Changeset 190522 in webkit for trunk/Source/JavaScriptCore/dfg/DFGJITCode.h

Timestamp:

Oct 2, 2015, 2:16:20 PM (10 years ago)

Author:

[email protected]

Message:

Unreviewed, rolling back in r190450
​https://bugs.webkit.org/show_bug.cgi?id=149727

The cause of the crash was a CodeBlock, after surviving a call to
deleteAllCode by virtue of being in the remembered set, trying to mark
its inlined CodeBlocks via pointers from its inlined executables.
Since deleteAllCode clears those pointers, the CodeBlock would ASSERT.
(Any other choice to retain a CodeBlock after deleteAllCode -- for
example, conservative marking -- could trigger the same bug.)

The fix is for InlineCallFrame to point directly to its inlined CodeBlock
instead of pointing indirectly via an executable. This guarantees that
CodeBlocks are GC safe regardless of whether we've called deleteAllCode.

Restored changesets:

"CodeBlock should be a GC object"
​https://bugs.webkit.org/show_bug.cgi?id=149727
​http://trac.webkit.org/changeset/190450

File:

• trunk/Source/JavaScriptCore/dfg/DFGJITCode.h (modified) (3 diffs)

trunk/Source/JavaScriptCore/dfg/DFGJITCode.h

@@ -29 +29 @@
 #if ENABLE(DFG_JIT)
 
+#include "CodeBlock.h"
 #include "CompilationResult.h"
 #include "DFGCommonData.h"
@@ -115 +116 @@
     
     void shrinkToFit();
+
+#if ENABLE(FTL_JIT)
+    CodeBlock* osrEntryBlock() { return m_osrEntryBlock.get(); }
+    void setOSREntryBlock(VM& vm, const JSCell* owner, CodeBlock* osrEntryBlock) { m_osrEntryBlock.set(vm, owner, osrEntryBlock); }
+    void clearOSREntryBlock() { m_osrEntryBlock.clear(); }
+#endif
     
 private:
@@ -129 +136 @@
     uint8_t nestedTriggerIsSet { 0 };
     UpperTierExecutionCounter tierUpCounter;
-    RefPtr<CodeBlock> osrEntryBlock;
+    WriteBarrier<CodeBlock> m_osrEntryBlock;
     unsigned osrEntryRetry;
     bool abandonOSREntry;