From 3f409b7ca59f0d656b092bd4b0eef7d0dbbf1d82 Mon Sep 17 00:00:00 2001
From: Joyee Cheung <joyeec9h3@gmail.com>
Date: Wed, 2 Jun 2021 13:47:40 +0800
Subject: [PATCH] bootstrap: implement --snapshot-blob and --snapshot-main

This patch introduces --snapshot-main and --snapshot-blob options
for creating and using user land snapshots. At the time of the
creation of this patch, user land CJS modules and ESM are not
yet supported in the snapshot, but a subset of builtins should
already work (in particular modules loaded during bootstrap
are shipped in the built-in snapshot, so they should work
in user land snapshot as well), and support for more builtins
are being added.

To generate a snapshot using main.js as entry point and write
the snapshot blob to snapshot.blob:

$ node --snapshot-main main.js --snapshot-blob snapshot.blob

To restore application state from snapshot.blob:

$ node --snapshot-blob snapshot.blob
---
 doc/api/cli.md                      | 25 ++++++++++
 src/node.cc                         | 73 ++++++++++++++++++++++++++---
 src/node_options.cc                 | 10 ++++
 src/node_options.h                  |  2 +
 src/node_snapshotable.cc            | 63 ++++++++++++++++++++++++-
 src/node_snapshotable.h             |  6 +++
 test/fixtures/snapshot/mutate-fs.js |  3 ++
 test/parallel/test-snapshot-blob.js | 63 +++++++++++++++++++++++++
 8 files changed, 238 insertions(+), 7 deletions(-)
 create mode 100644 test/fixtures/snapshot/mutate-fs.js
 create mode 100644 test/parallel/test-snapshot-blob.js

diff --git a/doc/api/cli.md b/doc/api/cli.md
index 38f72769d9d7ee..b0edc1e81790b4 100644
--- a/doc/api/cli.md
+++ b/doc/api/cli.md
@@ -897,6 +897,30 @@ minimum allocation from the secure heap. The minimum value is `2`.
 The maximum value is the lesser of `--secure-heap` or `2147483647`.
 The value given must be a power of two.
 
+### `--snapshot-blob=path`
+<!-- YAML
+added: REPLACEME
+-->
+
+When used with `--snapshot-main`, `--snapshot-blob` specifies the path
+where the generated snapshot blob will be written to. If not speficied,
+the generated blob will be written, by default, to `snapshot.blob`
+in the current working directory.
+
+When used without `--snapshot-main`, `--snapshot-blob` specifies the
+path to the blob that will be used to restore the application state.
+
+### `--snapshot-main=path`
+<!-- YAML
+added: REPLACEME
+-->
+
+Specifies the path of the entry point file used to build user land
+snapshot. If `--snapshot-blob` is not specified, the generated blob
+will be written, by default, to `snapshot.blob` in the current working
+directory. Otherwise it will be written to the path specified by
+`--snapshot-blob`.
+
 ### `--throw-deprecation`
 <!-- YAML
 added: v0.11.14
@@ -1429,6 +1453,7 @@ Node.js options that are allowed are:
 * `--require`, `-r`
 * `--secure-heap-min`
 * `--secure-heap`
+* `--snapshot-blob`
 * `--throw-deprecation`
 * `--title`
 * `--tls-cipher-list`
diff --git a/src/node.cc b/src/node.cc
index d2ba4734e1aaf6..e6501f4099087d 100644
--- a/src/node.cc
+++ b/src/node.cc
@@ -1100,18 +1100,76 @@ int Start(int argc, char** argv) {
     return result.exit_code;
   }
 
