developer asks for this. We should now be able to rely
on the doze modes to do better batching of jobs for us when it
is really important.
Also work on issue #28981330: Excessive JobScheduler wakeup alarms.
Use a work source with scheduled alarms to blame them on the app
whose job they are being scheduled for, and add a check for whether
a job's timing constraint has been satisfied before considering it
a possible candidate for the next alarm. (If it is satisified,
the time is in the past, so we should not schedule an alarm for it.)
Finally clean up a bunch of the dumpsys output to make it easier
to understand.
Change-Id: I06cf2c1310448f47cf386f393e9b267335fabaeb
/frameworks/base/services/core/java/com/android/server/job/controllers/ContentObserverController.java
|
| 2c4b8f9f11cce04ac246f2335d39243641c26b83 |
|
11-May-2016 |
Christopher Tate <ctate@google.com> |
Fix iterator double-advance in ContentObserverController
Bug 28725391
Change-Id: I1994e46645dbf716f3ef5d12f7d4833264e6937a
/frameworks/base/services/core/java/com/android/server/job/controllers/ContentObserverController.java
|
| ef3aa6ee53c5e4f1c50dd5a9b5821c54e449d4b3 |
|
30-Apr-2016 |
Dianne Hackborn <hackbod@google.com> |
Fix issue #28477006: Add small event log to job scheduler
Added. Also fixed dumpsys output when filtering, to apply the
filter to (almost) all of the output.
Change-Id: Iafb446599ad8fddbe8a766784deff618a6cfdbb7
/frameworks/base/services/core/java/com/android/server/job/controllers/ContentObserverController.java
|
| 8db0fc15b85c6501a0418b17edee2d9c447b408a |
|
12-Apr-2016 |
Dianne Hackborn <hackbod@google.com> |
More work on issue #26390151: Add new JobScheduler API...
...for monitoring content providers
We now have some delays before reporting URI changes, to allow
them to batch together.
Also clean up debug output, and fix some issues with how we
were managing the content observer state.
And while I am here, fix the device idle and app idle controllers
to no longer maintain their own list of jobs, but just directly
iterate over the JobStore.
Change-Id: If3fdff23c00c2f1b99901a9be096d851562d3439
/frameworks/base/services/core/java/com/android/server/job/controllers/ContentObserverController.java
|
| 141f11c82a2dbf042833f75aeae6f028e8ae2084 |
|
06-Apr-2016 |
Dianne Hackborn <hackbod@google.com> |
More work on issue #26390151: Add new JobScheduler API...
...for monitoring content providers
- Improve media provider change reporting so that observers can
avoid spurious reports of the top-level content directory changing.
- Fix a bug where collected content changes while a job was running
were not being properly propagated to the next job.
Change-Id: I29e3c2960e6fec75b16ee3ee6588d47342bf8c75
/frameworks/base/services/core/java/com/android/server/job/controllers/ContentObserverController.java
|
| b0001f6fb1383d9824c2733896b0b348e7f77240 |
|
16-Feb-2016 |
Dianne Hackborn <hackbod@google.com> |
Remove internal locking from JobStatus.
Now all state of JobStatus is implicitly protected by the lock
of whoever is using it -- in this case the global lock for the
JobSchedulerService. This allows us to remove all of the atomic
variables and just replace those with a simple bit field.
The required constraints for a job are now statically defined
once a JobStatus is created, and don't change. (They wouldn't
change before, but now this is absolutely specified to be the
case.) This required tweaking the constructors a bit so that
the earliest and latest run times are computed as part of the
core class initialization.
Also clarified methods on StateController that are called with
the lock held, and took advantage of that in the various
controllers to not now redundantly re-acquire the lock.
Change-Id: I595c5e7d1bff1bd2ff906d612581af82878a25ee
/frameworks/base/services/core/java/com/android/server/job/controllers/ContentObserverController.java
|
| 33d31c5b70c7d056e799e34bb6eccbe6939714ea |
|
16-Feb-2016 |
Dianne Hackborn <hackbod@google.com> |
Simplify job scheduler service locking.
Unify all locks to just one lock protecting the entire service.
There is really no need for more complicated locking -- there is
nothing in the code that can take a long time to complete. And
having a single lock will allow various parts of the code to be
much simpler and easier to maintain.
This is just the first step of the change, switching all of the
locking to use one lock. With this done, we can now start
simplifying the code. For example, JobStatus no longer needs
to do any locking (or have atomic variables and such), it can
just rely on its callers holding the global service lock.
Change-Id: I502916ed7f2994b601750c67a59a96b1a4e95c6d
/frameworks/base/services/core/java/com/android/server/job/controllers/ContentObserverController.java
|
| 1a30bd9b13fd127d9bbfdc5fd4cb2f80ab7ece21 |
|
11-Jan-2016 |
Dianne Hackborn <hackbod@google.com> |
Add and implement JobScheduler content observer APIs.
Allows one to schedule jobs to run when content URIs
change, and find out what changed when the job executes.
This required adding a new API to StateController to
tell it when we are about to start executing a job, so
we can transfer the currently collected changes out of
its internal state to fill it in to the JobParameters.
Also some additional dumpsys debug output to help
understand what is going on in the job scheduler.
Change-Id: I91f51b226ff4add7a271a8333beffa5e86c7bf18
/frameworks/base/services/core/java/com/android/server/job/controllers/ContentObserverController.java
|