feat: extract sri from fetch, upgrade to latest spec (#4307)

* feat: extract sri from fetch, upgrade to latest spec

* improve coverage

* fix

* remove ts-check

* Apply suggestions from code review

* fix

* switch order according to my change in the spec

* fix comment

* imrprove with better helper functions

* rename sri to subresource-integrity

* fix

* Update test/fetch/integrity.js

Co-authored-by: Khafra <maitken033380023@gmail.com>

* adapt

---------

Co-authored-by: Khafra <maitken033380023@gmail.com>

Author: Uzlopak

lib/web/fetch/index.js

@@ -14,7 +14,6 @@ const { HeadersList } = require('./headers')
 const { Request, cloneRequest, getRequestDispatcher, getRequestState } = require('./request')
 const zlib = require('node:zlib')
 const {
-  bytesMatch,
   makePolicyContainer,
   clonePolicyContainer,
   requestBadPort,
@@ -62,6 +61,7 @@ const { dataURLProcessor, serializeAMimeType, minimizeSupportedMimeType } = requ
 const { getGlobalDispatcher } = require('../../global')
 const { webidl } = require('../webidl')
 const { STATUS_CODES } = require('node:http')
+const { bytesMatch } = require('../subresource-integrity/subresource-integrity')
 const { createDeferredPromise } = require('../../util/promise')
 const GET_OR_HEAD = ['GET', 'HEAD']
 

lib/web/fetch/util.js

@@ -11,20 +11,6 @@ const assert = require('node:assert')
 const { isUint8Array } = require('node:util/types')
 const { webidl } = require('../webidl')
 
-let supportedHashes = []
-
-// https://nodejs.org/api/crypto.html#determining-if-crypto-support-is-unavailable
-/** @type {import('crypto')} */
-let crypto
-try {
-  crypto = require('node:crypto')
-  const possibleRelevantHashes = ['sha256', 'sha384', 'sha512']
-  supportedHashes = crypto.getHashes().filter((hash) => possibleRelevantHashes.includes(hash))
-/* c8 ignore next 3 */
-} catch {
-
-}
-
 function responseURL (response) {
   // https://fetch.spec.whatwg.org/#responses
   // A response has an associated URL. It is a pointer to the last URL
@@ -698,206 +684,6 @@ function isURLPotentiallyTrustworthy (url) {
   return isOriginPotentiallyTrustworthy(url.origin)
 }
 
