src: allow building the snapshot on top of a V8 startup blob

CommonEnvironmentSetup::CreateForSnapshotting() always created the
SnapshotCreator without a startup blob, so V8 set up its heap from
scratch. That is impossible when the V8 that Node.js is linked against
only carries the deserializer (external startup data, as Chromium builds
it), and it puts the resulting snapshot on a different read-only heap
lineage than isolates the embedder creates from its own blob. Embedders
in that situation cannot use the snapshot support at all today.

Add SnapshotConfig::base_blob, an optional caller-owned v8::StartupData
that is handed to the SnapshotCreator so that V8 deserializes its heap
from that blob and Node.js adds its isolate data and contexts on top,
the way Blink's context snapshot is built. node_mksnapshot accepts
--v8-snapshot-blob=<file> for hosts that build Node.js with such a V8;
Node.js's own build never passes it and is unchanged. Consuming the
result needs no changes.

embedtest can create a plain V8 startup blob and build the embedder
snapshot on top of one, and a test round-trips argv through a snapshot
built that way.

Signed-off-by: Shelley Vohr <shelley.vohr@gmail.com>
PR-URL: https://github.com/nodejs/node/pull/65374
Reviewed-By: James M Snell <jasnell@gmail.com>

src/api/embed_helpers.cc

@@ -131,6 +131,9 @@ CommonEnvironmentSetup::CommonEnvironmentSetup(
     isolate = impl_->isolate = Isolate::Allocate(GetOrCreateIsolateGroup());
     platform->RegisterIsolate(isolate, loop);
 
+    if (snapshot_config != nullptr && snapshot_config->base_blob != nullptr) {
+      params.snapshot_blob = snapshot_config->base_blob;
+    }
     impl_->snapshot_creator.emplace(isolate, params);
     isolate->SetCaptureStackTraceForUncaughtExceptions(
         true,

src/node.h

@@ -673,6 +673,13 @@ struct SnapshotConfig {
   // the snapshot builder can execute asynchronous operations as long as they
   // are run to completion when the snapshot is taken.
   std::optional<std::string> builder_script_path;
+
+  // A V8 startup blob (as produced by V8's mksnapshot) to build the snapshot
+  // on top of, instead of setting up the V8 heap from scratch. Needed when
+  // the V8 that Node.js is linked against can only deserialize (external
+  // startup data), and to keep the result on the same read-only heap lineage
+  // as the embedder's other isolates. Caller-owned; must outlive the setup.
+  const v8::StartupData* base_blob = nullptr;
 };
 
 struct InspectorParentHandle {

test/embedding/embedtest.cc

@@ -28,6 +28,52 @@ static int RunNodeInstance(MultiIsolatePlatform* platform,
                            const std::vector<std::string>& args,
                            const std::vector<std::string>& exec_args);
 
+// --create-v8-startup-blob <file>: a plain V8 startup blob (what V8's
+// mksnapshot produces), to test building the Node.js snapshot on top of one.
+static int CreateV8StartupBlob(MultiIsolatePlatform* platform,
+                               const std::string& path) {
+  std::unique_ptr<v8::ArrayBuffer::Allocator> allocator(
+      v8::ArrayBuffer::Allocator::NewDefaultAllocator());
+  v8::Isolate::CreateParams params;
+  params.array_buffer_allocator = allocator.get();
+  uv_loop_t loop;
+  assert(uv_loop_init(&loop) == 0);
+  v8::Isolate* isolate = v8::Isolate::Allocate();
+  platform->RegisterIsolate(isolate, &loop);
+  v8::StartupData blob;
+  {
+    v8::SnapshotCreator creator(isolate, params);
+    {
+      v8::HandleScope handle_scope(isolate);
+      creator.SetDefaultContext(v8::Context::New(isolate));
+    }
+    blob =
+        creator.CreateBlob(v8::SnapshotCreator::FunctionCodeHandling::kClear);
+  }
+  bool platform_finished = false;
+  platform->AddIsolateFinishedCallback(
+      isolate,
+      [](void* data) {
+        bool* finished = static_cast<bool*>(data);
+        *finished = true;
+      },
+      &platform_finished);
+  platform->DisposeIsolate(isolate);
+  while (!platform_finished) uv_run(&loop, UV_RUN_ONCE);
+  uv_loop_close(&loop);
+  assert(blob.data != nullptr);
+  FILE* fp = fopen(path.c_str(), "wb");
+  assert(fp != nullptr);
+  size_t written = fwrite(blob.data, blob.raw_size, 1, fp);
+  assert(written == 1);
+  fclose(fp);
+  delete[] blob.data;
+  return 0;
+}
+
+static std::vector<char> base_blob_bytes;
+static v8::StartupData base_blob{nullptr, 0};
+
 NODE_MAIN(int argc, node::argv_type raw_argv[]) {
   char** argv = nullptr;
   node::FixupMain(argc, raw_argv, &argv);
@@ -112,6 +158,27 @@ int RunNodeInstance(MultiIsolatePlatform* platform,
       assert(i + 1 < args.size());
       snapshot_blob_path = args[i + 1];
       i++;
+    } else if (arg == "--create-v8-startup-blob") {
+      assert(i + 1 < args.size());
+      return CreateV8StartupBlob(platform, args[i + 1]);
+    } else if (arg == "--embedder-snapshot-base-blob") {
+      assert(i + 1 < args.size());
+      FILE* fp = fopen(args[i + 1].c_str(), "rb");
+      assert(fp != nullptr);
+      fseek(fp, 0, SEEK_END);
+      base_blob_bytes.resize(ftell(fp));
+      fseek(fp, 0, SEEK_SET);
+      size_t read =
+          fread(base_blob_bytes.data(), base_blob_bytes.size(), 1, fp);
+      assert(read == 1);
+      fclose(fp);
+      base_blob = {base_blob_bytes.data(),
+                   static_cast<int>(base_blob_bytes.size())};
+      if (!snapshot_config.has_value()) {
+        snapshot_config = node::SnapshotConfig{};
+      }
+      snapshot_config.value().base_blob = &base_blob;
+      i++;
     } else {
       filtered_args.push_back(arg);
     }

