RFR: 8318462: [GenShen] Prevent unsafe access to displaced mark-word [v3]

[Kelvin Nilsen]

On Thu, 26 Oct 2023 16:03:12 GMT, Kelvin Nilsen <kdnilsen at openjdk.org> wrote:

>> Roman Kennke has updated the pull request with a new target base due to a merge or a rebase. The incremental webrev excludes the unrelated changes brought in by the merge/rebase. The pull request contains six additional commits since the last revision:
>> 
>>  - Fix inflating check with LW locking
>>  - Merge branch 'master' into JDK-8318462
>>  - Accept all-zero header with LW locking
>>  - Handle forwarded headers
>>  - Assert that no inflation happens
>>  - 8318462: [Shenandoah] Prevent unsafe access to displaced mark-word
>
> src/hotspot/share/gc/shenandoah/shenandoahHeap.inline.hpp line 544:
> 
>> 542:   // is thread-local and therefore safe to access. However, when the mark is
>> 543:   // displaced (i.e. stack-locked or monitor-locked), then it must be considered
>> 544:   //a shared memory location. It can/ be accessed by other threads.
> 
> Minor formatting issues in this line of comment.

and below.

> src/hotspot/share/gc/shenandoah/shenandoahMark.inline.hpp line 123:
> 
>> 121:       assert(region->is_young(), "Only for young objects");
>> 122:       uint age = ShenandoahHeap::get_object_age(obj);
>> 123:       CENSUS_NOISE(heap->age_census()->add(age, region->age(), region->youth(), size, worker_id);)
> 
> I'm thinking if get_object_age() returns max_age+1, the CENSUS_NOISE macros might end up with invalid data.  Don't we need a check here?

I think the safe thing to do is treat max_age+1 as if it were 0, insofar as promotion semantics is concerned.  We may delay promotion that way.

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

PR Review Comment: https://git.openjdk.org/shenandoah/pull/343#discussion_r1373431845
PR Review Comment: https://git.openjdk.org/shenandoah/pull/343#discussion_r1373436928