Bug 68921 – CSP doesn't work for a wide variety of cases

• Component: WebCore Misc.
• Status: RESOLVED FIXED

[Boris Smus]

In my own tests, as well as in more comprehensive test suites, CSP policy is not being followed as expected. 

This test suite: http://erlend.oftedal.no/blog/csp/readiness/latest.php provides both X-Content-Security-Policy and X-WebKit-CSP headers, and uses the new syntax ('default-src' rather than 'allow'), but fails for most test cases.

Do we have a CSP unit test suite?

[Adam Barth]

> Do we have a CSP unit test suite?

Our CSP tests are here:

http://trac.webkit.org/browser/trunk/LayoutTests/http/tests/security/contentSecurityPolicy

I'm not sure why we're doing so badly on the oftedal test suite.  I'll investigate.

[Sam Weinig]

This looks like an issue with our port matching code in CSPSource.  In the case that the source's port is set to 0 (for no port specified) and the URL to match has no port specified (also set to 0), the code checks if 0 is the default port for the protocol. 

        int port = url.port();
        return port ? port == m_port : isDefaultPortForProtocol(m_port, url.protocol());

In this case, we should be returning true, something like this should probably work:


    bool portMatches(const KURL& url) const
    {
        if (m_portHasWildcard)
            return true;
        
        int port = url.port();
        if (port == m_port)
            return true;
    
        if (!port)
            return isDefaultPortForProtocol(m_port, m_scheme);

        if (!m_port)
            return isDefaultPortForProtocol(port, m_scheme);

        return false;
    }

[Sam Weinig]

Fixing that issue, it looks like we pass all the same tests that Firefox does.

[Sam Weinig]

We probably have no tests for this case, since our testing harness requires using a non-default port :(.

[Adam Barth]

Ah, that makes a lot of sense.  Thanks Sam.  Are you planning to post a patch, or should I?  (Also, any thoughts on how we can test this bug?)

[Adam Barth]

Created attachment 109035 [details]
Patch

[WebKit Review Bot]

Comment on attachment 109035 [details]
Patch

Clearing flags on attachment: 109035

Committed r96239: <http://trac.webkit.org/changeset/96239>

[WebKit Review Bot]

All reviewed patches have been landed.  Closing bug.

[Adam Barth]

*** Bug 67008 has been marked as a duplicate of this bug. ***