RFR (S): 8025228: assert(new_entry->reserved_words() == vs_word_size) fails in nightly

Bengt Rutisson bengt.rutisson at oracle.com
Tue Sep 24 11:32:28 UTC 2013


Hi everyone,

Could I have a couple of reviews for this change?

http://cr.openjdk.java.net/~brutisso/8025228/webrev.00/

https://bugs.openjdk.java.net/browse/JDK-8025228

We hit this assert in VirtualSpaceList::get_new_chunk() when we have an 
allocation that is larger than VirtualSpaceSize and that is not aligned 
with vm_allocation_granularity. The reason is that the VirtualSpace will 
grow with sizes aligned to vm_allocation_granularity but we we only 
guarantee that the sizes we want to grow with are page aligned. This 
means that we only hit this assert on platforms (Windows) where 
vm_allocation_granularity > vm_page_size.

Since VirtualSpaceSize is vm_allocation_granularity aligned on all 
platforms we only hit the assert for allocations larger than that size 
due to this line:

1224           size_t grow_vs_words =
1225             MAX2((size_t)VirtualSpaceSize, aligned_expand_vs_by_words);

The fix is to make sure that the grow_vs_words value is 
vm_allocation_granularity aligned and not just page size aligned.


Testing:

Added a unit test that fails this way on Windows before my fix:

Running internal VM tests
Running test: TestReservedSpace_test()
Running test: TestReserveMemorySpecial_test()
Running test: TestVirtualSpace_test()
Running test: MetaspaceAux_test()
# To suppress the following error report, specify this argument
# after -XX: or in .hotspotrc:  SuppressErrorAt=\metaspace.cpp:1145
#
# A fatal error has been detected by the Java Runtime Environment:
#
#  Internal Error (...src\share\vm\memory\metaspace.cpp:1145), pid=220, 
tid=3352
#  assert(vs_byte_size % os::vm_allocation_granularity() == 0) failed: 
Not aligned

After my fix the test passes on all platforms.

Ran the UTE test list vm.quick-pcl.testlist. All test pass.

Thanks,
Bengt

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mail.openjdk.org/pipermail/hotspot-gc-dev/attachments/20130924/889d1b32/attachment.htm>


More information about the hotspot-gc-dev mailing list