[v14.x] Backport V8 8.6 LTS commits

Makefile

@@ -648,8 +648,7 @@ test-with-async-hooks:
 ifneq ("","$(wildcard deps/v8/tools/run-tests.py)")
 # Related CI job: node-test-commit-v8-linux
 test-v8: v8  ## Runs the V8 test suite on deps/v8.
-	deps/v8/tools/run-tests.py --gn --arch=$(V8_ARCH) \
-				--mode=$(BUILDTYPE_LOWER) $(V8_TEST_OPTIONS) \
+	deps/v8/tools/run-tests.py --gn --arch=$(V8_ARCH) $(V8_TEST_OPTIONS) \
 				mjsunit cctest debugger inspector message preparser \
 				$(TAP_V8)
 	$(info Testing hash seed)

common.gypi

@@ -36,7 +36,7 @@
 
     # Reset this number to 0 on major V8 upgrades.
     # Increment by one for each non-official patch applied to deps/v8.
-    'v8_embedder_string': '-node.29',
+    'v8_embedder_string': '-node.60',
 
     ##### V8 defaults for Node.js #####
 

deps/v8/AUTHORS

@@ -67,6 +67,7 @@ Ben Newman <[email protected]>
 Ben Noordhuis <[email protected]>
 Benjamin Tan <[email protected]>
 Bert Belder <[email protected]>
+Brendon Tiszka <[email protected]>
 Burcu Dogan <[email protected]>
 Caitlin Potter <[email protected]>
 Craig Schlenter <[email protected]>

deps/v8/infra/testing/builders.pyl

@@ -1078,7 +1078,7 @@
       {'name': 'v8testing', 'variant': 'slow_path', 'shards': 1},
     ],
   },
-  'V8 Linux64 TSAN - concurrent marking': {
+  'V8 Linux64 TSAN - stress-incremental-marking': {
     'swarming_dimensions' : {
       'os': 'Ubuntu-16.04',
     },
@@ -1105,7 +1105,7 @@
       {
         'name': 'v8testing',
         'test_args': ['--extra-flags=--stress-incremental-marking'],
-        'shards': 4,
+        'shards': 6,
       },
     ],
   },

deps/v8/src/ast/ast-function-literal-id-reindexer.cc

@@ -54,10 +54,10 @@ void AstFunctionLiteralIdReindexer::VisitClassLiteral(ClassLiteral* expr) {
     // Private fields have their key and value present in
     // instance_members_initializer_function, so they will
     // already have been visited.
-    if (prop->value()->IsFunctionLiteral()) {
-      Visit(prop->value());
-    } else {
+    if (prop->kind() == ClassLiteralProperty::Kind::FIELD) {
       CheckVisited(prop->value());
+    } else {
+      Visit(prop->value());
     }
   }
   ZonePtrList<ClassLiteral::Property>* props = expr->public_members();
@@ -67,7 +67,8 @@ void AstFunctionLiteralIdReindexer::VisitClassLiteral(ClassLiteral* expr) {
     // Public fields with computed names have their key
     // and value present in instance_members_initializer_function, so they will
     // already have been visited.
-    if (prop->is_computed_name() && !prop->value()->IsFunctionLiteral()) {
+    if (prop->is_computed_name() &&
+        prop->kind() == ClassLiteralProperty::Kind::FIELD) {
       if (!prop->key()->IsLiteral()) {
         CheckVisited(prop->key());
       }

deps/v8/src/builtins/builtins-array.cc

@@ -649,11 +649,14 @@ class ArrayConcatVisitor {
         index_offset_(0u),
         bit_field_(FastElementsField::encode(fast_elements) |
                    ExceedsLimitField::encode(false) |
-                   IsFixedArrayField::encode(storage->IsFixedArray()) |
+                   IsFixedArrayField::encode(storage->IsFixedArray(isolate)) |
                    HasSimpleElementsField::encode(
-                       storage->IsFixedArray() ||
-                       !storage->map().IsCustomElementsReceiverMap())) {
-    DCHECK(!(this->fast_elements() && !is_fixed_array()));
+                       storage->IsFixedArray(isolate) ||
+                       // Don't take fast path for storages that might have
+                       // side effects when storing to them.
+                       (!storage->map(isolate).IsCustomElementsReceiverMap() &&
+                        !storage->IsJSTypedArray(isolate)))) {
+    DCHECK_IMPLIES(this->fast_elements(), is_fixed_array());
   }
 
   ~ArrayConcatVisitor() { clear_storage(); }
@@ -1063,8 +1066,8 @@ bool IterateElements(Isolate* isolate, Handle<JSReceiver> receiver,
     return IterateElementsSlow(isolate, receiver, length, visitor);
   }
 
