On devices without /dev/alarm, use a new backend based on timerfd.
timerfd has near-equivalent syscalls for the /dev/alarm ioctls we care
about, with two key differences:
1) /dev/alarm uses one fd for all clocks, while timerfd needs one fd per
clock type.
AlarmManagerService addresses this by replacing the fd (int) with an
opaque pointer (long) to the backend-specific state.
2) When the RTC changes, the /dev/alarm WAIT ioctl always returns, while
timerfd cancels (and signals events) only on specially-flagged RTC
timerfds.
The timerfd backend masks this by creating an extraneous RTC timerfd,
specifically so there's always something to signal on RTC changes.
Change-Id: I5aef867748298610347f6e1479dd8bf569495832
Signed-off-by: Greg Hackmann <ghackmann@google.com>
Don't keep unused global references to classes, don't throw exceptions
when an exception is already pending, and fix a (harmless) misunderstanding
about how GetStringChars works.
Change-Id: Ie445036f057daa8a1c76aceb7bad2a84fb81d820
HAVE_ANDROID_OS was defined as "1" for targets, but never defined as "0"
for non-targets. Changing them to #ifdef should be safe and matches
all the other uses of HAVE_ANDROID_OS throughout the system.
Change-Id: I82257325a8ae5e4e4371ddfc4dbf51cea8ea0abb
When the wakeup time is negative, the kernel /dev/alarm driver
never triggers the alarm. This can cause alarms to back up in the
priority queue since an alarm at the head with a negative wakup time
will never be triggered. Now we use 0 as the wakup time which causes
an immediate triggering.
When the wakeup time is very large, it is possible for a numeric
overflow to occur when converting the timestamp from milliseconds
since epoch to nanoseconds. This has been fixed by avoiding the
intermediate conversion in the JNI call so that overflow cannot
occur.
Bug: b/2558820
Change-Id: I4f5b4646a04090cc749a9fc5d3982a68402954ef
