From b5ca290605a631b43d71fe47c01ea37cf6a40ba9 Mon Sep 17 00:00:00 2001
From: Ali Mohammad Pur <ali.mpfard@gmail.com>
Date: Sat, 4 Dec 2021 17:56:58 +0330
Subject: [PATCH] LibWasm: Implement memory.init and passive mode data

---
 .../AbstractMachine/AbstractMachine.cpp       | 33 ++++++++++++++++--
 .../LibWasm/AbstractMachine/AbstractMachine.h | 27 ++++++++++++++-
 .../AbstractMachine/BytecodeInterpreter.cpp   | 34 +++++++++++++++----
 .../AbstractMachine/BytecodeInterpreter.h     |  2 +-
 .../LibWasm/AbstractMachine/Validator.cpp     |  3 ++
 5 files changed, 88 insertions(+), 11 deletions(-)

diff --git a/Userland/Libraries/LibWasm/AbstractMachine/AbstractMachine.cpp b/Userland/Libraries/LibWasm/AbstractMachine/AbstractMachine.cpp
index 90789d8c09..6893296e29 100644
--- a/Userland/Libraries/LibWasm/AbstractMachine/AbstractMachine.cpp
+++ b/Userland/Libraries/LibWasm/AbstractMachine/AbstractMachine.cpp
@@ -54,6 +54,13 @@ Optional<GlobalAddress> Store::allocate(GlobalType const& type, Value value)
     return address;
 }
 
+Optional<DataAddress> Store::allocate_data(Vector<u8> initializer)
+{
+    DataAddress address { m_datas.size() };
+    m_datas.append(DataInstance { move(initializer) });
+    return address;
+}
+
 Optional<ElementAddress> Store::allocate(ValueType const& type, Vector<Reference> references)
 {
     ElementAddress address { m_elements.size() };
@@ -101,6 +108,14 @@ ElementInstance* Store::get(ElementAddress address)
     return &m_elements[value];
 }
 
+DataInstance* Store::get(DataAddress address)
+{
+    auto value = address.value();
+    if (m_datas.size() <= value)
+        return nullptr;
+    return &m_datas[value];
+}
+
 ErrorOr<void, ValidationError> AbstractMachine::validate(Module& module)
 {
     if (module.validation_status() != Module::ValidationStatus::Unchecked) {
@@ -304,7 +319,7 @@ InstantiationResult AbstractMachine::instantiate(Module const& module, Vector<Ex
                     size_t offset = 0;
                     result.values().first().value().visit(
                         [&](auto const& value) { offset = value; },
-                        [&](Reference const&) { instantiation_result = InstantiationError { "Data segment offset returned a reference" }; });
+                        [&](Reference const&) { instantiation_result = InstantiationError { "Data segment offset returned a reference"sv }; });
                     if (instantiation_result.has_value() && instantiation_result->is_error())
                         return;
                     if (main_module_instance.memories().size() <= data.index.value()) {
@@ -314,6 +329,13 @@ InstantiationResult AbstractMachine::instantiate(Module const& module, Vector<Ex
                         };
                         return;
                     }
+                    auto maybe_data_address = m_store.allocate_data(data.init);
+                    if (!maybe_data_address.has_value()) {
+                        instantiation_result = InstantiationError { "Failed to allocate a data instance for an active data segment"sv };
+                        return;
+                    }
+                    main_module_instance.datas().append(*maybe_data_address);
+
                     if (data.init.is_empty())
                         return;
                     auto address = main_module_instance.memories()[data.index.value()];
@@ -332,8 +354,13 @@ InstantiationResult AbstractMachine::instantiate(Module const& module, Vector<Ex
                         instance->data().overwrite(offset, data.init.data(), data.init.size());
                     }
                 },
-                [&](DataSection::Data::Passive const&) {
-                    // FIXME: What do we do here?
+                [&](DataSection::Data::Passive const& passive) {
+                    auto maybe_data_address = m_store.allocate_data(passive.init);
+                    if (!maybe_data_address.has_value()) {
+                        instantiation_result = InstantiationError { "Failed to allocate a data instance for a passive data segment"sv };
+                        return;
+                    }
+                    main_module_instance.datas().append(*maybe_data_address);
                 });
         }
     });
diff --git a/Userland/Libraries/LibWasm/AbstractMachine/AbstractMachine.h b/Userland/Libraries/LibWasm/AbstractMachine/AbstractMachine.h
index 381f7e4e21..4527bb4a10 100644
--- a/Userland/Libraries/LibWasm/AbstractMachine/AbstractMachine.h
+++ b/Userland/Libraries/LibWasm/AbstractMachine/AbstractMachine.h
@@ -34,6 +34,7 @@ TYPEDEF_DISTINCT_NUMERIC_GENERAL(u64, true, true, false, false, false, true, Ext
 TYPEDEF_DISTINCT_NUMERIC_GENERAL(u64, true, true, false, false, false, true, TableAddress);
 TYPEDEF_DISTINCT_NUMERIC_GENERAL(u64, true, true, false, false, false, true, GlobalAddress);
 TYPEDEF_DISTINCT_NUMERIC_GENERAL(u64, true, true, false, false, false, true, ElementAddress);
+TYPEDEF_DISTINCT_NUMERIC_GENERAL(u64, true, true, false, false, false, true, DataAddress);
 TYPEDEF_DISTINCT_NUMERIC_GENERAL(u64, true, true, false, false, false, true, MemoryAddress);
 
