RESOLVED FIXED 104480
CSP: XHR from an isolated world should bypass a page's policy. 
https://bugs.webkit.org/show_bug.cgi?id=104480
Summary CSP: XHR from an isolated world should bypass a page's policy.
Mike West
Reported 2012-12-09 04:33:04 PST
1. Yay! Facebook is using CSP! 2. Oh noes! Facebook extensions are breaking! See: http://stackoverflow.com/questions/13786918/refused-to-connect-to-url-because-it-violates-the-following-content-security-p for some detail. In a nutshell, we should do the same check for `connect-src` that we're currently doing for CachedResourceLoader.
Attachments
Patch (13.32 KB, patch)
2012-12-10 03:03 PST, Mike West 		no flags
DetailsFormatted DiffDiff
Patch (17.96 KB, patch)
2013-01-04 00:19 PST, Mike West 		no flags
DetailsFormatted DiffDiff
Show Obsolete (1) View All Add attachment proposed patch, testcase, etc.
Mike West
Comment 1 2012-12-10 03:03:17 PST
Created attachment 178501 [details] Patch
Mike West
Comment 2 2012-12-10 03:06:25 PST
Hey Eric! Are you a good person to review this patch in Adam's stead?
Eric Seidel (no email)
Comment 3 2012-12-10 11:45:11 PST
Sadly, no. This really should wait for Adam's return. Just add a note to your calendar to email him Jan 3rd. :)
Eric Seidel (no email)
Comment 4 2012-12-10 11:46:09 PST
If we absolutely need a fix before then, I can review it, but my understanding of the CSP security model is not nearly as deep as Adam's.
Mike West
Comment 5 2012-12-10 12:28:22 PST
(In reply to comment #4) > If we absolutely need a fix before then, I can review it, but my understanding of the CSP security model is not nearly as deep as Adam's. Hrm. We don't absolutely need anything. :) This just jumped onto my radar now that Facebook launched CSP support and we're getting more bug reports. This patch doesn't do anything really new, it simply applies the check we added to CachedResourceLoader to the areas covered by `connect-src` (WebSockers, XHR, and EventSource). I was hoping it'd be an easy patch to review, but I agree with you completely that we don't want to make mistakes in this area. *shrug* I'll get in line for Adam's return. He really is the right guy to review.
Build Bot
Comment 6 2012-12-11 13:01:45 PST
Comment on attachment 178501 [details] Patch Attachment 178501 [details] did not pass mac-ews (mac): Output: http://queues.webkit.org/results/15257797 New failing tests: http/tests/security/isolatedWorld/bypass-main-world-csp-for-xhr.html
Avi K
Comment 7 2012-12-13 03:21:27 PST
Thanks guys for handling this, this is really a major issue for many extensions
Oliver Schlöbe
Comment 8 2012-12-13 05:35:08 PST
Thanks also! I am the topic starter of the topic Mike West referenced in his initial post, and am grateful he went and created this bug report a few minutes later. :) Thanks guys!
Mike West
Comment 9 2013-01-04 00:19:30 PST
Created attachment 181279 [details] Patch
Adam Barth
Comment 10 2013-01-04 09:53:57 PST
Comment on attachment 181279 [details] Patch Great!
WebKit Review Bot
Comment 11 2013-01-04 11:10:31 PST
Comment on attachment 181279 [details] Patch Clearing flags on attachment: 181279 Committed r138817: <http://trac.webkit.org/changeset/138817>
WebKit Review Bot
Comment 12 2013-01-04 11:10:37 PST
All reviewed patches have been landed. Closing bug.
Note You need to log in before you can comment on or make changes to this bug.