1
|
/**
|
2
|
* This file is part of Haketilo.
|
3
|
*
|
4
|
* Function: Main content script that runs in all frames.
|
5
|
*
|
6
|
* Copyright (C) 2021 Wojtek Kosior
|
7
|
* Copyright (C) 2021 jahoti
|
8
|
*
|
9
|
* This program is free software: you can redistribute it and/or modify
|
10
|
* it under the terms of the GNU General Public License as published by
|
11
|
* the Free Software Foundation, either version 3 of the License, or
|
12
|
* (at your option) any later version.
|
13
|
*
|
14
|
* This program is distributed in the hope that it will be useful,
|
15
|
* but WITHOUT ANY WARRANTY; without even the implied warranty of
|
16
|
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
17
|
* GNU General Public License for more details.
|
18
|
*
|
19
|
* As additional permission under GNU GPL version 3 section 7, you
|
20
|
* may distribute forms of that code without the copy of the GNU
|
21
|
* GPL normally required by section 4, provided you include this
|
22
|
* license notice and, in case of non-source distribution, a URL
|
23
|
* through which recipients can access the Corresponding Source.
|
24
|
* If you modify file(s) with this exception, you may extend this
|
25
|
* exception to your version of the file(s), but you are not
|
26
|
* obligated to do so. If you do not wish to do so, delete this
|
27
|
* exception statement from your version.
|
28
|
*
|
29
|
* As a special exception to the GPL, any HTML file which merely
|
30
|
* makes function calls to this code, and for that purpose
|
31
|
* includes it by reference shall be deemed a separate work for
|
32
|
* copyright law purposes. If you modify this code, you may extend
|
33
|
* this exception to your version of the code, but you are not
|
34
|
* obligated to do so. If you do not wish to do so, delete this
|
35
|
* exception statement from your version.
|
36
|
*
|
37
|
* You should have received a copy of the GNU General Public License
|
38
|
* along with this program. If not, see <https://www.gnu.org/licenses/>.
|
39
|
*
|
40
|
* I, Wojtek Kosior, thereby promise not to sue for violation of this file's
|
41
|
* license. Although I request that you do not make use this code in a
|
42
|
* proprietary program, I am not going to enforce this in court.
|
43
|
*/
|
44
|
|
45
|
/*
|
46
|
* IMPORTS_START
|
47
|
* IMPORT handle_page_actions
|
48
|
* IMPORT extract_signed
|
49
|
* IMPORT sign_data
|
50
|
* IMPORT gen_nonce
|
51
|
* IMPORT is_privileged_url
|
52
|
* IMPORT is_chrome
|
53
|
* IMPORT is_mozilla
|
54
|
* IMPORT start_activity_info_server
|
55
|
* IMPORT make_csp_rule
|
56
|
* IMPORT csp_header_regex
|
57
|
* IMPORTS_END
|
58
|
*/
|
59
|
|
60
|
document.content_loaded = document.readyState === "complete";
|
61
|
const wait_loaded = e => e.content_loaded ? Promise.resolve() :
|
62
|
new Promise(c => e.addEventListener("DOMContentLoaded", c, {once: true}));
|
63
|
|
64
|
wait_loaded(document).then(() => document.content_loaded = true);
|
65
|
|
66
|
function extract_cookie_policy(cookie, min_time)
|
67
|
{
|
68
|
let best_result = {time: -1};
|
69
|
let policy = null;
|
70
|
const extracted_signatures = [];
|
71
|
|
72
|
for (const match of cookie.matchAll(/haketilo-(\w*)=([^;]*)/g)) {
|
73
|
const new_result = extract_signed(...match.slice(1, 3));
|
74
|
if (new_result.fail)
|
75
|
continue;
|
76
|
|
77
|
extracted_signatures.push(match[1]);
|
78
|
|
79
|
if (new_result.time < Math.max(min_time, best_result.time))
|
80
|
continue;
|
81
|
|
82
|
/* This should succeed - it's our self-produced valid JSON. */
|
83
|
const new_policy = JSON.parse(decodeURIComponent(new_result.data));
|
84
|
if (new_policy.url !== document.URL)
|
85
|
continue;
|
86
|
|
87
|
best_result = new_result;
|
88
|
policy = new_policy;
|
89
|
}
|
90
|
|
91
|
return [policy, extracted_signatures];
|
92
|
}
|
93
|
|
94
|
function extract_url_policy(url, min_time)
|
