Files
cacti/tests/Unit/Security/Csrf/CsrfBootstrapOrderTest.php
Thomas VincentandTheWitness cb49f31ff9 Replace csrf-magic with maintained CSRF protection (#7744)
* build: add symfony/security-csrf

Backs the CSRF token core that replaces the vendored csrf-magic fork.

Signed-off-by: Thomas Vincent <thomasvincent@gmail.com>

* test: fix security-csrf package path assertion

symfony/security-csrf ships CsrfTokenManager.php at the package root, not under src/.

Signed-off-by: Thomas Vincent <thomasvincent@gmail.com>

* refactor: register the composer autoloader before the CSRF include

CSRF validation ran before the autoloader existed, which blocks any class-based guard.

Signed-off-by: Thomas Vincent <thomasvincent@gmail.com>

* feat(csrf): add CactiCsrfGuard over symfony/security-csrf

Issues and validates session-scoped tokens; not yet wired into the request path.

Signed-off-by: Thomas Vincent <thomasvincent@gmail.com>

* feat(csrf): accept pre-upgrade csrf-magic tokens during the grace window

Legacy tokens carry their own expiry, so the fallback closes itself without a config flag.

Signed-off-by: Thomas Vincent <thomasvincent@gmail.com>

* feat(csrf): port the output buffer rewriter to CactiCsrfGuard

Keeps the hidden field name and JS globals that third-party plugins rely on.

Signed-off-by: Thomas Vincent <thomasvincent@gmail.com>

* fix(csrf): use json_encode for the JS script context in rewriteBuffer

htmlspecialchars() protects the HTML attribute context but not the
<script> string literal, which relied on Symfony's token generator
never emitting a quote rather than being safe by construction.

Signed-off-by: Thomas Vincent <thomasvincent@gmail.com>

* refactor(csrf): move the XHR shim to include/js as Cacti-owned code

It has been maintained by the Cacti Group since 2020 and is not an upstream package.

Signed-off-by: Thomas Vincent <thomasvincent@gmail.com>

* refactor(csrf): retire the vendored csrf-magic fork

The procedural csrf_* surface stays as a deprecated shim so plugins keep working.

Signed-off-by: Thomas Vincent <thomasvincent@gmail.com>

* feat(csrf): validate requests through CactiCsrfGuard

csrf-magic self-invoked from its own file; the guard is called explicitly.

Signed-off-by: Thomas Vincent <thomasvincent@gmail.com>

* fix(install): stop gating the permission step on the csrf secret

Retiring csrf-magic removed the directory its default path lives in, and setStep() clamps to the first errored step, so no fresh install could advance.

Signed-off-by: Thomas Vincent <thomasvincent@gmail.com>

* fix(csrf): key the legacy fallback on the raw secret bytes

refresh_csrf.php writes PHP source with a trailing newline, and trimming that yields a different HMAC key than csrf-magic used.

Signed-off-by: Thomas Vincent <thomasvincent@gmail.com>

* fix(csrf): log when the guard degrades to disabled in a web context

The installer is the only web path that can reach it, and failing open in silence is not something csrf-magic could do.

Signed-off-by: Thomas Vincent <thomasvincent@gmail.com>

* docs: changelog for the CSRF library replacement

Signed-off-by: Thomas Vincent <thomasvincent@gmail.com>

* fix(csrf): mint the token before the session write callback closes it

csrf_startup() minted lazily from the output buffer callback, which runs after session_write_close()'s shutdown-function registration, so a token minted there on a legacy-fallback POST was silently dropped.

Signed-off-by: Thomas Vincent <thomasvincent@gmail.com>

* fix(csrf): make refresh_csrf.php a no-op

Its default secret path no longer ships and live tokens are session-bound, not HMACs over the secret, so rotation neither worked nor did anything meaningful; running it during the upgrade grace window used to delete the legacy secret and kill validation for pre-upgrade tokens outright.

Signed-off-by: Thomas Vincent <thomasvincent@gmail.com>

* fix(csrf): fall back to the old default secret path when reading it

csrf-magic tried its own directory as a second candidate regardless of $path_csrf_secret; the new reader only tried the configured path, so a site pointing $path_csrf_secret at a file that was never created lost the legacy-token grace window entirely.

Signed-off-by: Thomas Vincent <thomasvincent@gmail.com>

* docs(csrf): point the JS header at its BSD license sidecar

The header reduces the original BSD terms to prose and the actual text sits in an unreferenced csrf.js.LICENSE, so add a pointer to it.

Signed-off-by: Thomas Vincent <thomasvincent@gmail.com>

* fix(csrf): restore csrf_check()'s bool return and $fatal contract

The shim ignored $fatal and always returned void, so audit/audit.php's `!csrf_check(false)` evaluated `!null` (always true) and denied every request instead of getting a real pass/fail result back.