-/**
- * @see https://w3c.github.io/webappsec-subresource-integrity/#does-response-match-metadatalist
- * @param {Uint8Array} bytes
- * @param {string} metadataList
- */
-function bytesMatch (bytes, metadataList) {
-  // If node is not built with OpenSSL support, we cannot check
-  // a request's integrity, so allow it by default (the spec will
-  // allow requests if an invalid hash is given, as precedence).
-  /* istanbul ignore if: only if node is built with --without-ssl */
-  if (crypto === undefined) {
-    return true
-  }
-
-  // 1. Let parsedMetadata be the result of parsing metadataList.
-  const parsedMetadata = parseMetadata(metadataList)
-
-  // 2. If parsedMetadata is no metadata, return true.
-  if (parsedMetadata === 'no metadata') {
-    return true
-  }
-
-  // 3. If response is not eligible for integrity validation, return false.
-  // TODO
-
-  // 4. If parsedMetadata is the empty set, return true.
-  if (parsedMetadata.length === 0) {
-    return true
-  }
-
-  // 5. Let metadata be the result of getting the strongest
-  //    metadata from parsedMetadata.
-  const strongest = getStrongestMetadata(parsedMetadata)
-  const metadata = filterMetadataListByAlgorithm(parsedMetadata, strongest)
-
-  // 6. For each item in metadata:
-  for (const item of metadata) {
-    // 1. Let algorithm be the alg component of item.
-    const algorithm = item.algo
-
-    // 2. Let expectedValue be the val component of item.
-    const expectedValue = item.hash
-
-    // See https://github.com/web-platform-tests/wpt/commit/e4c5cc7a5e48093220528dfdd1c4012dc3837a0e
-    // "be liberal with padding". This is annoying, and it's not even in the spec.
-
-    // 3. Let actualValue be the result of applying algorithm to bytes.
-    let actualValue = crypto.createHash(algorithm).update(bytes).digest('base64')
-
-    if (actualValue[actualValue.length - 1] === '=') {
-      if (actualValue[actualValue.length - 2] === '=') {
-        actualValue = actualValue.slice(0, -2)
-      } else {
-        actualValue = actualValue.slice(0, -1)
-      }
-    }
-
-    // 4. If actualValue is a case-sensitive match for expectedValue,
-    //    return true.
-    if (compareBase64Mixed(actualValue, expectedValue)) {
-      return true
-    }
-  }
-
-  // 7. Return false.
-  return false
-}
-
-// https://w3c.github.io/webappsec-subresource-integrity/#grammardef-hash-with-options
-// https://www.w3.org/TR/CSP2/#source-list-syntax
-// https://www.rfc-editor.org/rfc/rfc5234#appendix-B.1
-const parseHashWithOptions = /(?<algo>sha256|sha384|sha512)-((?<hash>[A-Za-z0-9+/]+|[A-Za-z0-9_-]+)={0,2}(?:\s|$)( +[!-~]*)?)?/i
-
-/**
- * @see https://w3c.github.io/webappsec-subresource-integrity/#parse-metadata
- * @param {string} metadata
- */
-function parseMetadata (metadata) {
-  // 1. Let result be the empty set.
-  /** @type {{ algo: string, hash: string }[]} */
-  const result = []
-
-  // 2. Let empty be equal to true.
-  let empty = true
-
-  // 3. For each token returned by splitting metadata on spaces:
-  for (const token of metadata.split(' ')) {
-    // 1. Set empty to false.
-    empty = false
-
-    // 2. Parse token as a hash-with-options.
-    const parsedToken = parseHashWithOptions.exec(token)
-
-    // 3. If token does not parse, continue to the next token.
-    if (
-      parsedToken === null ||
-      parsedToken.groups === undefined ||
-      parsedToken.groups.algo === undefined
-    ) {
-      // Note: Chromium blocks the request at this point, but Firefox
-      // gives a warning that an invalid integrity was given. The
-      // correct behavior is to ignore these, and subsequently not
-      // check the integrity of the resource.
-      continue
-    }
-
-    // 4. Let algorithm be the hash-algo component of token.
-    const algorithm = parsedToken.groups.algo.toLowerCase()
-
-    // 5. If algorithm is a hash function recognized by the user
-    //    agent, add the parsed token to result.
-    if (supportedHashes.includes(algorithm)) {
-      result.push(parsedToken.groups)
-    }
-  }
-
-  // 4. Return no metadata if empty is true, otherwise return result.
-  if (empty === true) {
-    return 'no metadata'
-  }
-
-  return result
-}
-
-/**
- * @param {{ algo: 'sha256' | 'sha384' | 'sha512' }[]} metadataList
- */
-function getStrongestMetadata (metadataList) {
-  // Let algorithm be the algo component of the first item in metadataList.
-  // Can be sha256
-  let algorithm = metadataList[0].algo
-  // If the algorithm is sha512, then it is the strongest
-  // and we can return immediately
-  if (algorithm[3] === '5') {
-    return algorithm
-  }
-
-  for (let i = 1; i < metadataList.length; ++i) {
-    const metadata = metadataList[i]
-    // If the algorithm is sha512, then it is the strongest
-    // and we can break the loop immediately
-    if (metadata.algo[3] === '5') {
-      algorithm = 'sha512'
-      break
-    // If the algorithm is sha384, then a potential sha256 or sha384 is ignored
-    } else if (algorithm[3] === '3') {
-      continue
-    // algorithm is sha256, check if algorithm is sha384 and if so, set it as
-    // the strongest
-    } else if (metadata.algo[3] === '3') {
-      algorithm = 'sha384'
-    }
-  }
-  return algorithm
-}
-
-function filterMetadataListByAlgorithm (metadataList, algorithm) {
-  if (metadataList.length === 1) {
-    return metadataList
-  }
-
-  let pos = 0
-  for (let i = 0; i < metadataList.length; ++i) {
-    if (metadataList[i].algo === algorithm) {
-      metadataList[pos++] = metadataList[i]
-    }
-  }
-
-  metadataList.length = pos
-
-  return metadataList
-}
-
-/**
- * Compares two base64 strings, allowing for base64url
- * in the second string.
- *
-* @param {string} actualValue always base64
- * @param {string} expectedValue base64 or base64url
- * @returns {boolean}
- */
-function compareBase64Mixed (actualValue, expectedValue) {
-  if (actualValue.length !== expectedValue.length) {
-    return false
-  }
-  for (let i = 0; i < actualValue.length; ++i) {
-    if (actualValue[i] !== expectedValue[i]) {
-      if (
-        (actualValue[i] === '+' && expectedValue[i] === '-') ||
-        (actualValue[i] === '/' && expectedValue[i] === '_')
-      ) {
-        continue
-      }
-      return false
-    }
-  }
-
-  return true
-}
-
 // https://w3c.github.io/webappsec-upgrade-insecure-requests/#upgrade-request
 function tryUpgradeRequestToAPotentiallyTrustworthyURL (request) {
   // TODO
@@ -1761,7 +1547,6 @@ module.exports = {
   isValidHeaderValue,
   isErrorLike,
   fullyReadBody,
-  bytesMatch,
   readableStreamClose,
   isomorphicEncode,
   urlIsLocal,
@@ -1770,7 +1555,6 @@ module.exports = {
   readAllBytes,
   simpleRangeHeaderValue,
   buildContentRange,
-  parseMetadata,
   createInflate,
   extractMimeType,
   getDecodeSplit,

lib/web/subresource-integrity/Readme.md

@@ -0,0 +1,9 @@
+# Subresource Integrity
+
+based on Editor’s Draft, 12 June 2025
+
+This module provides support for Subresource Integrity (SRI) in the context of web fetch operations. SRI is a security feature that allows clients to verify that fetched resources are delivered without unexpected manipulation.
+
+## Links
+
+- [Subresource Integrity](https://w3c.github.io/webappsec-subresource-integrity/)