Programmatic injection on nested iframes in extension page

Question

Summary: I need to find a way to accomplish with programmatic injection the same exact behaviour as using content_scripts > matches with "all_frames": true on a manifest. Why? because it is the only way I've found of injecting iframe's content in an extension page without having Cross-Origin errors.

---

I'm moving to optional_permissions on a Chrome extension and I'm on a dead end.

What I want:

Move this behaviour to optional_permissions in order to be able to add more hosts in the future. With the current code, by adding one new host on content_scripts > matches the extension is disabled by Chrome.

For the move, I removed content_scripts in the manifest and I added "optional_permissions": ["*://*/"],. Then, I successfully implemented a dialog asking new permissions to the user with chrome.permissions.request.

As I said before, the problem is how to inject the iframe's content in an extension page.

What I've tried:

1. chrome.declarativeContent.RequestContentScript (mentioned here) with allFrames: true. I can only see the script running if I enter the URL directly, nothing happens when that URL is set in an iframe.
2. chrome.tabs.onUpdated: url is undefined for an extension page. Also, the iframe url is not detected.
3. Call chrome.tabs.executeScript with allFrames: true as soon as I load the first iframe. By doing this I get an exception Cannot access contents of the page. Extension manifest must request permission to access the respective host. and the "respective host" is chrome-extension://, which is not a valid host if you want to add it to the permissions.

I'm lost. I couldn't find a way to simulate the same behaviour as content_scripts > matches with programmatic injection.

Note: using webNavigation API is not an option since the extension is live and it has thousands of users. Because of this, I can not use the frameId property for executeScript. Thus, my only option with executeScript was to inject all frames but the chrome-extension host issue do not let me continue.

---

Update: I was able to accomplish what I wanted but only on an HTTP host. I used chrome.tabs.executeScript (option 3).

The question remains on how to make this work on an extension page.

Answer 1

You cannot run content scripts in any extension page, including your own.

If you want to run code in a subframe of your extension page, then you have to use frameId. There are two ways to do this, with and without webNavigation.

I've put all code snippets in this answer together (with some buttons to invoke the individual code snippets) and shared it at https://robwu.nl/s/optional_permissions-script-subframe.zip
To try it out, download and extract the zip file, load the extension at chrome://extensions and click on the extension button to open the test page.

Request optional permissions

Since the goal is to programmatically run scripts with optional permissions, you need to request the permission. My example will use example.com. If you want to use the webNavigation API too, include its permission in the permission request too.

chrome.permissions.request({
    // permissions: ['webNavigation'], // uncomment if you want this.
    origins: ['*://*.example.com/*'],
}, function(granted) {
    alert('Permission was ' + (granted ? '' : 'not ') + 'granted!');
});

Inject script in subframe

Once you have a tab ID and frameId, injecting scripts in a specific frame is easy. Because of the tabId requirement, this method can only work for frames in tabs, not for frames in your browserAction/pageAction popup or background page!

To demonstrate that code execution succeeds, my examples below will call the next injectInFrame function once the tabId and frameId is known.

function injectInFrame(tabId, frameId) {
    chrome.tabs.executeScript(tabId, {
        frameId,
        code: 'document.body.textContent = "The document content replaced with content at " + new Date().toLocaleString();',
    });
}

If you want to run code not just in the specific frame, but all sub frames of that frame, just add allFrames: true to the chrome.tabs.executeScript call.

Option 1: Use webNavigation to find frameId

Use chrome.tabs.getCurrent to find the ID of the tab where the script runs (or chrome.tabs.query with {active:true,currentWindow:true} if you want to know the current tabId from another script (e.g. background script).

After that, use chrome.webNavigation.getAllFrames to query all frames in the tab. The primary way of identifying a frame is by the URL of the page, so you have a problem if the framed page redirects elsewhere, or if there are multiple frames with the same URL. Here is an example:

// Assuming that you already have a frame in your document,
// i.e. <iframe src="https://example.com"></iframe>
chrome.tabs.getCurrent(function(tab) {
    chrome.webNavigation.getAllFrames({
        tabId: tab.id,
    }, function(frames) {
        for (var frame of frames) {
            if (frame.url === 'https://example.com/') {
                injectInFrame(tab.id, frame.frameId);
                break;
            }
        }
    });
});

Option 2: Use helper page in the frame to find frameId

The option with webNavigation looks simple but has two main disadvantages:

• It requires the webNavigation permission (causing the "Read your browsing history" permission warning)
• The identification of the frame can fail if there are multiple frames with the same URL.

An alternative is to first open an extension page that sends an extension message, and find the frameId (and tab ID) in the metadata that is made available in the second parameter of the chrome.runtime.onMessage listener. This code is more complicated than the other option, but it is more reliable and does not require any additional permissions.

framehelper.html

<script src="framehelper.js"></script>

framehelper.js

var parentOrigin = location.ancestorOrigins[location.ancestorOrigins.length - 1];
if (parentOrigin === location.origin) {
    // Only send a message if the frame was opened by ourselves.
    chrome.runtime.sendMessage(location.hash.slice(1));
}

Code to be run in your extension page:

chrome.runtime.onMessage.addListener(frameMessageListener);
var randomMessage = 'Random message: ' + Math.random();

var f = document.createElement('iframe');
f.src = chrome.runtime.getURL('framehelper.html') + '#' + randomMessage;
document.body.appendChild(f);

function frameMessageListener(msg, sender) {
    if (msg !== randomMessage) return;
    var tabId = sender.tab.id;
    var frameId = sender.frameId;

    chrome.runtime.onMessage.removeListener(frameMessageListener);
    // Note: This will cause the script to be run on the first load.
    // If the frame redirects elsewhere, then the injection can seemingly fail.
    f.addEventListener('load', function onload() {
        f.removeEventListener('load', onload);
        injectInFrame(tabId, frameId);
    });
    f.src = 'https://example.com';
}