Shell execution integration (#3608)

---------

Co-authored-by: Ugaitz Urien <ugaitz.urien@datadoghq.com>
Co-authored-by: Igor Unanua <igor.unanua@datadoghq.com>

.github/workflows/plugins.yml

@@ -334,6 +334,22 @@ jobs:
         uses: ./.github/actions/testagent/logs
       - uses: codecov/codecov-action@v3
 
+  child_process:
+    runs-on: ubuntu-latest
+    env:
+      PLUGINS: child_process
+    steps:
+      - uses: actions/checkout@v2
+      - uses: ./.github/actions/node/setup
+      - run: yarn install
+      - uses: ./.github/actions/node/oldest
+      - run: yarn test:plugins:ci
+      - uses: ./.github/actions/node/20
+      - run: yarn test:plugins:ci
+      - uses: ./.github/actions/node/latest
+      - run: yarn test:plugins:ci
+      - uses: codecov/codecov-action@v2
+
   couchbase:
     runs-on: ubuntu-latest
     services:

LICENSE-3rdparty.csv

@@ -30,6 +30,7 @@ require,protobufjs,BSD-3-Clause,Copyright 2016 Daniel Wirtz
 require,tlhunter-sorted-set,MIT,Copyright (c) 2023 Datadog Inc.
 require,retry,MIT,Copyright 2011 Tim Koschützki Felix Geisendörfer
 require,semver,ISC,Copyright Isaac Z. Schlueter and Contributors
+require,shell-quote,mit,Copyright (c) 2013 James Halliday
 dev,@types/node,MIT,Copyright Authors
 dev,autocannon,MIT,Copyright 2016 Matteo Collina
 dev,aws-sdk,Apache 2.0,Copyright 2012-2017 Amazon.com, Inc. or its affiliates. All Rights Reserved.

ci/init.js

@@ -44,6 +44,7 @@ if (isJestWorker) {
 if (shouldInit) {
   tracer.init(options)
   tracer.use('fs', false)
+  tracer.use('child_process', false)
 }
 
 module.exports = tracer

package.json

@@ -99,6 +99,7 @@
     "protobufjs": "^7.2.5",
     "retry": "^0.13.1",
     "semver": "^7.5.4",
+    "shell-quote": "^1.8.1",
     "tlhunter-sorted-set": "^0.1.0"
   },
   "devDependencies": {

