[prev in list] [next in list] [prev in thread] [next in thread] 

List:       openjdk-hotspot-runtime-dev
Subject:    Re: RFR: 8295872: [PPC64] JfrGetCallTrace: Need pc == nullptr check before frame constructor [v2]
From:       Martin Doerr <mdoerr () openjdk ! org>
Date:       2022-10-31 11:55:39
Message-ID: 7fau1vI-JpvPByWdZfUaMI7YmpZ2ZqF_c3hmFHzzcl8=.c8bb2436-66ee-42b9-9562-a62df7948070 () github ! com
[Download RAW message or body]

On Mon, 31 Oct 2022 11:53:38 GMT, Martin Doerr <mdoerr@openjdk.org> wrote:

> > The check pc == nullptr is needed before the frame constructor.
> 
> Martin Doerr has updated the pull request incrementally with one additional commit \
> since the last revision: 
> Remove acquire barrier again. The null check is the clean fix.

I've spent more time for understanding the memory ordering aspect and I've come to \
the conclusion, that the acquire barrier from \
[JDK-8290004](https://bugs.openjdk.org/browse/JDK-8290004) is not really needed. The \
fields are never read while concurrently modified. They are only read by the current \
thread or when the thread is suspended (or stopped by a signal). So, the pc == null \
check before the frame constructor is a sufficient fix for the profiling usages. An \
unnecessary acquire barrier may be confusing and I think a comment is helpful. Please \
take a 2nd look. Thanks!

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

PR: https://git.openjdk.org/jdk/pull/10846


[prev in list] [next in list] [prev in thread] [next in thread]