Fix handling of bad requests to prevent unnecessary 500 errors (#3432)

Co-authored-by: Nicolas Dorseuil <nicolas@gitbook.io>

packages/gitbook/src/components/SitePage/fetch.ts

@@ -52,7 +52,8 @@ async function resolvePage(context: GitBookSiteContext, params: PagePathParams |
     }
 
     // We don't test path that are too long as GitBook doesn't support them and will return a 404 anyway.
-    if (rawPathname.length <= 512) {
+    // API has a limit of less than 512 characters for the source path, so we use the same limit here.
+    if (rawPathname.length < 512) {
         // Duplicated the regex pattern from SiteRedirectSourcePath API type.
         const SITE_REDIRECT_SOURCE_PATH_REGEX =
             /^\/(?:[A-Za-z0-9\-._~]|%[0-9A-Fa-f]{2})+(?:\/(?:[A-Za-z0-9\-._~]|%[0-9A-Fa-f]{2})+)*$/;

packages/gitbook/src/middleware.ts

@@ -81,6 +81,24 @@ async function validateServerActionRequest(request: NextRequest) {
     }
 }
 
+/**
+ * Filter malicious requests.
+ * @param requestURL The URL of the request to filter.
+ * @returns True if the request is malicious, false otherwise.
+ */
+function shouldFilterMaliciousRequests(requestURL: URL): boolean {
+    // We want to filter hostnames that contains a port here as this is likely a malicious request.
+    if (requestURL.host.includes(':')) {
+        return true;
+    }
+    // These requests will be rejected by the API anyway, we might as well do it right away.
+    if (requestURL.pathname.endsWith(';.jsp')) {
+        return true;
+    }
+
+    return false;
+}
+
 /**
  * Handle request that are targetting the site routes group.
  */
@@ -108,7 +126,7 @@ async function serveSiteRoutes(requestURL: URL, request: NextRequest) {
     }
 
     // We want to filter hostnames that contains a port here as this is likely a malicious request.
-    if (siteRequestURL.host.includes(':')) {
+    if (shouldFilterMaliciousRequests(siteRequestURL)) {
         return new Response('Invalid request', {
             status: 400,
             headers: { 'content-type': 'text/plain' },