infernap12/jdk
1
0
Fork 0
mirror of https://github.com/openjdk/jdk.git synced 2026-01-28 12:09:14 +00:00
Code Issues Packages Projects Releases Wiki Activity

8366038: Thread::SpinRelease should use Atomic::release_store

Browse Source 
Reviewed-by: dholmes, ayang
This commit is contained in:
Anton Artemov 2025-09-04 04:35:51 +00:00 committed by David Holmes
parent f4d73d2a3d
commit 4d1dfabcb4
1 changed files with 3 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
src/hotspot/share/runtime/thread.cpp
View File

@ -602,7 +602,6 @@ void Thread::SpinAcquire(volatile int * adr) {
void Thread::SpinRelease(volatile int * adr) {
assert(*adr != 0, "invariant");
OrderAccess::fence(); // guarantee at least release consistency.
// Roach-motel semantics.
// It's safe if subsequent LDs and STs float "up" into the critical section,
// but prior LDs and STs within the critical section can't be allowed
@ -610,8 +609,7 @@ void Thread::SpinRelease(volatile int * adr) {
// Loads and stores in the critical section - which appear in program
// order before the store that releases the lock - must also appear
// before the store that releases the lock in memory visibility order.
// Conceptually we need a #loadstore|#storestore "release" MEMBAR before
// the ST of 0 into the lock-word which releases the lock, so fence
// more than covers this on all platforms.
*adr = 0;
// So we need a #loadstore|#storestore "release" memory barrier before
// the ST of 0 into the lock-word which releases the lock.
Atomic::release_store(adr, 0);
}