test/embedding/test-embedding-snapshot-base-blob.js

@@ -0,0 +1,49 @@
+'use strict';
+
+// SnapshotConfig::base_blob: the embedder snapshot can be built on top of an
+// existing V8 startup blob instead of a heap set up from scratch.
+
+const common = require('../common');
+const tmpdir = require('../common/tmpdir');
+const assert = require('assert');
+const fs = require('fs');
+const fixtures = require('../common/fixtures');
+const {
+  spawnSyncAndAssert,
+  spawnSyncAndExitWithoutError,
+} = require('../common/child_process');
+
+const embedtest = common.resolveBuiltBinary('embedtest');
+const snapshotFixture = fixtures.path('snapshot', 'echo-args.js');
+const v8Blob = tmpdir.resolve('v8.blob');
+const nodeBlob = tmpdir.resolve('node-on-v8.blob');
+const buildSnapshotExecArgs = [
+  `eval(require("fs").readFileSync(${JSON.stringify(snapshotFixture)}, "utf8"))`,
+  'arg1', 'arg2',
+];
+
+tmpdir.refresh();
+
+spawnSyncAndExitWithoutError(embedtest, ['--', '--create-v8-startup-blob', v8Blob], { cwd: tmpdir.path });
+assert.ok(fs.statSync(v8Blob).size > 0);
+
+spawnSyncAndExitWithoutError(
+  embedtest,
+  ['--', ...buildSnapshotExecArgs, '--embedder-snapshot-blob', nodeBlob,
+   '--embedder-snapshot-base-blob', v8Blob, '--embedder-snapshot-create'],
+  { cwd: tmpdir.path });
+assert.ok(fs.statSync(nodeBlob).size > fs.statSync(v8Blob).size);
+
+spawnSyncAndAssert(
+  embedtest,
+  ['--', 'arg3', 'arg4', '--embedder-snapshot-blob', nodeBlob],
+  { cwd: tmpdir.path },
+  {
+    stdout(output) {
+      assert.deepStrictEqual(JSON.parse(output), {
+        originalArgv: [embedtest, '__node_anonymous_main', ...buildSnapshotExecArgs],
+        currentArgv: [embedtest, embedtest, 'arg3', 'arg4'],
+      });
+      return true;
+    },
+  });

tools/snapshot/node_mksnapshot.cc

@@ -53,18 +53,41 @@ int main(int argc, char* argv[]) {
   return BuildSnapshot(argc, argv);
 }
 
+static const char kBaseBlobFlag[] = "--v8-snapshot-blob=";
+
 int BuildSnapshot(int argc, char* argv[]) {
-  if (argc < 2) {
-    std::cerr << "Usage: " << argv[0] << " <path/to/output.cc>\n";
-    std::cerr << "       " << argv[0] << " --build-snapshot "
+  std::vector<std::string> args(argv, argv + argc);
+  // --v8-snapshot-blob=<file>: build on top of this V8 startup blob (for
+  // hosts whose V8 uses external startup data) instead of from scratch.
+  std::string base_blob_bytes;
+  v8::StartupData base_blob{nullptr, 0};
+  for (auto it = args.begin(); it != args.end(); ++it) {
+    if (it->starts_with(kBaseBlobFlag)) {
+      std::string path = it->substr(sizeof(kBaseBlobFlag) - 1);
+      args.erase(it);
+      if (node::ReadFileSync(path.c_str(), &base_blob_bytes) != 0) {
+        std::cerr << "Cannot read V8 snapshot blob " << path << "\n";
+        return 1;
+      }
+      base_blob = {base_blob_bytes.data(),
+                   static_cast<int>(base_blob_bytes.size())};
+      break;
+    }
+  }
+
+  if (args.size() < 2) {
+    std::cerr
+        << "Usage: " << argv[0]
+        << " [--v8-snapshot-blob=<path/to/blob.bin>] <path/to/output.cc>\n";
+    std::cerr << "       " << argv[0]
+              << " [--v8-snapshot-blob=<path/to/blob.bin>] --build-snapshot "
               << "<path/to/script.js> <path/to/output.cc>\n";
     return 1;
   }
 
   std::shared_ptr<node::InitializationResult> result =
       node::InitializeOncePerProcess(
-          std::vector<std::string>(argv, argv + argc),
-          node::ProcessInitializationFlags::kGeneratePredictableSnapshot);
+          args, node::ProcessInitializationFlags::kGeneratePredictableSnapshot);
 
   if (result->exit_code() != 0) {
     for (const std::string& error : result->errors()) {
@@ -94,6 +117,7 @@ int BuildSnapshot(int argc, char* argv[]) {
 
   node::SnapshotConfig snapshot_config;
   snapshot_config.builder_script_path = builder_script_path;
+  if (base_blob.data != nullptr) snapshot_config.base_blob = &base_blob;
 
 #ifdef NODE_USE_NODE_CODE_CACHE
   snapshot_config.flags = node::SnapshotFlags::kDefault;