95
|
{
|
96
|
const [base_url, payload, anchor] =
|
97
|
/^([^#]*)#?([^#]*)(#?.*)$/.exec(url).splice(1, 4);
|
98
|
|
99
|
const match = /^haketilo_([^_]+)_(.*)$/.exec(payload);
|
100
|
if (!match)
|
101
|
return [null, url];
|
102
|
|
103
|
const result = extract_signed(...match.slice(1, 3));
|
104
|
if (result.fail)
|
105
|
return [null, url];
|
106
|
|
107
|
const original_url = base_url + anchor;
|
108
|
const policy = result.time < min_time ? null :
|
109
|
JSON.parse(decodeURIComponent(result.data));
|
110
|
|
111
|
return [policy.url === original_url ? policy : null, original_url];
|
112
|
}
|
113
|
|
114
|
function employ_nonhttp_policy(policy)
|
115
|
{
|
116
|
if (!policy.allow)
|
117
|
return;
|
118
|
|
119
|
policy.nonce = gen_nonce();
|
120
|
const [base_url, target] = /^([^#]*)(#?.*)$/.exec(policy.url).slice(1, 3);
|
121
|
const encoded_policy = encodeURIComponent(JSON.stringify(policy));
|
122
|
const payload = "haketilo_" +
|
123
|
sign_data(encoded_policy, new Date().getTime()).join("_");
|
124
|
const resulting_url = `${base_url}#${payload}${target}`;
|
125
|
location.href = resulting_url;
|
126
|
location.reload();
|
127
|
}
|
128
|
|
129
|
/*
|
130
|
* In the case of HTML documents:
|
131
|
* 1. When injecting some payload we need to sanitize <meta> CSP tags before
|
132
|
* they reach the document.
|
133
|
* 2. Only <meta> tags inside <head> are considered valid by the browser and
|
134
|
* need to be considered.
|
135
|
* 3. We want to detach <html> from document, wait until its <head> completes
|
136
|
* loading, sanitize it and re-attach <html>.
|
137
|
* 4. We shall wait for anything to appear in or after <body> and take that as
|
138
|
* a sign <head> has finished loading.
|
139
|
* 5. Otherwise, getting the `DOMContentLoaded' event on the document shall also
|
140
|
* be a sign that <head> is fully loaded.
|
141
|
*/
|
142
|
|
143
|
function make_body_start_observer(DOM_element, waiting)
|
144
|
{
|
145
|
const observer = new MutationObserver(() => try_body_started(waiting));
|
146
|
observer.observe(DOM_element, {childList: true});
|
147
|
return observer;
|
148
|
}
|
149
|
|
150
|
function try_body_started(waiting)
|
151
|
{
|
152
|
const body = waiting.detached_html.querySelector("body");
|
153
|
|
154
|
if ((body && (body.firstChild || body.nextSibling)) ||
|
155
|
waiting.doc.documentElement.nextSibling) {
|
156
|
finish_waiting(waiting);
|
157
|
return true;
|
158
|
}
|
159
|
|
160
|
if (body && waiting.observers.length < 2)
|
161
|
waiting.observers.push(make_body_start_observer(body, waiting));
|
162
|
}
|
163
|
|
164
|
function finish_waiting(waiting)
|
165
|
{
|
166
|
if (waiting.finished)
|
167
|
return;
|
168
|
waiting.finished = true;
|
169
|
waiting.observers.forEach(observer => observer.disconnect());
|
170
|
setTimeout(waiting.callback, 0);
|
171
|
}
|
172
|
|
173
|
function _wait_for_head(doc, detached_html, callback)
|
174
|
{
|
175
|
const waiting = {doc, detached_html, callback, observers: []};
|
176
|
|
177
|
if (try_body_started(waiting))
|
178
|
return;
|
179
|
|
180
|
waiting.observers = [make_body_start_observer(detached_html, waiting)];
|
181
|
|
182
|
wait_loaded(doc).then(() => finish_waiting(waiting));
|
183
|
}
|
184
|
|
185
|
function wait_for_head(doc, detached_html)
|
186
|
{
|
187
|
return new Promise(cb => _wait_for_head(doc, detached_html, cb));
|
188
|
}
|
189
|
|
190
|
const blocked_str = "blocked";
|
191
|
|
192
|
function block_attribute(node, attr, ns=null)
|
193
|
{
|
194
|
const [hasa, geta, seta, rema] = ["has", "get", "set", "remove"]
|
195
|
.map(m => (n, ...args) => typeof ns === "string" ?
|
196
|
n[`${m}AttributeNS`](ns, ...args) : n[`${m}Attribute`](...args));
|
197
|
/*
|
198
|
* Disabling attributes by prepending `-blocked' allows them to still be
|
199
|
* relatively easily accessed in case they contain some useful data.
