[8u] request for approval: "8170409: CMS: Crash in CardTableModRefBSForCTRS::process_chunk_boundaries"

[Volker Simonis]

Hi,

could you please approve the backport of the following GC fix to jdk8u-dev:

8170409: CMS: Crash in CardTableModRefBSForCTRS::process_chunk_boundaries

This fix is critical for linux/ppc64 where it impacts Hadoop/Terasort [1].

Bug: https://bugs.openjdk.java.net/browse/JDK-8170409
Webrev: http://cr.openjdk.java.net/~simonis/webrevs/2016/8170409.v3/
Review: http://mail.openjdk.java.net/pipermail/hotspot-gc-dev/2016-November/019313.html,
http://mail.openjdk.java.net/pipermail/hotspot-gc-dev/2016-December/019332.html
URL: http://hg.openjdk.java.net/jdk9/hs/hotspot/rev/fe86ccf9132f

The original patch cleanly applies with only two minor, non-functional changes:

- we have to revert the file renaming from
"share/vm/gc_implementation/parNew/parCardTableModRefBS.cpp" and
"share/vm/memory/cardTableModRefBS.hpp" in jdk8 to
"share/vm/gc/cms/parCardTableModRefBS.cpp" and
"share/vm/gc/shared/cardTableModRefBSForCTRS.hpp" in jdk9.

- in jkd9, jdk8's Universe::heap() was moved to
GenCollectedHeap::heap(). While the call to this function is not part
of the patch, it has to be renamed in the context of the change such
that the patch cleanly applies.

The updated webrev can be found here:

http://cr.openjdk.java.net/~simonis/webrevs/2016/8170409.8u/

Thank you and best regards,
Volker

[1] http://mail.openjdk.java.net/pipermail/hotspot-dev/2016-December/025483.html