 // FIXME: These should probably be made generic/virtual if/when we decide to do something more
@@ -214,12 +215,14 @@ class ModuleInstance {
 public:
     explicit ModuleInstance(
         Vector<FunctionType> types, Vector<FunctionAddress> function_addresses, Vector<TableAddress> table_addresses,
-        Vector<MemoryAddress> memory_addresses, Vector<GlobalAddress> global_addresses, Vector<ExportInstance> exports)
+        Vector<MemoryAddress> memory_addresses, Vector<GlobalAddress> global_addresses, Vector<DataAddress> data_addresses,
+        Vector<ExportInstance> exports)
         : m_types(move(types))
         , m_functions(move(function_addresses))
         , m_tables(move(table_addresses))
         , m_memories(move(memory_addresses))
         , m_globals(move(global_addresses))
+        , m_datas(move(data_addresses))
         , m_exports(move(exports))
     {
     }
@@ -232,6 +235,7 @@ public:
     auto& memories() const { return m_memories; }
     auto& globals() const { return m_globals; }
     auto& elements() const { return m_elements; }
+    auto& datas() const { return m_datas; }
     auto& exports() const { return m_exports; }
 
     auto& types() { return m_types; }
@@ -240,6 +244,7 @@ public:
     auto& memories() { return m_memories; }
     auto& globals() { return m_globals; }
     auto& elements() { return m_elements; }
+    auto& datas() { return m_datas; }
     auto& exports() { return m_exports; }
 
 private:
@@ -249,6 +254,7 @@ private:
     Vector<MemoryAddress> m_memories;
     Vector<GlobalAddress> m_globals;
     Vector<ElementAddress> m_elements;
+    Vector<DataAddress> m_datas;
     Vector<ExportInstance> m_exports;
 };
 
@@ -385,6 +391,22 @@ private:
     Value m_value;
 };
 
+class DataInstance {
+public:
+    explicit DataInstance(Vector<u8> data)
+        : m_data(move(data))
+    {
+    }
+
+    size_t size() const { return m_data.size(); }
+
+    Vector<u8>& data() { return m_data; }
+    Vector<u8> const& data() const { return m_data; }
+
+private:
+    Vector<u8> m_data;
+};
+
 class ElementInstance {
 public:
     explicit ElementInstance(ValueType type, Vector<Reference> references)
@@ -409,6 +431,7 @@ public:
     Optional<FunctionAddress> allocate(HostFunction&&);
     Optional<TableAddress> allocate(TableType const&);
     Optional<MemoryAddress> allocate(MemoryType const&);
+    Optional<DataAddress> allocate_data(Vector<u8>);
     Optional<GlobalAddress> allocate(GlobalType const&, Value);
     Optional<ElementAddress> allocate(ValueType const&, Vector<Reference>);
 
@@ -416,6 +439,7 @@ public:
     TableInstance* get(TableAddress);
     MemoryInstance* get(MemoryAddress);
     GlobalInstance* get(GlobalAddress);
+    DataInstance* get(DataAddress);
     ElementInstance* get(ElementAddress);
 
 private:
@@ -424,6 +448,7 @@ private:
     Vector<MemoryInstance> m_memories;
     Vector<GlobalInstance> m_globals;
     Vector<ElementInstance> m_elements;
+    Vector<DataInstance> m_datas;
 };
 
 class Label {
diff --git a/Userland/Libraries/LibWasm/AbstractMachine/BytecodeInterpreter.cpp b/Userland/Libraries/LibWasm/AbstractMachine/BytecodeInterpreter.cpp
index 73ba496167..1b64d8704e 100644
--- a/Userland/Libraries/LibWasm/AbstractMachine/BytecodeInterpreter.cpp
+++ b/Userland/Libraries/LibWasm/AbstractMachine/BytecodeInterpreter.cpp
@@ -227,17 +227,17 @@ void BytecodeInterpreter::pop_and_store(Configuration& configuration, Instructio
     auto entry = configuration.stack().pop();
     auto value = ConvertToRaw<StoreT> {}(*entry.get<Value>().to<PopT>());
     dbgln_if(WASM_TRACE_DEBUG, "stack({}) -> temporary({}b)", value, sizeof(StoreT));
-    store_to_memory(configuration, instruction, { &value, sizeof(StoreT) });
+    auto base_entry = configuration.stack().pop();
+    auto base = base_entry.get<Value>().to<i32>();
+    store_to_memory(configuration, instruction, { &value, sizeof(StoreT) }, *base);
 }
 
