finalize: strip segments that contain only EM_ASM/EM_JS data (WebAssembly#3557)

If we find a data segment whose entire contents is EM_JS or EM_ASM
strings then strip it from the binary.

See: emscripten-core/emscripten#13443

Author: sbc100

src/binaryen-c.h

@@ -1633,9 +1633,9 @@ BINARYEN_API void BinaryenMemoryInitSetSize(BinaryenExpressionRef expr,
 
 // DataDrop
 
-// Gets the index of the segment being dropped by a `memory.drop` expression.
+// Gets the index of the segment being dropped by a `data.drop` expression.
 BINARYEN_API uint32_t BinaryenDataDropGetSegment(BinaryenExpressionRef expr);
-// Sets the index of the segment being dropped by a `memory.drop` expression.
+// Sets the index of the segment being dropped by a `data.drop` expression.
 BINARYEN_API void BinaryenDataDropSetSegment(BinaryenExpressionRef expr,
                                              uint32_t segmentIndex);
 

src/wasm/wasm-emscripten.cpp

@@ -76,8 +76,7 @@ Global* getStackPointerGlobal(Module& wasm) {
 
 const Address UNKNOWN_OFFSET(uint32_t(-1));
 
-std::string escape(const char* input) {
-  std::string code = input;
+std::string escape(std::string code) {
   // replace newlines quotes with escaped newlines
   size_t curr = 0;
   while ((curr = code.find("\\n", curr)) != std::string::npos) {
@@ -109,14 +108,21 @@ class StringConstantTracker {
 public:
   StringConstantTracker(Module& wasm) : wasm(wasm) { calcSegmentOffsets(); }
 
-  std::string codeForConstAddr(int64_t address) {
-    const char* str = stringAtAddr(address);
-    if (!str) {
-      Fatal() << "unable to find data for ASM/EM_JS const at: " << address;
+  const char* stringAtAddr(Address address) {
+    for (unsigned i = 0; i < wasm.memory.segments.size(); ++i) {
+      Memory::Segment& segment = wasm.memory.segments[i];
+      Address offset = segmentOffsets[i];
+      if (offset != UNKNOWN_OFFSET && address >= offset &&
+          address < offset + segment.data.size()) {
+        return &segment.data[address - offset];
+      }
     }
-    return escape(str);
+    Fatal() << "unable to find data for ASM/EM_JS const at: " << address;
+    return nullptr;
   }
 
+  std::vector<Address> segmentOffsets; // segment index => address offset
+
 private:
   void calcSegmentOffsets() {
     std::unordered_map<Index, Address> passiveOffsets;
@@ -172,32 +178,19 @@ class StringConstantTracker {
     }
   }
 
-  const char* stringAtAddr(Address address) {
-    for (unsigned i = 0; i < wasm.memory.segments.size(); ++i) {
-      Memory::Segment& segment = wasm.memory.segments[i];
-      Address offset = segmentOffsets[i];
-      if (offset != UNKNOWN_OFFSET && address >= offset &&
-          address < offset + segment.data.size()) {
-        return &segment.data[address - offset];
-      }
-    }
-    return nullptr;
-  }
-
   Module& wasm;
-  std::vector<Address> segmentOffsets; // segment index => address offset
+};
+
+struct AsmConst {
+  Address id;
+  std::string code;
 };
 
 struct AsmConstWalker : public LinearExecutionWalker<AsmConstWalker> {
   Module& wasm;
   bool minimizeWasmChanges;
   StringConstantTracker stringTracker;
 
-  struct AsmConst {
-    Address id;
-    std::string code;
-  };
-
   std::vector<AsmConst> asmConsts;
   // last sets in the current basic block, per index
   std::map<Index, LocalSet*> sets;
@@ -292,9 +285,8 @@ void AsmConstWalker::visitCall(Call* curr) {
   }
 
   auto* value = arg->cast<Const>();
-  int64_t address = value->value.getInteger();
-  auto code = stringTracker.codeForConstAddr(address);
-  createAsmConst(address, code);
+  Address address = value->value.getInteger();
+  asmConsts.push_back({address, stringTracker.stringAtAddr(address)});
 }
 
 void AsmConstWalker::process() {
@@ -305,24 +297,105 @@ void AsmConstWalker::process() {
   addImports();
 }
 