|
200
|
*/
|
201
|
const construct_name = [attr];
|
202
|
while (hasa(node, construct_name.join("")))
|
203
|
construct_name.unshift(blocked_str);
|
204
|
|
205
|
while (construct_name.length > 1) {
|
206
|
construct_name.shift();
|
207
|
const name = construct_name.join("");
|
208
|
seta(node, `${blocked_str}-${name}`, geta(node, name));
|
209
|
}
|
210
|
|
211
|
rema(node, attr);
|
212
|
}
|
213
|
|
214
|
/*
|
215
|
* Used to disable `<script>'s and `<meta>'s that have not yet been added to
|
216
|
* live DOM (doesn't work for those already added).
|
217
|
*/
|
218
|
function sanitize_meta(meta)
|
219
|
{
|
220
|
if (csp_header_regex.test(meta.httpEquiv) && meta.content)
|
221
|
block_attribute(meta, "content");
|
222
|
}
|
223
|
|
224
|
function sanitize_script(script)
|
225
|
{
|
226
|
script.haketilo_blocked_type = script.getAttribute("type");
|
227
|
script.type = "text/plain";
|
228
|
}
|
229
|
|
230
|
/*
|
231
|
* Executed after `<script>' has been connected to the DOM, when it is no longer
|
232
|
* eligible for being executed by the browser.
|
233
|
*/
|
234
|
function desanitize_script(script)
|
235
|
{
|
236
|
script.setAttribute("type", script.haketilo_blocked_type);
|
237
|
|
238
|
if ([null, undefined].includes(script.haketilo_blocked_type))
|
239
|
script.removeAttribute("type");
|
240
|
|
241
|
delete script.haketilo_blocked_type;
|
242
|
}
|
243
|
|
244
|
const bad_url_reg = /^data:([^,;]*ml|unknown-content-type)/i;
|
245
|
function sanitize_urls(element)
|
246
|
{
|
247
|
for (const attr of [...element.attributes || []]
|
248
|
.filter(attr => /^(href|src|data)$/i.test(attr.localName))
|
249
|
.filter(attr => bad_url_reg.test(attr.value)))
|
250
|
block_attribute(element, attr.localName, attr.namespaceURI);
|
251
|
}
|
252
|
|
253
|
function start_data_urls_sanitizing(doc)
|
254
|
{
|
255
|
doc.querySelectorAll("*[href], *[src], *[data]").forEach(sanitize_urls);
|
256
|
if (!doc.content_loaded) {
|
257
|
const mutation_handler = m => m.addedNodes.forEach(sanitize_urls);
|
258
|
const mo = new MutationObserver(ms => ms.forEach(mutation_handler));
|
259
|
mo.observe(doc, {childList: true, subtree: true});
|
260
|
wait_loaded(doc).then(() => mo.disconnect());
|
261
|
}
|
262
|
}
|
263
|
|
264
|
/*
|
265
|
* Normally, we block scripts with CSP. However, Mozilla does optimizations that
|
266
|
* cause part of the DOM to be loaded when our content scripts get to run. Thus,
|
267
|
* before the CSP rules we inject (for non-HTTP pages) become effective, we need
|
268
|
* to somehow block the execution of `<script>'s and intrinsics that were
|
269
|
* already there. Additionally, some browsers (IceCat 60) seem to have problems
|
270
|
* applying this CSP to non-inline `<scripts>' in certain scenarios.
|
271
|
*/
|
272
|
function prevent_script_execution(event)
|
273
|
{
|
274
|
if (!event.target.haketilo_payload)
|
275
|
event.preventDefault();
|
276
|
}
|
277
|
|
278
|
function mozilla_initial_block(doc)
|
279
|
{
|
280
|
doc.addEventListener("beforescriptexecute", prevent_script_execution);
|
281
|
|
282
|
for (const elem of doc.querySelectorAll("*")) {
|
283
|
[...elem.attributes].map(attr => attr.localName)
|
284
|
.filter(attr => /^on/.test(attr) && elem.wrappedJSObject[attr])
|
285
|
.forEach(attr => elem.wrappedJSObject[attr] = null);
|
286
|
}
|
287
|
}
|
288
|
|
289
|
/*
|
290
|
* Here we block all scripts of a document which might be either and
|
291
|
* HTMLDocument or an XMLDocument. Modifying an XML document might disrupt
|
292
|
* Mozilla's XML preview. This is an unfortunate thing we have to accept for
|
293
|
* now. XML documents *have to* be sanitized as well because they might
|
294
|
* contain `<script>' tags (or on* attributes) with namespace declared as
|
295
|
* "http://www.w3.org/1999/xhtml" or "http://www.w3.org/2000/svg" which allows
|
296
|
* javascript execution.
|
297
|
*/
|
298
|
async function sanitize_document(doc, policy)
|
299
|
{
|
300
|
/*
|
301
|
* Blocking of scripts that are in the DOM from the beginning. Needed for
|
302
|
* Mozilla.