-void BytecodeInterpreter::store_to_memory(Configuration& configuration, Instruction const& instruction, ReadonlyBytes data)
+void BytecodeInterpreter::store_to_memory(Configuration& configuration, Instruction const& instruction, ReadonlyBytes data, i32 base)
 {
     auto& address = configuration.frame().module().memories().first();
     auto memory = configuration.store().get(address);
     auto& arg = instruction.arguments().get<Instruction::MemoryArgument>();
-    auto entry = configuration.stack().pop();
-    auto base = entry.get<Value>().to<i32>();
-    u64 instance_address = static_cast<u64>(bit_cast<u32>(base.value())) + arg.offset;
+    u64 instance_address = static_cast<u64>(bit_cast<u32>(base)) + arg.offset;
     Checked addition { instance_address };
     addition += data.size();
     if (addition.has_overflow() || addition.value() > memory->size()) {
@@ -888,7 +888,29 @@ void BytecodeInterpreter::interpret(Configuration& configuration, InstructionPoi
         return unary_operation<double, i64, Operators::SaturatingTruncate<i64>>(configuration);
     case Instructions::i64_trunc_sat_f64_u.value():
         return unary_operation<double, i64, Operators::SaturatingTruncate<u64>>(configuration);
-    case Instructions::memory_init.value():
+    case Instructions::memory_init.value(): {
+        auto data_index = instruction.arguments().get<DataIndex>();
+        auto& data_address = configuration.frame().module().datas()[data_index.value()];
+        auto& data = *configuration.store().get(data_address);
+        auto count = *configuration.stack().pop().get<Value>().to<i32>();
+        auto source_offset = *configuration.stack().pop().get<Value>().to<i32>();
+        auto destination_offset = *configuration.stack().pop().get<Value>().to<i32>();
+
+        TRAP_IF_NOT(count > 0);
+        TRAP_IF_NOT(source_offset + count > 0);
+        TRAP_IF_NOT(static_cast<size_t>(source_offset + count) <= data.size());
+
+        Instruction synthetic_store_instruction {
+            Instructions::i32_store8,
+            Instruction::MemoryArgument { 0, 0 }
+        };
+
+        for (size_t i = 0; i < (size_t)count; ++i) {
+            auto value = data.data()[source_offset + i];
+            store_to_memory(configuration, synthetic_store_instruction, { &value, sizeof(value) }, destination_offset + i);
+        }
+        return;
+    }
     case Instructions::data_drop.value():
     case Instructions::memory_copy.value():
     case Instructions::memory_fill.value():
diff --git a/Userland/Libraries/LibWasm/AbstractMachine/BytecodeInterpreter.h b/Userland/Libraries/LibWasm/AbstractMachine/BytecodeInterpreter.h
index 8528578275..b3e46b05bb 100644
--- a/Userland/Libraries/LibWasm/AbstractMachine/BytecodeInterpreter.h
+++ b/Userland/Libraries/LibWasm/AbstractMachine/BytecodeInterpreter.h
@@ -39,7 +39,7 @@ protected:
     void load_and_push(Configuration&, Instruction const&);
     template<typename PopT, typename StoreT>
     void pop_and_store(Configuration&, Instruction const&);
-    void store_to_memory(Configuration&, Instruction const&, ReadonlyBytes data);
+    void store_to_memory(Configuration&, Instruction const&, ReadonlyBytes data, i32 base);
     void call_address(Configuration&, FunctionAddress);
 
     template<typename PopType, typename PushType, typename Operator>
diff --git a/Userland/Libraries/LibWasm/AbstractMachine/Validator.cpp b/Userland/Libraries/LibWasm/AbstractMachine/Validator.cpp
index 857ca555be..e573e11dfd 100644
--- a/Userland/Libraries/LibWasm/AbstractMachine/Validator.cpp
+++ b/Userland/Libraries/LibWasm/AbstractMachine/Validator.cpp
@@ -103,6 +103,9 @@ ErrorOr<void, ValidationError> Validator::validate(Module& module)
         for (auto& segment : section.segments())
             m_context.elements.unchecked_append(segment.type);
     });
+    module.for_each_section_of_type<DataSection>([this](DataSection const& section) {
+        m_context.datas.resize(section.data().size());
+    });
 
     // FIXME: C.refs is the set funcidx(module with funcs=ϵ with start=ϵ),
     //        i.e., the set of function indices occurring in the module, except in its functions or start function.