-void AsmConstWalker::createAsmConst(uint64_t id, std::string code) {
-  AsmConst asmConst;
-  asmConst.id = id;
-  asmConst.code = code;
-  asmConsts.push_back(asmConst);
-}
-
 void AsmConstWalker::addImports() {
   for (auto& import : queuedImports) {
     wasm.addFunction(import.release());
   }
 }
 
-static AsmConstWalker findEmAsmConstsAndReturnWalker(Module& wasm,
-                                                     bool minimizeWasmChanges) {
-  AsmConstWalker walker(wasm, minimizeWasmChanges);
-  walker.process();
-  return walker;
+struct SegmentRemover : WalkerPass<PostWalker<SegmentRemover>> {
+  SegmentRemover(Index segment) : segment(segment) {}
+
+  bool isFunctionParallel() override { return true; }
+
+  Pass* create() override { return new SegmentRemover(segment); }
+
+  void visitMemoryInit(MemoryInit* curr) {
+    if (segment == curr->segment) {
+      Builder builder(*getModule());
+      replaceCurrent(builder.blockify(builder.makeDrop(curr->dest),
+                                      builder.makeDrop(curr->offset),
+                                      builder.makeDrop(curr->size)));
+    }
+  }
+
+  void visitDataDrop(DataDrop* curr) {
+    if (segment == curr->segment) {
+      Builder builder(*getModule());
+      replaceCurrent(builder.makeNop());
+    }
+  }
+
+  Index segment;
+};
+
+static void removeSegment(Module& wasm, Index segment) {
+  PassRunner runner(&wasm);
+  SegmentRemover(segment).run(&runner, &wasm);
+  // Resize the segment to zero.  In theory we should completely remove it
+  // but that would mean re-numbering the segments that follow which is
+  // non-trivial.
+  wasm.memory.segments[segment].data.resize(0);
+}
+
+static Address getExportedAddress(Module& wasm, Export* export_) {
+  Global* g = wasm.getGlobal(export_->value);
+  auto* addrConst = g->init->dynCast<Const>();
+  return addrConst->value.getInteger();
+}
+
+static std::vector<AsmConst> findEmAsmConsts(Module& wasm,
+                                             bool minimizeWasmChanges) {
+  Export* start = wasm.getExportOrNull("__start_em_asm");
+  Export* end = wasm.getExportOrNull("__stop_em_asm");
+
+  // Older versions of emscripten don't export these symbols.  Instead
+  // we run AsmConstWalker in an attempt to derive the string addresses
+  // from the code.
+  if (!start || !end) {
+    AsmConstWalker walker(wasm, minimizeWasmChanges);
+    walker.process();
+    return walker.asmConsts;
+  }
+
+  // Newer version of emscripten export this symbols and we
+  // can use it ot find all the EM_ASM constants.   Sadly __start_em_asm and
+  // __stop_em_asm don't alwasy mark the start and end of segment because in
+  // dynamic linking we merge all data segments into one.
+  std::vector<AsmConst> asmConsts;
+  StringConstantTracker stringTracker(wasm);
+  Address startAddress = getExportedAddress(wasm, start);
+  Address endAddress = getExportedAddress(wasm, end);
+  for (Index i = 0; i < wasm.memory.segments.size(); i++) {
+    Address segmentStart = stringTracker.segmentOffsets[i];
+    size_t segmentSize = wasm.memory.segments[i].data.size();
+    if (segmentStart <= startAddress &&
+        segmentStart + segmentSize >= endAddress) {
+      Address address = startAddress;
+      while (address < endAddress) {
+        auto code = stringTracker.stringAtAddr(address);
+        asmConsts.push_back({address, code});
+        address.addr += strlen(code) + 1;
+      }
+
+      if (segmentStart == startAddress &&
+          segmentStart + segmentSize == endAddress) {
+        removeSegment(wasm, i);
+      } else {
+        // If we can't remove the whole segment then just set the string
+        // data to zero.
+        size_t segmentOffset = startAddress - segmentStart;
+        char* startElem = &wasm.memory.segments[i].data[segmentOffset];
+        memset(startElem, 0, endAddress - startAddress);
+      }
+      break;
+    }
+  }
+
+  assert(asmConsts.size());
+  wasm.removeExport("__start_em_asm");
+  wasm.removeExport("__stop_em_asm");
+  return asmConsts;
 }
 
