RFR: 8305819: LogConfigurationTest intermittently fails on AArch64 [v2]

[Dan Gao]

On Fri, 5 May 2023 18:32:46 GMT, Xin Liu <xliu at openjdk.org> wrote:

>> Dan Gao has updated the pull request incrementally with one additional commit since the last revision:
>> 
>>   comment fixme
>
> src/hotspot/share/logging/logOutputList.hpp line 52:
> 
>> 50:   struct LogOutputNode : public CHeapObj<mtLogging> {
>> 51:     LogOutput*      _value;
>> 52:     LogOutputNode* volatile _next;
> 
> Atomic::load_acquire will cast its argument to const volatile T* no matter what.
>  
> another reason is that there's no other load of _next field. I don't see any chance compiler elides load instruction for it. 
> so I don't think we need 'volatile' qualifier for it. same reasons for  _level_start[LogLevel::Count]

I have referred to the usage of Atomic::load_acquire in hotspot, and found that most of them use 'volatile' qualifier even if there is no multiple load of the field in some situation. Furthermore, I think adding volatile will not cause performance or correctness issue (harmless), thus I tend to add it.

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

PR Review Comment: https://git.openjdk.org/jdk/pull/13421#discussion_r1186606089