[External] : Re: AArch64: loading the thread-local poll value when waking up from native with ldAr in case -XX:+UseSystemMemoryBarrier

[Erik Osterlund]

Hi,

A key ingredient of the race I pointed out is no longer around. The VMThread used to disarm the Java threads. It’s observing that disarmed value that required ldar.

Today, Java threads always disarm themselves. That removes the need for ldar entirely, with or without sys_membarrier.

Thanks for directing my attention to this.

/Erik

> On 25 Jul 2024, at 17:45, Andrew Haley <aph-open at littlepinkcloud.com> wrote:
> 
> On 7/25/24 14:43, Dmitry Chuyko wrote:
>> I wonder if it would be correct to emit the safepoint_poll with
>> acquire=false in case UseSystemMemoryBarrier is true at least on some known
>> CPUs. Regression tests don't reveal problems, but they don't produce eager
>> safepoint/jni races.
> 
> Mmm, good point. That sounds reasonable, but I'll defer to others on Cc:
> 
> --
> Andrew Haley  (he/him)
> Java Platform Lead Engineer
> Red Hat UK Ltd. <https://urldefense.com/v3/__https://www.redhat.com__;!!ACWV5N9M2RV99hQ!OzZDdGStquMIZXrytYV3LNK6l1u5Lg5U2nomVggyZwXZMm3JIDrJFxFYv8xDCcttK-yEKVwFTqS-P_J-ntB1LgMoAtHd$ >
> https://urldefense.com/v3/__https://keybase.io/andrewhaley__;!!ACWV5N9M2RV99hQ!OzZDdGStquMIZXrytYV3LNK6l1u5Lg5U2nomVggyZwXZMm3JIDrJFxFYv8xDCcttK-yEKVwFTqS-P_J-ntB1LriHa8IO$ EAC8 43EB D3EF DB98 CC77 2FAD A5CD 6035 332F A671
>