-  if (!HasOnlySimpleElements(isolate, *receiver) ||
-      !visitor->has_simple_elements()) {
+  if (!visitor->has_simple_elements() ||
+      !HasOnlySimpleElements(isolate, *receiver)) {
     return IterateElementsSlow(isolate, receiver, length, visitor);
   }
   Handle<JSObject> array = Handle<JSObject>::cast(receiver);
@@ -1080,6 +1083,9 @@ bool IterateElements(Isolate* isolate, Handle<JSReceiver> receiver,
     case HOLEY_SEALED_ELEMENTS:
     case HOLEY_NONEXTENSIBLE_ELEMENTS:
     case HOLEY_ELEMENTS: {
+      // Disallow execution so the cached elements won't change mid execution.
+      DisallowJavascriptExecution no_js(isolate);
+
       // Run through the elements FixedArray and use HasElement and GetElement
       // to check the prototype for missing elements.
       Handle<FixedArray> elements(FixedArray::cast(array->elements()), isolate);
@@ -1106,6 +1112,9 @@ bool IterateElements(Isolate* isolate, Handle<JSReceiver> receiver,
     }
     case HOLEY_DOUBLE_ELEMENTS:
     case PACKED_DOUBLE_ELEMENTS: {
+      // Disallow execution so the cached elements won't change mid execution.
+      DisallowJavascriptExecution no_js(isolate);
+
       // Empty array is FixedArray but not FixedDoubleArray.
       if (length == 0) break;
       // Run through the elements FixedArray and use HasElement and GetElement
@@ -1142,6 +1151,9 @@ bool IterateElements(Isolate* isolate, Handle<JSReceiver> receiver,
     }
 
     case DICTIONARY_ELEMENTS: {
+      // Disallow execution so the cached dictionary won't change mid execution.
+      DisallowJavascriptExecution no_js(isolate);
+
       Handle<NumberDictionary> dict(array->element_dictionary(), isolate);
       std::vector<uint32_t> indices;
       indices.reserve(dict->Capacity() / 2);

deps/v8/src/codegen/arm/macro-assembler-arm.cc

@@ -722,23 +722,22 @@ void TurboAssembler::CallEphemeronKeyBarrier(Register object, Operand offset,
 void TurboAssembler::CallRecordWriteStub(
     Register object, Operand offset, RememberedSetAction remembered_set_action,
     SaveFPRegsMode fp_mode) {
-  CallRecordWriteStub(
-      object, offset, remembered_set_action, fp_mode,
-      isolate()->builtins()->builtin_handle(Builtins::kRecordWrite),
-      kNullAddress);
+  CallRecordWriteStub(object, offset, remembered_set_action, fp_mode,
+                      Builtins::kRecordWrite, kNullAddress);
 }
 
 void TurboAssembler::CallRecordWriteStub(
     Register object, Operand offset, RememberedSetAction remembered_set_action,
     SaveFPRegsMode fp_mode, Address wasm_target) {
   CallRecordWriteStub(object, offset, remembered_set_action, fp_mode,
-                      Handle<Code>::null(), wasm_target);
+                      Builtins::kNoBuiltinId, wasm_target);
 }
 
 void TurboAssembler::CallRecordWriteStub(
     Register object, Operand offset, RememberedSetAction remembered_set_action,
-    SaveFPRegsMode fp_mode, Handle<Code> code_target, Address wasm_target) {
-  DCHECK_NE(code_target.is_null(), wasm_target == kNullAddress);
+    SaveFPRegsMode fp_mode, int builtin_index, Address wasm_target) {
+  DCHECK_NE(builtin_index == Builtins::kNoBuiltinId,
+            wasm_target == kNullAddress);
   // TODO(albertnetymk): For now we ignore remembered_set_action and fp_mode,
   // i.e. always emit remember set and save FP registers in RecordWriteStub. If
   // large performance regression is observed, we should use these values to
@@ -762,9 +761,13 @@ void TurboAssembler::CallRecordWriteStub(
 
