Linked URL: add sanitize and normalize

package.json

@@ -230,6 +230,7 @@
     "services/history-v1",
     "services/idp",
     "services/latexqc",
+    "services/linked-url-proxy",
     "services/notifications",
     "services/project-history",
     "services/real-time",

services/linked-url-proxy/Dockerfile

@@ -2,7 +2,7 @@
 # Instead run bin/update_build_scripts from
 # https://github.com/overleaf/internal/
 
-FROM node:20.18.2 AS base
+FROM node:22.18.0 AS base
 
 WORKDIR /overleaf/services/linked-url-proxy
 
@@ -13,16 +13,20 @@ RUN mkdir /home/node/.config && chown node:node /home/node/.config
 FROM base AS app
 
 COPY package.json package-lock.json /overleaf/
+COPY libraries/fetch-utils/package.json /overleaf/libraries/fetch-utils/package.json
 COPY libraries/logger/package.json /overleaf/libraries/logger/package.json
 COPY libraries/metrics/package.json /overleaf/libraries/metrics/package.json
+COPY libraries/o-error/package.json /overleaf/libraries/o-error/package.json
 COPY libraries/settings/package.json /overleaf/libraries/settings/package.json
 COPY services/linked-url-proxy/package.json /overleaf/services/linked-url-proxy/package.json
 COPY patches/ /overleaf/patches/
 
 RUN cd /overleaf && npm ci --quiet
 
+COPY libraries/fetch-utils/ /overleaf/libraries/fetch-utils/
 COPY libraries/logger/ /overleaf/libraries/logger/
 COPY libraries/metrics/ /overleaf/libraries/metrics/
+COPY libraries/o-error/ /overleaf/libraries/o-error/
 COPY libraries/settings/ /overleaf/libraries/settings/
 COPY services/linked-url-proxy/ /overleaf/services/linked-url-proxy/
 

services/linked-url-proxy/app/js/LinkedUrlProxyController.mjs

@@ -1,4 +1,6 @@
 import dns from 'dns/promises'
+import { sanitizeUrl } from 'strict-url-sanitise'
+import normalizeUrlPath from 'als-normalize-urlpath'
 import ipaddr from 'ipaddr.js'
 import { URL } from 'node:url'
 import { Transform } from 'node:stream'
@@ -18,15 +20,7 @@ function isBlockedIp(ipStr, targetUrl) {
   }
 
   const range = addr.range()
-  if ([
-    'loopback',
-    'private',
-    'linkLocal',
-    'multicast',
-    'reserved',
-    'broadcast',
-    'unspecified'
-  ].includes(range)) {
+  if (!['unicast'].includes(addr.range())) {
     return true
   }
 
@@ -44,7 +38,7 @@ function isBlockedIp(ipStr, targetUrl) {
   return false
 }
 
-async function validateSourceUrl(hostname, targetUrl) {
+async function checkUrlAccess(hostname, targetUrl) {
   const records = await dns.lookup(hostname, { all: true }).catch(() => [])
   if (!records.length) {
     const err = new Error(`DNS lookup failed for ${hostname}`)
@@ -62,22 +56,40 @@ async function validateSourceUrl(hostname, targetUrl) {
   }
 }
 
-async function fetchValidated(urlStr, redirectCount = 0) {
+async function validateAndFetch(rawUrl, redirectCount = 0) {
   if (redirectCount > Settings.maxRedirects) {
     const err = new Error('Too many redirects')
     err.info = { status: 421 }
     throw err
   }
 
-  const url = new URL(urlStr)
+  const sanitizedUrl = sanitizeUrl(rawUrl)
+  if (!sanitizedUrl) {
+    const err = new Error(`Invalid or unsafe URL: ${rawUrl}`)
+    err.info = { status: 400 }
+    throw err
+  }
+
+  const url = new URL(sanitizedUrl)
+
   if (!['http:', 'https:'].includes(url.protocol)) {
     const err = new Error(`${url.protocol} protocol is not allowed`)
     err.info = { status: 400 }
     throw err
   }
 