-  {
+  if (!per_process::cli_options->snapshot_main.empty()) {
+    SnapshotData data;
+    {
+      std::string entry;
+      int r =
+          ReadFileSync(&entry, per_process::cli_options->snapshot_main.c_str());
+      if (r != 0) {
+        const char* code = uv_err_name(r);
+        const char* message = uv_strerror(r);
+        FPrintF(stderr,
+                "Failed to open %s. %s: %s\n",
+                per_process::cli_options->snapshot_main.c_str(),
+                code,
+                message);
+        return 1;
+      }
+      node::SnapshotBuilder::Generate(
+          &data, entry, result.args, result.exec_args);
+    }
+
+    std::string snapshot_blob_path;
+    if (!per_process::cli_options->snapshot_blob.empty()) {
+      snapshot_blob_path = per_process::cli_options->snapshot_blob;
+    } else {
+      snapshot_blob_path = std::string("snapshot.blob");
+      char buf[PATH_MAX_BYTES];
+      size_t cwd_size = sizeof(buf);
+      if (uv_cwd(buf, &cwd_size)) {
+        snapshot_blob_path =
+            std::string(buf) + kPathSeparator + std::string("snapshot.blob");
+      }
+    }
+
+    FILE* fp = fopen(snapshot_blob_path.c_str(), "w");
+    if (fp != nullptr) {
+      data.ToBlob(fp);
+      fclose(fp);
+    } else {
+      fprintf(stderr, "Cannot open %s", snapshot_blob_path.c_str());
+      result.exit_code = 1;
+    }
+  } else {
+    SnapshotData snapshot_data;
     Isolate::CreateParams params;
     const std::vector<size_t>* indices = nullptr;
     const EnvSerializeInfo* env_info = nullptr;
     bool force_no_snapshot =
         per_process::cli_options->per_isolate->no_node_snapshot;
     if (!force_no_snapshot) {
-      v8::StartupData* blob = NodeMainInstance::GetEmbeddedSnapshotBlob();
-      if (blob != nullptr) {
-        params.snapshot_blob = blob;
-        indices = NodeMainInstance::GetIsolateDataIndices();
-        env_info = NodeMainInstance::GetEnvSerializeInfo();
+      // TODO(joyee): return const SnapshotData* from the generated source
+      if (per_process::cli_options->snapshot_blob.empty()) {
+        v8::StartupData* blob = NodeMainInstance::GetEmbeddedSnapshotBlob();
+        if (blob != nullptr) {
+          params.snapshot_blob = blob;
+          indices = NodeMainInstance::GetIsolateDataIndices();
+          env_info = NodeMainInstance::GetEnvSerializeInfo();
+        }
+      } else {
+        std::string filename = per_process::cli_options->snapshot_blob;
+        FILE* fp = fopen(filename.c_str(), "r");
+        if (fp != nullptr) {
+          SnapshotData::FromBlob(&snapshot_data, fp);
+          params.snapshot_blob = &(snapshot_data.blob);
+          indices = &(snapshot_data.isolate_data_indices);
+          env_info = &(snapshot_data.env_info);
+          fclose(fp);
+        } else {
+          fprintf(stderr, "Cannot open %s", filename.c_str());
+          result.exit_code = 1;
+        }
       }
     }
     uv_loop_configure(uv_default_loop(), UV_METRICS_IDLE_TIME);
@@ -1123,6 +1181,9 @@ int Start(int argc, char** argv) {
                                    result.exec_args,
                                    indices);
     result.exit_code = main_instance.Run(env_info);
+    if (snapshot_data.blob.data != nullptr) {
+      delete snapshot_data.blob.data;
+    }
   }
 
   TearDownOncePerProcess();
diff --git a/src/node_options.cc b/src/node_options.cc
index b8809df58f0552..70a6f11b9482af 100644
--- a/src/node_options.cc
+++ b/src/node_options.cc
@@ -687,6 +687,16 @@ PerProcessOptionsParser::PerProcessOptionsParser(
             "disable Object.prototype.__proto__",
             &PerProcessOptions::disable_proto,
             kAllowedInEnvironment);
+  AddOption("--snapshot-main",
+            "Path to the entry point file used to build user snapshot",
+            &PerProcessOptions::snapshot_main,
+            kDisallowedInEnvironment);
+  AddOption("--snapshot-blob",
+            "Path to the snapshot blob that's either the result of snapshot"
+            "building, or the blob that is used to restore the application "
+            "state",
+            &PerProcessOptions::snapshot_blob,
+            kAllowedInEnvironment);
 
