WebAssembly: skip unknown sections, add names

Sets the code up so it'll be easier to have section names as strings instead of
hard-coded numbers. Using strings will require synchronizing with sexpr-wasm.

Mostly NFC (besides now skipping *all* unknown sections).

R=titzer@chromium.org

Review URL: https://codereview.chromium.org/1743773002

Cr-Commit-Position: refs/heads/master@{#34464}

src/wasm/decoder.h

@@ -231,9 +231,22 @@ class Decoder {
     return traceOffEnd<uint32_t>();
   }
 
+  // Consume {size} bytes and send them to the bit bucket, advancing {pc_}.
+  void consume_bytes(int size) {
+    if (checkAvailable(size)) {
+      pc_ += size;
+    } else {
+      pc_ = limit_;
+    }
+  }
+
   // Check that at least {size} bytes exist between {pc_} and {limit_}.
   bool checkAvailable(int size) {
-    if (pc_ < start_ || (pc_ + size) > limit_) {
+    intptr_t pc_overflow_value = std::numeric_limits<intptr_t>::max() - size;
+    if (size < 0 || (intptr_t)pc_ > pc_overflow_value) {
+      error(pc_, nullptr, "reading %d bytes would underflow/overflow", size);
+      return false;
+    } else if (pc_ < start_ || limit_ < (pc_ + size)) {
       error(pc_, nullptr, "expected %d bytes, fell off end", size);
       return false;
     } else {

src/wasm/module-decoder.cc

@@ -78,14 +78,21 @@ class ModuleDecoder : public Decoder {
     // Decode the module sections.
     while (pc_ < limit_) {
       TRACE("DecodeSection\n");
-      WasmSectionDeclCode section =
-          static_cast<WasmSectionDeclCode>(consume_u8("section"));
-      // Each section should appear at most once.
-      if (section < kMaxModuleSectionCode) {
-        CheckForPreviousSection(sections, section, false);
-        sections[section] = true;
+      uint8_t section_u8 = consume_u8("section");
+
+      if (section_u8 >= kMaxModuleSectionCode) {
+        // Skip unknown section.
+        int length;
+        uint32_t section_bytes = consume_u32v(&length, "section size");
+        consume_bytes(section_bytes);
+        continue;
       }
 
+      // Each section should appear at most once.
+      auto section = static_cast<WasmSectionDeclCode>(section_u8);
+      CheckForPreviousSection(sections, section, false);
+      sections[section] = true;
+
       switch (section) {
         case kDeclEnd:
           // Terminate section decoding.
@@ -279,23 +286,8 @@ class ModuleDecoder : public Decoder {
           }
           break;
         }
-        case kDeclWLL: {
-          // Reserved for experimentation by the Web Low-level Language project
-          // which is augmenting the binary encoding with source code meta
-          // information. This section does not affect the semantics of the code
-          // and can be ignored by the runtime. https://github.com/JSStats/wll
-          int length = 0;
-          uint32_t section_size = consume_u32v(&length, "section size");
-          if (pc_ + section_size > limit_ || pc_ + section_size < pc_) {
-            error(pc_ - length, "invalid section size");
-            break;
-          }
-          pc_ += section_size;
-          break;
-        }
-        default:
-          error(pc_ - 1, nullptr, "unrecognized section 0x%02x", section);
-          break;
+        case kMaxModuleSectionCode:
+          UNREACHABLE();  // Already skipped unknown sections.
       }
     }
 

src/wasm/wasm-module.h

@@ -25,22 +25,29 @@ const size_t kMaxStringSize = 256;
 const uint32_t kWasmMagic = 0x6d736100;
 const uint32_t kWasmVersion = 0x0a;
 
-enum WasmSectionDeclCode {
-  kDeclMemory = 0x00,
-  kDeclSignatures = 0x01,
-  kDeclFunctions = 0x02,
-  kDeclGlobals = 0x03,
-  kDeclDataSegments = 0x04,
-  kDeclFunctionTable = 0x05,
-  kDeclEnd = 0x06,
-  kDeclStartFunction = 0x07,
-  kDeclImportTable = 0x08,
-  kDeclExportTable = 0x09,
-  kDeclWLL = 0x11,
+// WebAssembly sections are named as strings in the binary format, but
+// internally V8 uses an enum to handle them.
+//
+// Entries have the form F(enumerator, string).
+#define FOR_EACH_WASM_SECTION_TYPE(F)     \
+  F(kDeclMemory, "memory")                \
+  F(kDeclSignatures, "signatures")        \
+  F(kDeclFunctions, "functions")          \
+  F(kDeclGlobals, "globals")              \
+  F(kDeclDataSegments, "data_segments")   \
+  F(kDeclFunctionTable, "function_table") \
+  F(kDeclEnd, "end")                      \
+  F(kDeclStartFunction, "start_function") \
+  F(kDeclImportTable, "import_table")     \
+  F(kDeclExportTable, "export_table")
+
+enum WasmSectionDeclCode : uint32_t {
+#define F(enumerator, string) enumerator,
+  FOR_EACH_WASM_SECTION_TYPE(F)
+#undef F
+      kMaxModuleSectionCode
 };
 
-static const int kMaxModuleSectionCode = 0x11;
-
 enum WasmFunctionDeclBit {
   kDeclFunctionName = 0x01,
   kDeclFunctionImport = 0x02,

test/unittests/wasm/module-decoder-unittest.cc

@@ -932,28 +932,27 @@ TEST_F(WasmFunctionVerifyTest, Ok_v_v_empty) {
   if (result.val) delete result.val;
 }
 
