We were calling STS join and leave during a STW pause and we are not suppoesed to. I now only call those during concurrent phase. I also added stress code in the non-product builds to force an overflows (the condition that ws uncovering the bug) to make sure it does not happen again.
Reviewed-by: johnc, brutisso
Repurpose the existing flag G1PrintRegionLivenessInfo to print out the liveness distribution across the regions in the heap at the end of marking.
Reviewed-by: iveresov, jwilhelm
Use _max_num_q = max(discovery_degree, processing_degree), and let balance_queues() redistribute from discovery_degree to processing_degree of queues. This should also allow a more dynamic and flexible parallelism policy in the future.
Reviewed-by: jmasa, johnc
A race between the concurrent cleanup thread and the VM thread while it is processing the "expanded sparse table list" causes both threads to try to free the same sparse table entry and either causes one of the threads to fail or leaves the entry in an inconsistent state. The solution is purge all entries on the expanded list that correspond go regions that are being cleaned up.
Reviewed-by: brutisso, johnc
This changeset removes the zero-filling thread from G1 and collapses the two free region lists we had before (the "free" and "unclean" lists) into one. The new free list uses the new heap region sets / lists abstractions that we'll ultimately use it to keep track of all regions in the heap. A heap region set was also introduced for the humongous regions. Finally, this change increases the concurrency between the thread that completes freeing regions (after a cleanup pause) and the rest of the system (before we'd have to wait for said thread to complete before allocating a new region). The changest also includes a lot of refactoring and code simplification.
Reviewed-by: jcoomes, johnc
This changeset introduces filtering of SATB buffers at the point when they are about to be enqueued. If this filtering clears enough entries on each buffer, the buffer can then be re-used and not enqueued. This cuts down the number of SATB buffers that need to be processed by the concurrent marking threads.
Reviewed-by: johnc, ysr
Initialize the _is_alive_non_header field of G1's reference processor with an instance of the G1CMIsAliveClosure. This will stop adding reference objects with live referents to the discovered reference lists unnecessarily.
Reviewed-by: tonyp, ysr, jwilhelm, brutisso
Enable reference discovery during concurrent marking by setting the reference processor field of the concurrent marking closure. Keep reference objects on the discovered reference lists alive during incremental evacuation pauses until they are processed at the end of concurrent marking.
Reviewed-by: ysr, tonyp
The changes from 6941395 did not clear the CMTask::_aborted_region fields when concurrent marking aborted because of overflow. As a result, the next time around we could see a memory region whose start address was above the global finger and the assertion tripped. Moved the clearing of the aborted regions to ConcurrentMark::clear_marking_state, which is executed on all of the exit paths.
Reviewed-by: tonyp, ysr, jmasa
Re-enable use of the lock-free versions of region stack push() and pop() by recording aborted regions in a thread-local structure, which are then processed when scanning of the region stack restarts. The previous locking versions of these routines are retained for diagnostic purposes.
Reviewed-by: tonyp, ysr
We're renaming "evacuation failure" to "to-space overflow". I'm also piggy-backing a small additional change which removes the "Mark closure took..." output.
Reviewed-by: ysr, johnc
Make sure that two marking cycles do not overlap, i.e., a new one can only start after the concurrent marking thread finishes all its work. In the fix I piggy-back a couple of minor extra fixes: some general code reformatting for consistency (only around the code I modified), the removal of a field (G1CollectorPolicy::_should_initiate_conc_mark) which doesn't seem to be used at all (it's only set but never read), as well as moving the "is GC locker active" test earlier into the G1 pause / Full GC and using a more appropriate method for it.
Reviewed-by: johnc, jmasa, jcoomes, ysr
Calling the methods region_stack_push() and region_stack_pop() concurrent is not MT-safe. The assumption is that we will only call region_stack_push() during a GC pause and region_stack_pop() during marking. Unfortunately, we also call region_stack_push() during marking which seems to be introducing subtle marking failures. This change introduces lock-based methods for pushing / popping to be called during marking.
Reviewed-by: iveresov, johnc
Newly-created threads always had the active field of their SATB queue initialized to false, even if they were created during marking. As a result, updates from threads created during a marking cycle were never enqueued and never processed. The fix includes remaining a method from active() to is_active() for readability and naming consistency.
Reviewed-by: ysr, johnc
If CM task was aborted while scanning the last object of the specified region and the size of that object is equal to bitmap's granularity then the next offset would be equal or over the region limit which is exactly what the assertion states.
Reviewed-by: ysr, tonyp, jmasa
Removing the concurrent overhead tracker from G1, along with the GC overhead reporter and the G1AccountConcurrentOverhead (both of which rely on the the concurrent overhead tracker).
Reviewed-by: iveresov, johnc
RSet scrubbing incorrectly deletes RSet entries that point to regions tagged as "continues humongous" due to a race when RSet scrubbing iterates over regions in parallel.
Reviewed-by: apetrusenko, iveresov