   Move(remembered_set_parameter, Smi::FromEnum(remembered_set_action));
   Move(fp_mode_parameter, Smi::FromEnum(fp_mode));
-  if (code_target.is_null()) {
+  if (builtin_index == Builtins::kNoBuiltinId) {
     Call(wasm_target, RelocInfo::WASM_STUB_CALL);
+  } else if (options().inline_offheap_trampolines) {
+    CallBuiltin(builtin_index);
   } else {
+    Handle<Code> code_target =
+        isolate()->builtins()->builtin_handle(Builtins::kRecordWrite);
     Call(code_target, RelocInfo::CODE_TARGET);
   }
 

deps/v8/src/codegen/arm/macro-assembler-arm.h

@@ -588,7 +588,7 @@ class V8_EXPORT_PRIVATE TurboAssembler : public TurboAssemblerBase {
 
   void CallRecordWriteStub(Register object, Operand offset,
                            RememberedSetAction remembered_set_action,
-                           SaveFPRegsMode fp_mode, Handle<Code> code_target,
+                           SaveFPRegsMode fp_mode, int builtin_index,
                            Address wasm_target);
 };
 

deps/v8/src/codegen/arm64/macro-assembler-arm64.cc

@@ -2738,23 +2738,22 @@ void TurboAssembler::CallEphemeronKeyBarrier(Register object, Operand offset,
 void TurboAssembler::CallRecordWriteStub(
     Register object, Operand offset, RememberedSetAction remembered_set_action,
     SaveFPRegsMode fp_mode) {
-  CallRecordWriteStub(
-      object, offset, remembered_set_action, fp_mode,
-      isolate()->builtins()->builtin_handle(Builtins::kRecordWrite),
-      kNullAddress);
+  CallRecordWriteStub(object, offset, remembered_set_action, fp_mode,
+                      Builtins::kRecordWrite, kNullAddress);
 }
 
 void TurboAssembler::CallRecordWriteStub(
     Register object, Operand offset, RememberedSetAction remembered_set_action,
     SaveFPRegsMode fp_mode, Address wasm_target) {
   CallRecordWriteStub(object, offset, remembered_set_action, fp_mode,
-                      Handle<Code>::null(), wasm_target);
+                      Builtins::kNoBuiltinId, wasm_target);
 }
 
 void TurboAssembler::CallRecordWriteStub(
     Register object, Operand offset, RememberedSetAction remembered_set_action,
-    SaveFPRegsMode fp_mode, Handle<Code> code_target, Address wasm_target) {
-  DCHECK_NE(code_target.is_null(), wasm_target == kNullAddress);
+    SaveFPRegsMode fp_mode, int builtin_index, Address wasm_target) {
+  DCHECK_NE(builtin_index == Builtins::kNoBuiltinId,
+            wasm_target == kNullAddress);
   // TODO(albertnetymk): For now we ignore remembered_set_action and fp_mode,
   // i.e. always emit remember set and save FP registers in RecordWriteStub. If
   // large performance regression is observed, we should use these values to
@@ -2778,9 +2777,13 @@ void TurboAssembler::CallRecordWriteStub(
 
   Mov(remembered_set_parameter, Smi::FromEnum(remembered_set_action));
   Mov(fp_mode_parameter, Smi::FromEnum(fp_mode));
-  if (code_target.is_null()) {
+  if (builtin_index == Builtins::kNoBuiltinId) {
     Call(wasm_target, RelocInfo::WASM_STUB_CALL);
+  } else if (options().inline_offheap_trampolines) {
+    CallBuiltin(builtin_index);
   } else {
+    Handle<Code> code_target =
+        isolate()->builtins()->builtin_handle(Builtins::kRecordWrite);
     Call(code_target, RelocInfo::CODE_TARGET);
   }
 

deps/v8/src/codegen/arm64/macro-assembler-arm64.h

@@ -1419,7 +1419,7 @@ class V8_EXPORT_PRIVATE TurboAssembler : public TurboAssemblerBase {
 
   void CallRecordWriteStub(Register object, Operand offset,
                            RememberedSetAction remembered_set_action,
-                           SaveFPRegsMode fp_mode, Handle<Code> code_target,
+                           SaveFPRegsMode fp_mode, int builtin_index,
                            Address wasm_target);
 };
 

deps/v8/src/codegen/code-stub-assembler.cc

@@ -1801,12 +1801,13 @@ TNode<IntPtrT> CodeStubAssembler::LoadJSReceiverIdentityHash(
   return var_hash.value();
 }
 
-TNode<Uint32T> CodeStubAssembler::LoadNameHashField(SloppyTNode<Name> name) {
-  CSA_ASSERT(this, IsName(name));
-  return LoadObjectField<Uint32T>(name, Name::kHashFieldOffset);
+TNode<Uint32T> CodeStubAssembler::LoadNameHashAssumeComputed(TNode<Name> name) {
+  TNode<Uint32T> hash_field = LoadNameHashField(name);
+  CSA_ASSERT(this, IsClearWord32(hash_field, Name::kHashNotComputedMask));
+  return Unsigned(Word32Shr(hash_field, Int32Constant(Name::kHashShift)));
 }
 
