[jdk18] RFR: 8278627: Shenandoah: TestHeapDump test failed

[Roman Kennke]

On Tue, 21 Dec 2021 15:49:56 GMT, Zhengyu Gu <zgu at openjdk.org> wrote:

> This intermittent failure seems to be observed only on Linux ppc64le, but it is **not** ppc specific problem.
> 
> With concurrent thread stack scanning/updating, nmtehod entry barrier has to be applied prior to thread stack/code roots scanning during heap iteration, to ensure scanner only sees to-space objects.
> 
> The patch also cleanup the use of `nmethod::oops_do_marking_prologue/epilogue()`, which should be used with `nmethod::oops_do_try_claim()` and closures that use the nmethod, e.g. `MarkingCodeBlobClosure`. Since `Shenandoah` switched to concurrent code root iteration, they are no longer applied.
> 
> Test:
> The failure on Linux ppc64le is intermittent, usually sees a failure about 3 `hotspot_gc_shenandoah` runs. I have yet seen any failures after 10 runs with the patch.
> 
> SAP tested this patch in jdk/jdk repo in nightly tests, did not see failure so far.

Marked as reviewed by rkennke (Reviewer).

-------------

PR: https://git.openjdk.java.net/jdk18/pull/56