-
-TEST_F(WasmModuleVerifyTest, WLLSectionNoLen) {
+TEST_F(WasmModuleVerifyTest, UnknownSectionNoLen) {
   const byte data[] = {
-      kDeclWLL,  // section without length.
+      kMaxModuleSectionCode,  // unknown section without length.
   };
   EXPECT_FAILURE(data);
 }
 
-
-TEST_F(WasmModuleVerifyTest, WLLSectionEmpty) {
-  static const byte data[] = {
-      kDeclWLL, 0,  // empty section
-  };
-  ModuleResult result = DecodeModule(data, data + arraysize(data));
-  EXPECT_TRUE(result.ok());
-  if (result.val) delete result.val;
+TEST_F(WasmModuleVerifyTest, UnknownSectionEmpty) {
+  for (int i = 0; i < 255 - kMaxModuleSectionCode; ++i) {
+    const byte data[] = {
+        byte(kMaxModuleSectionCode + i), 0,  // empty unknown section
+    };
+    ModuleResult result = DecodeModule(data, data + arraysize(data));
+    EXPECT_TRUE(result.ok());
+    if (result.val) delete result.val;
+  }
 }
 
-
-TEST_F(WasmModuleVerifyTest, WLLSectionOne) {
+TEST_F(WasmModuleVerifyTest, UnknownSectionOne) {
   static const byte data[] = {
-      kDeclWLL,
+      kMaxModuleSectionCode,
       1,  // LEB128 1
       0,  // one byte section
   };
@@ -962,47 +961,108 @@ TEST_F(WasmModuleVerifyTest, WLLSectionOne) {
   if (result.val) delete result.val;
 }
 
-
-TEST_F(WasmModuleVerifyTest, WLLSectionTen) {
+TEST_F(WasmModuleVerifyTest, UnknownSectionTen) {
   static const byte data[] = {
-      kDeclWLL,
-      10,                                    // LEB128 10
-      1,        2, 3, 4, 5, 6, 7, 8, 9, 10,  // 10 byte section
+      kMaxModuleSectionCode,
+      10,  // LEB128 10
+      1,
+      2,
+      3,
+      4,
+      5,
+      6,
+      7,
+      8,
+      9,
+      10,  // 10 byte section
   };
   ModuleResult result = DecodeModule(data, data + arraysize(data));
   EXPECT_TRUE(result.ok());
   if (result.val) delete result.val;
 }
 
-
-TEST_F(WasmModuleVerifyTest, WLLSectionOverflow) {
+TEST_F(WasmModuleVerifyTest, UnknownSectionOverflow) {
   static const byte data[] = {
-      kDeclWLL,
-      11,                                    // LEB128 11
-      1,        2, 3, 4, 5, 6, 7, 8, 9, 10,  // 10 byte section
+      kMaxModuleSectionCode,
+      11,  // LEB128 11
+      1,
+      2,
+      3,
+      4,
+      5,
+      6,
+      7,
+      8,
+      9,
+      10,  // 10 byte section
   };
   EXPECT_FAILURE(data);
 }
 
-
-TEST_F(WasmModuleVerifyTest, WLLSectionUnderflow) {
+TEST_F(WasmModuleVerifyTest, UnknownSectionUnderflow) {
   static const byte data[] = {
-      kDeclWLL, 0xff, 0xff, 0xff, 0xff, 0x0f,  // LEB128 0xffffffff
-      1,        2,    3,    4,                 // 4 byte section
+      kMaxModuleSectionCode,
+      0xff,
+      0xff,
+      0xff,
+      0xff,
+      0x0f,  // LEB128 0xffffffff
+      1,
+      2,
+      3,
+      4,  // 4 byte section
   };
   EXPECT_FAILURE(data);
 }
 
-
-TEST_F(WasmModuleVerifyTest, WLLSectionLoop) {
+TEST_F(WasmModuleVerifyTest, UnknownSectionLoop) {
   // Would infinite loop decoding if wrapping and allowed.
   static const byte data[] = {
-      kDeclWLL, 0xfa, 0xff, 0xff, 0xff, 0x0f,  // LEB128 0xfffffffa
-      1,        2,    3,    4,                 // 4 byte section
+      kMaxModuleSectionCode,
+      0xfa,
+      0xff,
+      0xff,
+      0xff,
+      0x0f,  // LEB128 0xfffffffa
+      1,
+      2,
+      3,
+      4,  // 4 byte section
   };
   EXPECT_FAILURE(data);
 }
 
+TEST_F(WasmModuleVerifyTest, UnknownSectionSkipped) {
+  static const byte data[] = {
+      kMaxModuleSectionCode,
+      1,  // LEB128 1
+      0,  // one byte section
+      kDeclGlobals,
+      1,
+      0,
+      0,
+      0,
+      0,        // name offset
+      kMemI32,  // memory type
+      0,        // exported
+  };
+  ModuleResult result = DecodeModule(data, data + arraysize(data));
+  EXPECT_TRUE(result.ok());
+
+  EXPECT_EQ(1, result.val->globals.size());
+  EXPECT_EQ(0, result.val->functions.size());
+  EXPECT_EQ(0, result.val->data_segments.size());
+
+  WasmGlobal* global = &result.val->globals.back();
+
+  EXPECT_EQ(0, global->name_offset);
+  EXPECT_EQ(MachineType::Int32(), global->type);
+  EXPECT_EQ(0, global->offset);
+  EXPECT_FALSE(global->exported);
+
+  if (result.val) delete result.val;
+}
+
 TEST_F(WasmModuleVerifyTest, ImportTable_empty) {
   static const byte data[] = {kDeclSignatures, 0, kDeclImportTable, 0};
   EXPECT_VERIFIES(data);