-  // Validate DNS and blocked IPs
-  await validateSourceUrl(url.hostname, urlStr)
+  const normalizedPath = normalizeUrlPath(url.pathname).pathname
+
+  if (!normalizedPath) {
+    const err = new Error(`Invalid or unsafe URL path: ${url.pathname}`)
+    err.info = { status: 400 }
+    throw err
+  }
+
+  const normalizedUrl = url.toString()
+
+  // check DNS and allowed resources
+  await checkUrlAccess(url.hostname, normalizedUrl)
 
   const opts = {
     redirect: 'manual',
@@ -86,7 +98,7 @@ async function fetchValidated(urlStr, redirectCount = 0) {
   }
 
   try {
-    const { stream, response } = await fetchStreamWithResponse(urlStr, opts)
+    const { stream, response } = await fetchStreamWithResponse(normalizedUrl, opts)
 
     const contentLengthHeader = response.headers.get('content-length')
     if (contentLengthHeader) {
@@ -112,8 +124,8 @@ async function fetchValidated(urlStr, redirectCount = 0) {
       if (status >= 300 && status < 400) {
         const location = err.response.headers.get('Location')
         if (location) {
-          const nextUrl = new URL(location, url).toString()
-          return fetchValidated(nextUrl, redirectCount + 1)
+          const nextUrl = new URL(location, normalizedUrl).toString()
+          return validateAndFetch(nextUrl, redirectCount + 1)
         } else {
           const e = new Error('Redirect response missing Location header')
 	  e.info = { status: 421 }
@@ -141,14 +153,14 @@ async function proxy(req, res) {
       return
     }
 
-    const { stream: upstreamStream, response, headers } = await fetchValidated(targetUrl)
+    const { stream: upstreamStream, response, headers } = await validateAndFetch(targetUrl)
 
     res.statusCode = response.status || 200
     res.setHeader('Content-Type', headers['content-type'] || 'application/octet-stream')
     res.setHeader('Cache-Control', 'no-store')
 
     function onError(err) {
-      logger.warn({ err, url: req.url }, 'linked-url-proxy request failed')
+      logger.info({ err, url: req.url }, 'linked-url-proxy request failed')
       try { upstreamStream.destroy() } catch (_) {}
       if (!res.headersSent) {
 	let body = `Error: ${err?.message ?? String(err)}`
@@ -163,11 +175,9 @@ async function proxy(req, res) {
     upstreamStream.pipe(res)
 
   } catch (err) {
-    logger.warn({ err, url: req.url }, 'linked-url-proxy request failed')
-
-    let status = err.info.status
+    const status = err.info?.status || 500
+    logger.info({ linkedUrl: err.message, status, url: req.url }, 'linked-url-proxy request failed')
     let body = `Error: ${err.message || String(err)}`
-
     try {
       res.writeHead(status, { 'Content-Type': 'text/plain' })
       res.end(body)

services/linked-url-proxy/package.json

@@ -2,18 +2,24 @@
   "name": "@overleaf/linked-url-proxy",
   "description": "An API for providing linked url proxy",
   "private": true,
-  "type": "module",
   "main": "app.mjs",
   "scripts": {
-    "start": "node app.mjs"
+    "start": "node app.mjs",
+    "nodemon": "node --watch app.mjs"
   },
-  "version": "0.1.0",
+  "version": "0.1.1",
   "dependencies": {
     "@overleaf/settings": "*",
     "@overleaf/logger": "*",
     "@overleaf/metrics": "*",
     "async": "^3.2.5",
-    "express": "^4.21.2"
-    "ipaddr.js": "^1.9.1"
+    "express": "^4.22.1",
+    "ipaddr.js": "^2.1.0",
+    "als-normalize-urlpath": "^2.3.0",
+    "strict-url-sanitise": "^0.0.1"
+  },
+  "devDependencies": {
+    "als-normalize-urlpath": "^2.3.0",
+    "strict-url-sanitise": "^0.0.1"
   }
 }

services/linked-url-proxy/tsconfig.json

@@ -1,7 +1,7 @@
 {
   "extends": "../../tsconfig.backend.json",
   "include": [
-    "app.js",
+    "app.mjs",
     "app.ts",
     "app/js/**/*",
     "benchmarks/**/*",

services/web/package.json

@@ -74,6 +74,7 @@
     "last 1 year",
     "safari > 14"
   ],
+
   "dependencies": {
     "@ai-sdk/google-vertex": "^4.0.113",
     "@ai-sdk/mcp": "patch:@ai-sdk/mcp@npm%3A1.0.37#~/.yarn/patches/@ai-sdk-mcp-npm-1.0.37-8cd89b8972.patch",
@@ -109,6 +110,7 @@
     "accepts": "^1.3.7",
     "ai": "^6.0.169",
     "ajv": "^8.12.0",
+    "als-normalize-urlpath": "^2.3.0",
     "archiver": "^5.3.0",
     "async": "^3.2.5",
     "base-x": "^4.0.1",
@@ -182,6 +184,7 @@
     "request": "2.88.2",
     "requestretry": "7.1.0",
     "sanitize-html": "^2.8.1",
+    "strict-url-sanitise": "^0.0.1",
     "stripe": "^18.4.0",
     "tough-cookie": "^4.0.0",
     "tsscmp": "^1.0.6",