|
303
|
*/
|
304
|
if (is_mozilla)
|
305
|
mozilla_initial_block(doc);
|
306
|
|
307
|
/*
|
308
|
* Ensure our CSP rules are employed from the beginning. This CSP injection
|
309
|
* method is, when possible, going to be applied together with CSP rules
|
310
|
* injected using webRequest.
|
311
|
* Using elements namespaced as HTML makes this CSP injection also work for
|
312
|
* non-HTML documents.
|
313
|
*/
|
314
|
const html = new DOMParser().parseFromString(`<html><head><meta \
|
315
|
http-equiv="Content-Security-Policy" content="${make_csp_rule(policy)}"\
|
316
|
/></head><body>Loading...</body></html>`, "text/html").documentElement;
|
317
|
|
318
|
/*
|
319
|
* Root node gets hijacked now, to be re-attached after <head> is loaded
|
320
|
* and sanitized.
|
321
|
*/
|
322
|
const root = doc.documentElement;
|
323
|
root.replaceWith(html);
|
324
|
|
325
|
/*
|
326
|
* When we don't inject payload, we neither block document's CSP `<meta>'
|
327
|
* tags nor wait for `<head>' to be parsed.
|
328
|
*/
|
329
|
if (policy.has_payload) {
|
330
|
await wait_for_head(doc, root);
|
331
|
|
332
|
root.querySelectorAll("head meta")
|
333
|
.forEach(m => sanitize_meta(m, policy));
|
334
|
}
|
335
|
|
336
|
root.querySelectorAll("script").forEach(s => sanitize_script(s, policy));
|
337
|
html.replaceWith(root);
|
338
|
root.querySelectorAll("script").forEach(s => desanitize_script(s, policy));
|
339
|
|
340
|
start_data_urls_sanitizing(doc);
|
341
|
}
|
342
|
|
343
|
async function disable_service_workers()
|
344
|
{
|
345
|
if (!navigator.serviceWorker)
|
346
|
return;
|
347
|
|
348
|
const registrations = await navigator.serviceWorker.getRegistrations();
|
349
|
if (registrations.length === 0)
|
350
|
return;
|
351
|
|
352
|
console.warn("Service Workers detected on this page! Unregistering and reloading");
|
353
|
|
354
|
try {
|
355
|
await Promise.all(registrations.map(r => r.unregister()));
|
356
|
} finally {
|
357
|
location.reload();
|
358
|
}
|
359
|
|
360
|
/* Never actually return! */
|
361
|
return new Promise(() => 0);
|
362
|
}
|
363
|
|
364
|
if (!is_privileged_url(document.URL)) {
|
365
|
let policy_received_callback = () => undefined;
|
366
|
let policy;
|
367
|
|
368
|
/* Signature valid for half an hour. */
|
369
|
const min_time = new Date().getTime() - 1800 * 1000;
|
370
|
|
371
|
if (/^https?:/.test(document.URL)) {
|
372
|
let signatures;
|
373
|
[policy, signatures] = extract_cookie_policy(document.cookie, min_time);
|
374
|
for (const signature of signatures)
|
375
|
document.cookie = `haketilo-${signature}=; Max-Age=-1;`;
|
376
|
} else {
|
377
|
const scheme = /^([^:]*)/.exec(document.URL)[1];
|
378
|
const known_scheme = ["file", "ftp"].includes(scheme);
|
379
|
|
380
|
if (!known_scheme)
|
381
|
console.warn(`Unknown url scheme: \`${scheme}'!`);
|
382
|
|
383
|
let original_url;
|
384
|
[policy, original_url] = extract_url_policy(document.URL, min_time);
|
385
|
history.replaceState(null, "", original_url);
|
386
|
|
387
|
if (known_scheme && !policy)
|
388
|
policy_received_callback = employ_nonhttp_policy;
|
389
|
}
|
390
|
|
391
|
if (!policy) {
|
392
|
console.debug("Using fallback policy!");
|
393
|
policy = {allow: false, nonce: gen_nonce()};
|
394
|
}
|
395
|
|
396
|
if (!(document instanceof HTMLDocument))
|
397
|
policy.has_payload = false;
|
398
|
|
399
|
console.debug("current policy", policy);
|
400
|
|
401
|
const doc_ready = Promise.all([
|
402
|
policy.allow ? Promise.resolve() : sanitize_document(document, policy),
|
403
|
policy.allow ? Promise.resolve() : disable_service_workers(),
|
404
|
wait_loaded(document)
|
405
|
]);
|
406
|
|
407
|
handle_page_actions(policy.nonce, policy_received_callback, doc_ready);
|
408
|
|
409
|
start_activity_info_server();
|
410
|
}
|