 struct EmJsWalker : public PostWalker<EmJsWalker> {
@@ -331,6 +404,7 @@ struct EmJsWalker : public PostWalker<EmJsWalker> {
   std::vector<Export> toRemove;
 
   std::map<std::string, std::string> codeByName;
+  std::map<Address, size_t> codeAddresses; // map from address to string len
 
   EmJsWalker(Module& _wasm) : wasm(_wasm), stringTracker(_wasm) {}
 
@@ -353,8 +427,9 @@ struct EmJsWalker : public PostWalker<EmJsWalker> {
     }
     auto* addrConst = consts.list[0];
     int64_t address = addrConst->value.getInteger();
-    auto code = stringTracker.codeForConstAddr(address);
+    auto code = stringTracker.stringAtAddr(address);
     codeByName[funcName] = code;
+    codeAddresses[address] = strlen(code) + 1;
   }
 };
 
@@ -366,6 +441,27 @@ EmJsWalker findEmJsFuncsAndReturnWalker(Module& wasm) {
     wasm.removeExport(exp.name);
     wasm.removeFunction(exp.value);
   }
+
+  // With newer versions of emscripten/llvm we pack all EM_JS strings into
+  // single segment.
+  // We can detect this by checking for segments that contain only JS strings.
+  // When we find such segements we remove them from the final binary.
+  for (Index i = 0; i < wasm.memory.segments.size(); i++) {
+    Address start = walker.stringTracker.segmentOffsets[0];
+    Address cur = start;
+
+    while (cur < start + wasm.memory.segments[i].data.size()) {
+      if (walker.codeAddresses.count(cur) == 0) {
+        break;
+      }
+      cur.addr += walker.codeAddresses[cur];
+    }
+
+    if (cur == start + wasm.memory.segments[i].data.size()) {
+      // Entire segment is contains JS strings.  Remove it.
+      removeSegment(wasm, i);
+    }
+  }
   return walker;
 }
 
@@ -383,16 +479,15 @@ std::string EmscriptenGlueGenerator::generateEmscriptenMetadata() {
   std::stringstream meta;
   meta << "{\n";
 
-  AsmConstWalker emAsmWalker =
-    findEmAsmConstsAndReturnWalker(wasm, minimizeWasmChanges);
+  std::vector<AsmConst> asmConsts = findEmAsmConsts(wasm, minimizeWasmChanges);
 
   // print
   commaFirst = true;
-  if (!emAsmWalker.asmConsts.empty()) {
+  if (!asmConsts.empty()) {
     meta << "  \"asmConsts\": {";
-    for (auto& asmConst : emAsmWalker.asmConsts) {
+    for (auto& asmConst : asmConsts) {
       meta << nextElement();
-      meta << '"' << asmConst.id << "\": \"" << asmConst.code << "\"";
+      meta << '"' << asmConst.id << "\": \"" << escape(asmConst.code) << "\"";
     }
     meta << "\n  },\n";
   }
@@ -405,7 +500,7 @@ std::string EmscriptenGlueGenerator::generateEmscriptenMetadata() {
       auto& name = pair.first;
       auto& code = pair.second;
       meta << nextElement();
-      meta << '"' << name << "\": \"" << code << '"';
+      meta << '"' << name << "\": \"" << escape(code) << '"';
     }
     meta << "\n  },\n";
   }

