changeset 53004:27ebaa5566ea

8214302: Allow safely calling is_unloading() on zombie nmethods Reviewed-by: kvn, pliden
author eosterlund
date Wed, 05 Dec 2018 11:01:44 +0100
parents 6ed72482de52
children 9144c0b5c1e1
files src/hotspot/share/code/nmethod.cpp
diffstat 1 files changed, 33 insertions(+), 3 deletions(-) [+]
line wrap: on
line diff
--- a/src/hotspot/share/code/nmethod.cpp	Wed Dec 05 08:55:42 2018 +0100
+++ b/src/hotspot/share/code/nmethod.cpp	Wed Dec 05 11:01:44 2018 +0100
@@ -1575,14 +1575,44 @@
   if (state_is_unloading) {
     return true;
   }
-  if (state_unloading_cycle == CodeCache::unloading_cycle()) {
+  uint8_t current_cycle = CodeCache::unloading_cycle();
+  if (state_unloading_cycle == current_cycle) {
     return false;
   }
 
   // The IsUnloadingBehaviour is responsible for checking if there are any dead
   // oops in the CompiledMethod, by calling oops_do on it.
-  state_unloading_cycle = CodeCache::unloading_cycle();
-  state_is_unloading = IsUnloadingBehaviour::current()->is_unloading(this);
+  state_unloading_cycle = current_cycle;
+
+  if (is_zombie()) {
+    // Zombies without calculated unloading epoch are never unloading due to GC.
+
+    // There are no races where a previously observed is_unloading() nmethod
+    // suddenly becomes not is_unloading() due to here being observed as zombie.
+
+    // With STW unloading, all is_alive() && is_unloading() nmethods are unlinked
+    // and unloaded in the safepoint. That makes races where an nmethod is first
+    // observed as is_alive() && is_unloading() and subsequently observed as
+    // is_zombie() impossible.
+
+    // With concurrent unloading, all references to is_unloading() nmethods are
+    // first unlinked (e.g. IC caches and dependency contexts). Then a global
+    // handshake operation is performed with all JavaThreads before finally
+    // unloading the nmethods. The sweeper never converts is_alive() && is_unloading()
+    // nmethods to zombies; it waits for them to become is_unloaded(). So before
+    // the global handshake, it is impossible for is_unloading() nmethods to
+    // racingly become is_zombie(). And is_unloading() is calculated for all is_alive()
+    // nmethods before taking that global handshake, meaning that it will never
+    // be recalculated after the handshake.
+
+    // After that global handshake, is_unloading() nmethods are only observable
+    // to the iterators, and they will never trigger recomputation of the cached
+    // is_unloading_state, and hence may not suffer from such races.
+
+    state_is_unloading = false;
+  } else {
+    state_is_unloading = IsUnloadingBehaviour::current()->is_unloading(this);
+  }
 
   state = IsUnloadingState::create(state_is_unloading, state_unloading_cycle);