packages/datadog-instrumentations/src/child_process.js

@@ -0,0 +1,150 @@
+'use strict'
+
+const util = require('util')
+
+const {
+  addHook,
+  AsyncResource
+} = require('./helpers/instrument')
+const shimmer = require('../../datadog-shimmer')
+const dc = require('dc-polyfill')
+
+const childProcessChannel = dc.tracingChannel('datadog:child_process:execution')
+
+// ignored exec method because it calls to execFile directly
+const execAsyncMethods = ['execFile', 'spawn']
+const execSyncMethods = ['execFileSync', 'spawnSync']
+
+const names = ['child_process', 'node:child_process']
+
+// child_process and node:child_process returns the same object instance, we only want to add hooks once
+let patched = false
+names.forEach(name => {
+  addHook({ name }, childProcess => {
+    if (!patched) {
+      patched = true
+      shimmer.massWrap(childProcess, execAsyncMethods, wrapChildProcessAsyncMethod())
+      shimmer.massWrap(childProcess, execSyncMethods, wrapChildProcessSyncMethod())
+      shimmer.wrap(childProcess, 'execSync', wrapChildProcessSyncMethod(true))
+    }
+
+    return childProcess
+  })
+})
+
+function normalizeArgs (args, shell) {
+  const childProcessInfo = {
+    command: args[0]
+  }
+
+  if (Array.isArray(args[1])) {
+    childProcessInfo.command = childProcessInfo.command + ' ' + args[1].join(' ')
+    if (args[2] != null && typeof args[2] === 'object') {
+      childProcessInfo.options = args[2]
+    }
+  } else if (args[1] != null && typeof args[1] === 'object') {
+    childProcessInfo.options = args[1]
+  }
+  childProcessInfo.shell = shell ||
+    childProcessInfo.options?.shell === true ||
+    typeof childProcessInfo.options?.shell === 'string'
+
+  return childProcessInfo
+}
+
+function wrapChildProcessSyncMethod (shell = false) {
+  return function wrapMethod (childProcessMethod) {
+    return function () {
+      if (!childProcessChannel.start.hasSubscribers || arguments.length === 0) {
+        return childProcessMethod.apply(this, arguments)
+      }
+
+      const childProcessInfo = normalizeArgs(arguments, shell)
+
+      return childProcessChannel.traceSync(
+        childProcessMethod,
+        {
+          command: childProcessInfo.command,
+          shell: childProcessInfo.shell
+        },
+        this,
+        ...arguments)
+    }
+  }
+}
+
+function wrapChildProcessCustomPromisifyMethod (customPromisifyMethod, shell) {
+  return function () {
+    if (!childProcessChannel.start.hasSubscribers || arguments.length === 0) {
+      return customPromisifyMethod.apply(this, arguments)
+    }
+
+    const childProcessInfo = normalizeArgs(arguments, shell)
+
+    return childProcessChannel.tracePromise(
+      customPromisifyMethod,
+      {
+        command: childProcessInfo.command,
+        shell: childProcessInfo.shell
+      },
+      this,
+      ...arguments)
+  }
+}
+
+function wrapChildProcessAsyncMethod (shell = false) {
+  return function wrapMethod (childProcessMethod) {
+    function wrappedChildProcessMethod () {
+      if (!childProcessChannel.start.hasSubscribers || arguments.length === 0) {
+        return childProcessMethod.apply(this, arguments)
+      }
+
+      const childProcessInfo = normalizeArgs(arguments, shell)
+
+      const innerResource = new AsyncResource('bound-anonymous-fn')
+      return innerResource.runInAsyncScope(() => {
+        childProcessChannel.start.publish({ command: childProcessInfo.command, shell: childProcessInfo.shell })
+
+        const childProcess = childProcessMethod.apply(this, arguments)
+        if (childProcess) {
+          let errorExecuted = false
+
+          childProcess.on('error', (e) => {
+            errorExecuted = true
+            childProcessChannel.error.publish(e)
+          })
+
+          childProcess.on('close', (code) => {
+            code = code || 0
+            if (!errorExecuted && code !== 0) {
+              childProcessChannel.error.publish()
+            }
+            childProcessChannel.asyncEnd.publish({
+              command: childProcessInfo.command,
+              shell: childProcessInfo.shell,
+              result: code
+            })
+          })
+        }
+
+        return childProcess
+      })
+    }
+
+    if (childProcessMethod[util.promisify.custom]) {
+      const wrapedChildProcessCustomPromisifyMethod =
+        shimmer.wrap(childProcessMethod[util.promisify.custom],
+          wrapChildProcessCustomPromisifyMethod(childProcessMethod[util.promisify.custom]), shell)
+
+      // should do it in this way because the original property is readonly
+      const descriptor = Object.getOwnPropertyDescriptor(childProcessMethod, util.promisify.custom)
+      Object.defineProperty(wrappedChildProcessMethod,
+        util.promisify.custom,
+        {
+          ...descriptor,
+          value: wrapedChildProcessCustomPromisifyMethod
+        })
+    }
+    return wrappedChildProcessMethod
+  }
+}

packages/datadog-instrumentations/src/helpers/hooks.js

@@ -30,7 +30,7 @@ module.exports = {
   'body-parser': () => require('../body-parser'),
   'bunyan': () => require('../bunyan'),
   'cassandra-driver': () => require('../cassandra-driver'),
-  'child_process': () => require('../child-process'),
+  'child_process': () => require('../child_process'),
   'connect': () => require('../connect'),
   'cookie': () => require('../cookie'),
   'cookie-parser': () => require('../cookie-parser'),
@@ -78,7 +78,7 @@ module.exports = {
   'mysql2': () => require('../mysql2'),
   'net': () => require('../net'),
   'next': () => require('../next'),
-  'node:child_process': () => require('../child-process'),
+  'node:child_process': () => require('../child_process'),
   'node:crypto': () => require('../crypto'),
   'node:dns': () => require('../dns'),
   'node:http': () => require('../http'),