-TNode<Uint32T> CodeStubAssembler::LoadNameHash(SloppyTNode<Name> name,
+TNode<Uint32T> CodeStubAssembler::LoadNameHash(TNode<Name> name,
                                                Label* if_hash_not_computed) {
   TNode<Uint32T> hash_field = LoadNameHashField(name);
   if (if_hash_not_computed != nullptr) {
@@ -1994,13 +1995,13 @@ TNode<T> CodeStubAssembler::LoadArrayElement(TNode<Array> array,
   }
 }
 
-template TNode<MaybeObject>
+template V8_EXPORT_PRIVATE TNode<MaybeObject>
 CodeStubAssembler::LoadArrayElement<TransitionArray>(TNode<TransitionArray>,
                                                      int, Node*, int,
                                                      ParameterMode,
                                                      LoadSensitivity);
 
-template TNode<MaybeObject>
+template V8_EXPORT_PRIVATE TNode<MaybeObject>
 CodeStubAssembler::LoadArrayElement<DescriptorArray>(TNode<DescriptorArray>,
                                                      int, Node*, int,
                                                      ParameterMode,
@@ -8063,7 +8064,7 @@ void CodeStubAssembler::LookupBinary(TNode<Name> unique_name,
   TNode<Uint32T> limit =
       Unsigned(Int32Sub(NumberOfEntries<Array>(array), Int32Constant(1)));
   TVARIABLE(Uint32T, var_high, limit);
-  TNode<Uint32T> hash = LoadNameHashField(unique_name);
+  TNode<Uint32T> hash = LoadNameHashAssumeComputed(unique_name);
   CSA_ASSERT(this, Word32NotEqual(hash, Int32Constant(0)));
 
   // Assume non-empty array.
@@ -8081,7 +8082,7 @@ void CodeStubAssembler::LookupBinary(TNode<Name> unique_name,
     TNode<Uint32T> sorted_key_index = GetSortedKeyIndex<Array>(array, mid);
     TNode<Name> mid_name = GetKey<Array>(array, sorted_key_index);
 
-    TNode<Uint32T> mid_hash = LoadNameHashField(mid_name);
+    TNode<Uint32T> mid_hash = LoadNameHashAssumeComputed(mid_name);
 
     Label mid_greater(this), mid_less(this), merge(this);
     Branch(Uint32GreaterThanOrEqual(mid_hash, hash), &mid_greater, &mid_less);
@@ -8108,7 +8109,7 @@ void CodeStubAssembler::LookupBinary(TNode<Name> unique_name,
     TNode<Uint32T> sort_index =
         GetSortedKeyIndex<Array>(array, var_low.value());
     TNode<Name> current_name = GetKey<Array>(array, sort_index);
-    TNode<Uint32T> current_hash = LoadNameHashField(current_name);
+    TNode<Uint32T> current_hash = LoadNameHashAssumeComputed(current_name);
     GotoIf(Word32NotEqual(current_hash, hash), if_not_found);
     Label next(this);
     GotoIf(TaggedNotEqual(current_name, unique_name), &next);

deps/v8/src/codegen/code-stub-assembler.h

@@ -1353,13 +1353,12 @@ class V8_EXPORT_PRIVATE CodeStubAssembler
   // Check if the map is set for slow properties.
   TNode<BoolT> IsDictionaryMap(SloppyTNode<Map> map);
 
-  // Load the hash field of a name as an uint32 value.
-  TNode<Uint32T> LoadNameHashField(SloppyTNode<Name> name);
-  // Load the hash value of a name as an uint32 value.
+  // Load the Name::hash() value of a name as an uint32 value.
   // If {if_hash_not_computed} label is specified then it also checks if
   // hash is actually computed.
-  TNode<Uint32T> LoadNameHash(SloppyTNode<Name> name,
+  TNode<Uint32T> LoadNameHash(TNode<Name> name,
                               Label* if_hash_not_computed = nullptr);
+  TNode<Uint32T> LoadNameHashAssumeComputed(TNode<Name> name);
 