test/lit/wasm-emscripten-finalize/em_asm.wat

@@ -0,0 +1,29 @@
+;; Test that em_asm string are extracted correctly when the __start_em_asm
+;; and __stop_em_asm globals are exported.
+
+;; RUN: wasm-emscripten-finalize %s -S | filecheck %s
+
+;; Check that the data segment that contains only EM_ASM strings resized to
+;; zero, and that the string are extracted into the metadata.
+
+;; CHECK:      (data (i32.const 100) "normal data")
+;; CHECK-NEXT: (data (i32.const 512) "")
+;; CHECK-NEXT: (data (i32.const 1024) "more data")
+
+;; CHECK:       "asmConsts": {
+;; CHECK-NEXT:     "512": "{ console.log('JS hello'); }",
+;; CHECK-NEXT:     "541": "{ console.log('hello again'); }"
+;; CHECK-NEXT:   },
+
+;; Check that the exports are removed
+;; CHECK-NOT: export
+
+(module
+ (memory 1 1)
+ (global (export "__start_em_asm") i32 (i32.const 512))
+ (global (export "__stop_em_asm") i32 (i32.const 573))
+
+ (data (i32.const 100) "normal data")
+ (data (i32.const 512) "{ console.log('JS hello'); }\00{ console.log('hello again'); }\00")
+ (data (i32.const 1024) "more data")
+)

test/lit/wasm-emscripten-finalize/em_asm_partial.wat

@@ -0,0 +1,24 @@
+;; Test that em_asm string are extraced correctly when the __start_em_asm
+;; and __stop_em_asm globals are exported.
+
+;; RUN: wasm-emscripten-finalize %s -S | filecheck %s
+
+;; Check for the case when __start_em_asm and __stop_em_asm don't define an
+;; entire segment. In this case we preserve the segment but zero the data.
+
+;; CHECK: (data (i32.const 512) "xx\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00yy")
+
+;; CHECK:       "asmConsts": {
+;; CHECK-NEXT:     "514": "{ console.log('JS hello'); }",
+;; CHECK-NEXT:     "543": "{ console.log('hello again'); }"
+;; CHECK-NEXT:   },
+
+;; Check that the exports are removed
+;; CHECK-NOT: export
+
+(module
+ (memory 1 1)
+ (global (export "__start_em_asm") i32 (i32.const 514))
+ (global (export "__stop_em_asm") i32 (i32.const 575))
+ (data (i32.const 512) "xx{ console.log('JS hello'); }\00{ console.log('hello again'); }\00yy")
+)

test/lit/wasm-emscripten-finalize/em_js.wat

@@ -3,20 +3,32 @@
 
 ;; RUN: wasm-emscripten-finalize %s -S | filecheck %s
 
-;; Both functions should be stripped from the binary
+;; All functions should be stripped from the binary, regardless
+;; of internal name
 ;; CHECK-NOT:  (func
 
+;; The data section that contains only em_js strings should
+;; be stripped.
+;; CHECK-NOT: (i32.const 512) "Only em_js strings here\00")
+
+;; Data sections that also contain other stuff should not be stripped
+;; CHECK: (data (i32.const 1024) "some JS string data\00xxx")
+;; CHECK: (data (i32.const 2048) "more JS string data\00yyy")
+
 ;;      CHECK:  "emJsFuncs": {
-;; CHECK-NEXT:    "bar": "more JS string dara",
-;; CHECK-NEXT:    "foo": "some JS string"
+;; CHECK-NEXT:    "bar": "more JS string data",
+;; CHECK-NEXT:    "baz": "Only em_js strings here
+;; CHECK-NEXT:    "foo": "some JS string data"
 ;; CHECK-NEXT:  },
 
 (module
  (memory 1 1)
- (data (i32.const 1024) "some JS string\00")
- (data (i32.const 2048) "more JS string dara\00")
+ (data (i32.const 512) "Only em_js strings here\00")
+ (data (i32.const 1024) "some JS string data\00xxx")
+ (data (i32.const 2048) "more JS string data\00yyy")
  (export "__em_js__foo" (func $__em_js__foo))
  (export "__em_js__bar" (func $bar))
+ (export "__em_js__baz" (func $baz))
  ;; Name matches export name
  (func $__em_js__foo (result i32)
   (i32.const 1024)
@@ -25,4 +37,7 @@
  (func $bar (result i32)
   (i32.const 2048)
  )
+ (func $baz (result i32)
+  (i32.const 512)
+ )
 )