   // 12.x renamed this inadvertently, so alias it for consistency within the
   // release line, while using the original name for consistency with older
diff --git a/src/node_options.h b/src/node_options.h
index f20234daf7e5f8..1cb4fc482e5e02 100644
--- a/src/node_options.h
+++ b/src/node_options.h
@@ -220,6 +220,8 @@ class PerProcessOptions : public Options {
   bool zero_fill_all_buffers = false;
   bool debug_arraybuffer_allocations = false;
   std::string disable_proto;
+  std::string snapshot_main;
+  std::string snapshot_blob;
 
   std::vector<std::string> security_reverts;
   bool print_bash_completion = false;
diff --git a/src/node_snapshotable.cc b/src/node_snapshotable.cc
index 4e5a53c24b5ecb..4c219f2f18f844 100644
--- a/src/node_snapshotable.cc
+++ b/src/node_snapshotable.cc
@@ -16,12 +16,17 @@
 namespace node {
 
 using v8::Context;
+using v8::Function;
 using v8::HandleScope;
 using v8::Isolate;
 using v8::Local;
+using v8::MaybeLocal;
 using v8::Object;
+using v8::ScriptCompiler;
+using v8::ScriptOrigin;
 using v8::SnapshotCreator;
 using v8::StartupData;
+using v8::String;
 using v8::TryCatch;
 using v8::Value;
 
@@ -522,6 +527,7 @@ const EnvSerializeInfo* NodeMainInstance::GetEnvSerializeInfo() {
 }
 
 void SnapshotBuilder::Generate(SnapshotData* out,
+                               const std::string& entry_file,
                                const std::vector<std::string> args,
                                const std::vector<std::string> exec_args) {
   Isolate* isolate = Isolate::Allocate();
@@ -573,13 +579,62 @@ void SnapshotBuilder::Generate(SnapshotData* out,
       // Run scripts in lib/internal/bootstrap/
       {
         TryCatch bootstrapCatch(isolate);
-        v8::MaybeLocal<Value> result = env->RunBootstrapping();
+        MaybeLocal<Value> result = env->RunBootstrapping();
         if (bootstrapCatch.HasCaught()) {
           PrintCaughtException(isolate, context, bootstrapCatch);
         }
         result.ToLocalChecked();
       }
 
+      // Run the entry point file
+      if (!entry_file.empty()) {
+        TryCatch bootstrapCatch(isolate);
+        std::string filename_s = std::string("node:snapshot_main");
+        Local<String> filename =
+            OneByteString(isolate, filename_s.c_str(), filename_s.size());
+        ScriptOrigin origin(isolate, filename, 0, 0, true);
+        Local<String> source = ToV8Value(context, entry_file, isolate)
+                                   .ToLocalChecked()
+                                   .As<String>();
+        // TODO(joyee): do we need all of these? Maybe we would want a less
+        // internal version of them.
+        std::vector<Local<String>> parameters = {env->require_string(),
+                                                 env->process_string(),
+                                                 env->internal_binding_string(),
+                                                 env->primordials_string()};
+        ScriptCompiler::Source script_source(source, origin);
+        Local<Function> fn;
+        if (!ScriptCompiler::CompileFunctionInContext(
+                 context,
+                 &script_source,
+                 parameters.size(),
+                 parameters.data(),
+                 0,
+                 nullptr,
+                 ScriptCompiler::kEagerCompile)
+                 .ToLocal(&fn)) {
+          if (bootstrapCatch.HasCaught()) {
+            PrintCaughtException(isolate, context, bootstrapCatch);
+          }
+          abort();
+        }
+        std::vector<Local<Value>> args = {env->native_module_require(),
+                                          env->process_object(),
+                                          env->internal_binding_loader(),
+                                          env->primordials()};
+        Local<Value> result;
+        if (!fn->Call(context, Undefined(isolate), args.size(), args.data())
+                 .ToLocal(&result)) {
+          if (bootstrapCatch.HasCaught()) {
+            PrintCaughtException(isolate, context, bootstrapCatch);
+          }
+          abort();
+        }
+        // TODO(joyee): we could use the result for something special, like
+        // setting up initializers that should be invoked at snapshot
+        // dehydration.
+      }
+
       if (per_process::enabled_debug_list.enabled(DebugCategory::MKSNAPSHOT)) {
         env->PrintAllBaseObjects();
         printf("Environment = %p\n", env);
@@ -606,6 +661,12 @@ void SnapshotBuilder::Generate(SnapshotData* out,
   per_process::v8_platform.Platform()->UnregisterIsolate(isolate);
 }
 