   // Load length field of a String object as Smi value.
   TNode<Smi> LoadStringLengthAsSmi(TNode<String> string);

deps/v8/src/codegen/ia32/assembler-ia32.h

@@ -959,6 +959,9 @@ class V8_EXPORT_PRIVATE Assembler : public AssemblerBase {
   void movapd(XMMRegister dst, Operand src) {
     sse2_instr(dst, src, 0x66, 0x0F, 0x28);
   }
+  void movupd(XMMRegister dst, Operand src) {
+    sse2_instr(dst, src, 0x66, 0x0F, 0x10);
+  }
 
   void movmskpd(Register dst, XMMRegister src);
   void movmskps(Register dst, XMMRegister src);
@@ -1331,6 +1334,7 @@ class V8_EXPORT_PRIVATE Assembler : public AssemblerBase {
   void vmovapd(XMMRegister dst, Operand src) { vpd(0x28, dst, xmm0, src); }
   void vmovups(XMMRegister dst, XMMRegister src) { vmovups(dst, Operand(src)); }
   void vmovups(XMMRegister dst, Operand src) { vps(0x10, dst, xmm0, src); }
+  void vmovupd(XMMRegister dst, Operand src) { vpd(0x10, dst, xmm0, src); }
   void vshufps(XMMRegister dst, XMMRegister src1, XMMRegister src2, byte imm8) {
     vshufps(dst, src1, Operand(src2), imm8);
   }

deps/v8/src/codegen/ia32/macro-assembler-ia32.cc

@@ -415,25 +415,24 @@ void TurboAssembler::CallEphemeronKeyBarrier(Register object, Register address,
 void TurboAssembler::CallRecordWriteStub(
     Register object, Register address,
     RememberedSetAction remembered_set_action, SaveFPRegsMode fp_mode) {
-  CallRecordWriteStub(
-      object, address, remembered_set_action, fp_mode,
-      isolate()->builtins()->builtin_handle(Builtins::kRecordWrite),
-      kNullAddress);
+  CallRecordWriteStub(object, address, remembered_set_action, fp_mode,
+                      Builtins::kRecordWrite, kNullAddress);
 }
 
 void TurboAssembler::CallRecordWriteStub(
     Register object, Register address,
     RememberedSetAction remembered_set_action, SaveFPRegsMode fp_mode,
     Address wasm_target) {
   CallRecordWriteStub(object, address, remembered_set_action, fp_mode,
-                      Handle<Code>::null(), wasm_target);
+                      Builtins::kNoBuiltinId, wasm_target);
 }
 
 void TurboAssembler::CallRecordWriteStub(
     Register object, Register address,
     RememberedSetAction remembered_set_action, SaveFPRegsMode fp_mode,
-    Handle<Code> code_target, Address wasm_target) {
-  DCHECK_NE(code_target.is_null(), wasm_target == kNullAddress);
+    int builtin_index, Address wasm_target) {
+  DCHECK_NE(builtin_index == Builtins::kNoBuiltinId,
+            wasm_target == kNullAddress);
   // TODO(albertnetymk): For now we ignore remembered_set_action and fp_mode,
   // i.e. always emit remember set and save FP registers in RecordWriteStub. If
   // large performance regression is observed, we should use these values to
@@ -461,10 +460,14 @@ void TurboAssembler::CallRecordWriteStub(
 
   Move(remembered_set_parameter, Smi::FromEnum(remembered_set_action));
   Move(fp_mode_parameter, Smi::FromEnum(fp_mode));
-  if (code_target.is_null()) {
+  if (builtin_index == Builtins::kNoBuiltinId) {
     // Use {wasm_call} for direct Wasm call within a module.
     wasm_call(wasm_target, RelocInfo::WASM_STUB_CALL);
+  } else if (options().inline_offheap_trampolines) {
+    CallBuiltin(builtin_index);
   } else {
+    Handle<Code> code_target =
+        isolate()->builtins()->builtin_handle(Builtins::kRecordWrite);
     Call(code_target, RelocInfo::CODE_TARGET);
   }
 

deps/v8/src/codegen/ia32/macro-assembler-ia32.h

@@ -292,6 +292,7 @@ class V8_EXPORT_PRIVATE TurboAssembler : public TurboAssemblerBase {
   AVX_OP2_WITH_TYPE(Movaps, movaps, XMMRegister, XMMRegister)
   AVX_OP2_WITH_TYPE(Movapd, movapd, XMMRegister, XMMRegister)
   AVX_OP2_WITH_TYPE(Movapd, movapd, XMMRegister, const Operand&)
+  AVX_OP2_WITH_TYPE(Movupd, movupd, XMMRegister, const Operand&)
   AVX_OP2_WITH_TYPE(Pmovmskb, pmovmskb, Register, XMMRegister)
   AVX_OP2_WITH_TYPE(Movmskps, movmskps, Register, XMMRegister)
 
@@ -566,7 +567,7 @@ class V8_EXPORT_PRIVATE TurboAssembler : public TurboAssemblerBase {
 
   void CallRecordWriteStub(Register object, Register address,
                            RememberedSetAction remembered_set_action,
-                           SaveFPRegsMode fp_mode, Handle<Code> code_target,
+                           SaveFPRegsMode fp_mode, int builtin_index,
                            Address wasm_target);
 };