[aarch64-port-dev ] [NEW BUG]Missing LoadStore barrier in interpreter?

[Patrick Zhang]

Thanks for reviewed this, Andrew.

Hi Felix,
Could you please help sponsor this extra patch as well? It is quite similar as your previous one.
I found JDK-8218185 has been resolved and the subject was for "putfield_or_static". Probably a new Jira ticket is needed for this "fast_storefield" and two issues are closely corelated. I have no commit permission so appreciate your help very much! (both jdk/jdk and jdk8u)

Regarding the tests, I have run jtreg jdk_core and jcstress on my aarch64 platform, no problem found. Please help double check at your side too if possible. Thanks.

diff -r 2a78b2963793 src/hotspot/cpu/aarch64/templateTable_aarch64.cpp
--- a/src/hotspot/cpu/aarch64/templateTable_aarch64.cpp Wed Feb 13 13:27:17 2019 +0100
+++ b/src/hotspot/cpu/aarch64/templateTable_aarch64.cpp Fri Feb 22 17:18:20 2019 +0800
@@ -2981,7 +2981,7 @@
   {
     Label notVolatile;
     __ tbz(r3, ConstantPoolCacheEntry::is_volatile_shift, notVolatile);
-    __ membar(MacroAssembler::StoreStore);
+    __ membar(MacroAssembler::StoreStore | MacroAssembler::LoadStore);
     __ bind(notVolatile);
   }

Regards
Patrick

-----Original Message-----
From: Andrew Haley <aph at redhat.com> 
Sent: Friday, February 22, 2019 5:01 PM
To: Patrick Zhang <patrick at os.amperecomputing.com>; Yangfei (Felix) <felix.yang at huawei.com>
Cc: aarch64-port-dev at openjdk.java.net
Subject: Re: [aarch64-port-dev ] [NEW BUG]Missing LoadStore barrier in interpreter?

On 2/22/19 4:04 AM, Patrick Zhang wrote:
> Perhaps TemplateTable::fast_storefield needs the additional " | MacroAssembler::LoadStore" as well, please double check, thanks.
> http://hg.openjdk.java.net/jdk/jdk/file/f41793b5b83f/src/hotspot/cpu/aarch64/templateTable_aarch64.cpp#l2984

Yes.

-- 
Andrew Haley
Java Platform Lead Engineer
Red Hat UK Ltd. <https://www.redhat.com>
EAC8 43EB D3EF DB98 CC77 2FAD A5CD 6035 332F A671