Signed-off-by: Thomas Vincent <thomasvincent@gmail.com>

* docs: reference the pull request number in the changelog entry

Signed-off-by: Thomas Vincent <thomasvincent@gmail.com>

* test: resolve includes through Cacti path constants

Signed-off-by: Thomas Vincent <thomasvincent@gmail.com>

* test(csrf): enforce full coverage

Signed-off-by: Thomas Vincent <thomasvincent@gmail.com>

* Harden CSRF fallback handling

Signed-off-by: Thomas Vincent <thomasvincent@gmail.com>

* test(csrf): assert the component stays out of the index

Committing it would leave a second copy to drift against the lock.

Signed-off-by: Thomas Vincent <thomasvincent@gmail.com>

* ci: align CSRF checks with PHP 8.2

Signed-off-by: Thomas Vincent <thomasvincent@gmail.com>

* docs: normalize the CSRF changelog entry

* Fix CHANGELOG entry ordering

Signed-off-by: Thomas Vincent <thomasvincent@gmail.com>

* test(csrf): support constrained test environments

* fix(csrf): narrow nullable guard state

* test: place CSRF coverage in security domain

Signed-off-by: Thomas Vincent <thomasvincent@gmail.com>

* security(test): match disabled exec exactly

Signed-off-by: Thomas Vincent <thomasvincent@gmail.com>

* ci: run CSRF coverage on PHP 8.3

Signed-off-by: Thomas Vincent <thomasvincent@gmail.com>

---------

Signed-off-by: Thomas Vincent <thomasvincent@gmail.com>
Co-authored-by: TheWitness <thewitness@cacti.net>
2026-09-18 00:16:39 -04:00

78 lines
3.3 KiB
PHP

<?php
/*
+-------------------------------------------------------------------------+
| Copyright (C) 2004-2026 The Cacti Group |
| |
| This program is free software; you can redistribute it and/or |
| modify it under the terms of the GNU General Public License |
| as published by the Free Software Foundation; either version 2 |
| of the License, or (at your option) any later version. |
+-------------------------------------------------------------------------+
| Cacti: The Complete RRDtool-based Graphing Solution |
+-------------------------------------------------------------------------+
*/
/*
* CactiCsrfGuard resolves Symfony classes while include/csrf.php is being
* evaluated, so the Composer autoloader has to be registered first. This is an
* include-ordering property of a single file; there is no runtime seam to
* assert against, so the source order is the assertion.
*/
$root = dirname(__DIR__, 4);
test('the composer autoloader is required before include/csrf.php', function () use ($root) {
$src = file_get_contents($root . '/include/global.php');
$autoload = strpos($src, 'require_once($vendor_autoload);');
$csrf = strpos($src, "require_once(CACTI_PATH_INCLUDE . '/csrf.php');");
expect($autoload)->not->toBeFalse();
expect($csrf)->not->toBeFalse();
expect($autoload)->toBeLessThan($csrf);
});
/*
* csrf-magic validated the request from the bottom of its own file, so merely
* including it protected every POST. Nothing does that implicitly now: if this
* call goes missing, every page still renders and every form still carries a
* token, but no submission is ever checked.
*/
test('include/global.php invokes the guard after loading it', function () use ($root) {
$src = file_get_contents($root . '/include/global.php');
$include = strpos($src, "require_once(CACTI_PATH_INCLUDE . '/csrf.php');");
$startup = strpos($src, 'csrf_startup();');
expect($startup)->not->toBeFalse()
->and($include)->not->toBeFalse()
->and($include)->toBeLessThan($startup);
});
/*
* include/session.php registers session_write_close() as a shutdown
* function, and shutdown functions run before output buffers flush. A token
* minted lazily from inside the ob_start() callback would therefore write
* $_SESSION after the session already closed on any request where that
* callback runs the first mint of the request. Exercising this end to end
* would require a real shutdown-function pass, which Pest cannot simulate
* mid-test, so this asserts the fix at the source level: csrf_startup()
* must call token() before it registers the output handler.
*/
test('csrf_startup mints a token before registering the output handler', function () use ($root) {
$src = file_get_contents($root . '/include/csrf.php');
$start = strpos($src, 'function csrf_startup(');
expect($start)->not->toBeFalse();
$end = strpos($src, "\nfunction ", $start + 1);
$body = substr($src, $start, ($end === false ? strlen($src) : $end) - $start);
$token = strpos($body, '$guard->token();');
$obStart = strpos($body, 'ob_start(');
expect($token)->not->toBeFalse()
->and($obStart)->not->toBeFalse()
->and($token)->toBeLessThan($obStart);
});