+void SnapshotBuilder::Generate(SnapshotData* out,
+                               const std::vector<std::string> args,
+                               const std::vector<std::string> exec_args) {
+  Generate(out, "", args, exec_args);
+}
+
 std::string SnapshotBuilder::Generate(
     const std::vector<std::string> args,
     const std::vector<std::string> exec_args) {
diff --git a/src/node_snapshotable.h b/src/node_snapshotable.h
index 38da68f6d28eb2..5e6b9c5592ad32 100644
--- a/src/node_snapshotable.h
+++ b/src/node_snapshotable.h
@@ -125,6 +125,12 @@ class SnapshotBuilder {
  public:
   static std::string Generate(const std::vector<std::string> args,
                               const std::vector<std::string> exec_args);
+  // Generate the snapshot into out.
+  // entry_file should be the content of the UTF-8 encoded entry files.
+  static void Generate(SnapshotData* out,
+                       const std::string& entry_file,
+                       const std::vector<std::string> args,
+                       const std::vector<std::string> exec_args);
   static void Generate(SnapshotData* out,
                        const std::vector<std::string> args,
                        const std::vector<std::string> exec_args);
diff --git a/test/fixtures/snapshot/mutate-fs.js b/test/fixtures/snapshot/mutate-fs.js
new file mode 100644
index 00000000000000..ae17874455fa5f
--- /dev/null
+++ b/test/fixtures/snapshot/mutate-fs.js
@@ -0,0 +1,3 @@
+const fs = require('fs');
+
+fs.foo = 'I am from the snapshot';
diff --git a/test/parallel/test-snapshot-blob.js b/test/parallel/test-snapshot-blob.js
new file mode 100644
index 00000000000000..e3a23a36c0e010
--- /dev/null
+++ b/test/parallel/test-snapshot-blob.js
@@ -0,0 +1,63 @@
+'use strict';
+
+require('../common');
+const assert = require('assert');
+const { spawnSync } = require('child_process');
+const tmpdir = require('../common/tmpdir');
+const fixtures = require('../common/fixtures');
+const path = require('path');
+const fs = require('fs');
+
+tmpdir.refresh();
+const blobPath = path.join(tmpdir.path, 'my-snapshot.blob');
+const file = fixtures.path('snapshot', 'mutate-fs.js');
+
+{
+  const child = spawnSync(process.execPath, [
+    '--snapshot-main',
+    file,
+  ], {
+    cwd: tmpdir.path
+  });
+  if (child.status !== 0) {
+    console.log(child.stderr.toString());
+    console.log(child.stdout.toString());
+    assert.strictEqual(child.status, 0);
+  }
+  const stats = fs.statSync(path.join(tmpdir.path, 'snapshot.blob'));
+  assert(stats.isFile());
+}
+
+{
+  let child = spawnSync(process.execPath, [
+    '--snapshot-main',
+    file,
+    '--snapshot-blob',
+    blobPath,
+  ], {
+    cwd: tmpdir.path
+  });
+  if (child.status !== 0) {
+    console.log(child.stderr.toString());
+    console.log(child.stdout.toString());
+    assert.strictEqual(child.status, 0);
+  }
+  const stats = fs.statSync(blobPath);
+  assert(stats.isFile());
+
+  child = spawnSync(process.execPath, [
+    '--snapshot-blob',
+    blobPath,
+    '-p',
+    'require("fs").foo',
+  ], {
+    cwd: tmpdir.path
+  });
+
+  if (child.status !== 0) {
+    console.log(child.stderr.toString());
+    console.log(child.stdout.toString());
+    assert.strictEqual(child.status, 0);
+  }
+  assert(/I am from the snapshot/.test(child.stdout.toString()));
+}