505988 - (CVE-2009-3374) It's possible to access a double-wrapped object's underlying privileged object

Status: RESOLVED FIXED

(Core :: XPConnect, defect, P1)

Description

[moz_bug_r_a4]

XPCVariant::VariantDataToJS() unwraps double-wrapped objects even when it
shouldn't.

Comment 1

[moz_bug_r_a4]

Attachment: testcase - Arbitrary code execution

Comment 2

[Samuel Sidler (old account; do not CC)]

Do we need this on 1.9.0? This definitely block 1.9.1.x but not sure if it's for .2 or .3.

Comment 3

[Daniel Veditz [:dveditz]]

Yes, this affects Firefox 3.0.x

Comment 4

[Blake Kaplan (:mrbkap) (inactive)]

Attachment: Work in progress

This patch fixes this bug, but I'm seeing spurious errors in the livemark service with it applied.

Comment 5

[Blake Kaplan (:mrbkap) (inactive)]

Attachment: Proposed fix

The spurious errors were due to my incorrectly backing out bug 384632.

Comment 6

[Daniel Veditz [:dveditz]]

The blocking1.9.0.13+ is a soft block, we'll take it on the 1.9.0 branch when it gets fixed on the 1.9.1 branch.

Comment 7

[Boris Zbarsky [:bzbarsky]]

Comment on attachment 390563 [details] [diff] [review]
Proposed fix

Seems ok.

Comment 8

[Robert O'Callahan (:roc) (email my personal email if necessary)]

http://hg.mozilla.org/mozilla-central/rev/03c40c5a2d4b

Comment 9

[Boris Zbarsky [:bzbarsky]]

This caused a bunch of password manager test failures along these lines:

/test/test_basic_form_observer_autofillForms.html | Checking foundLogins is array - got "Object", expected "Array"
82398 ERROR TEST-UNEXPECTED-FAIL | /tests/toolkit/components/passwordmgr
82407 ERROR TEST-UNEXPECTED-FAIL | /tests/toolkit/components/passwordmgr/test/test_basic_form_observer_foundLogins.html | Checking foundLogins is array - got "Object", expected "Array"

and various other stuff, but I suspect most of it is due to the tests that got the errors above failing and due to the fact that the tests in this dir all depend on each other(!).  Which means you have to run with

   --test-path=toolkit/components/passwordmgr/test/

to test reasonably; just running a single test won't work.  And if you run a single test and then the whole dir that will also fail...

It's not clear to me whether this is a problem in the test or in the patch yet, but for now backed out to get the tree green: http://hg.mozilla.org/mozilla-central/rev/c1ab8650e0ce

Comment 10

[Mike Beltzner [:beltzner, not reading bugmail]]

This is a P1 1.9.2 blocker, thus blocking Firefox 3.6a1 - any ETA on re-landing or analysing why this caused so many test failures?

Comment 11

[Mike Beltzner [:beltzner, not reading bugmail]]

Justin: can you take a second to see if you can figure out why the PW Mgr tests failed?

Comment 12

[Justin Dolske [:Dolske]]

So, this was a  test added for bug 492153. That bug added notifications for when pwmgr does things when processing a form. The notification's subject is a nsIPropertyBag, with various properties set. One of these is a JS array of relevant logins.

The notification is built in nsLoginManager.js's _notifyFoundLogins() as

  propbag.setPropertyAsInterface("foundLogins", foundLogins.concat());

And the test is checking

  propbag.get("foundLogins").constructor.name, "Array", "Checking foundLogins is array");

I'd guess that either something is going wrong in setting the property (I think there's some XPCOM magic being used to shove an Array into an nsISupports param), or that checking .constructor.name isn't a reliable thing to do. [ISTR Firebug having run into similar problems.]

mrbkap, does this help?

Comment 13

[Mike Beltzner [:beltzner, not reading bugmail]]

I need to know if this needs to land for alpha 1 in order to make 1.9.2; it's literally holding our alpha production and branch at this point.

Please respond ASAP.

Comment 14

[Blake Kaplan (:mrbkap) (inactive)]

(In reply to comment #13)
> I need to know if this needs to land for alpha 1 in order to make 1.9.2; it's
> literally holding our alpha production and branch at this point.
> 
> Please respond ASAP.

Sorry, I should have commented here... The reason for the MochiTest failures are now blocking this bug. Once those bugs land, this one should be able to as well.

Comment 15

[Daniel Veditz [:dveditz]]

This probably needs to wait for the next round of releases rather than trying to rush this into 1.9.1.3/1.9.0.14

Comment 16

[Blake Kaplan (:mrbkap) (inactive)]

http://hg.mozilla.org/mozilla-central/rev/5c913c4662d8

Comment 17

[Mike Beltzner [:beltzner, not reading bugmail]]

Mass change: adding fixed1.9.2 keyword

(This bug was identified as a mozilla1.9.2 blocker which was fixed before the mozilla-1.9.2 repository was branched (August 13th, 2009) as per this query: http://is.gd/2ydcb - if this bug is not actually fixed on mozilla1.9.2, please remove the keyword. Apologies for the bugspam)

Comment 18

[Daniel Veditz [:dveditz]]

This is blocking 1.9.1.4 and 1.9.0.15 (code freeze Sept 22!). We probably need a branch patch that includes any regression fixes, or we need branch approval requests on those bugs too.

Comment 19

[Daniel Veditz [:dveditz]]

Comment on attachment 390563 [details] [diff] [review]
Proposed fix

Approved for 1.9.1.4 and 1.9.0.15, a=dveditz

Comment 20

[Blake Kaplan (:mrbkap) (inactive)]

http://hg.mozilla.org/releases/mozilla-1.9.1/rev/ed60afa878e9

Checking in js/src/xpconnect/src/xpcprivate.h;
/cvsroot/mozilla/js/src/xpconnect/src/xpcprivate.h,v  <--  xpcprivate.h
new revision: 1.286; previous revision: 1.285
done
Checking in js/src/xpconnect/src/xpcvariant.cpp;
/cvsroot/mozilla/js/src/xpconnect/src/xpcvariant.cpp,v  <--  xpcvariant.cpp
new revision: 1.32; previous revision: 1.31
done
Checking in js/src/xpconnect/tests/mochitest/Makefile.in;
/cvsroot/mozilla/js/src/xpconnect/tests/mochitest/Makefile.in,v  <--  Makefile.in
new revision: 1.12; previous revision: 1.11
done
RCS file: /cvsroot/mozilla/js/src/xpconnect/tests/mochitest/test_bug384632.html,v
done
Checking in js/src/xpconnect/tests/mochitest/test_bug384632.html;
/cvsroot/mozilla/js/src/xpconnect/tests/mochitest/test_bug384632.html,v  <--  test_bug384632.html
initial revision: 1.1
done

Comment 21

[Al Billings [:abillings - ex-MoCo]]

Verified fixed for 1.9.0.15 with Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.9.0.15pre) Gecko/2009092805 GranParadiso/3.0.15pre (.NET CLR 3.5.30729). 

Verified fixed for 1.9.1 with Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.9.1.4pre) Gecko/20090924 Shiretoko/3.5.4pre (.NET CLR 3.5.30729).

Comment 22

[Daniel Veditz [:dveditz]]

This testcase doesn't run on the 1.8 branch because I get NS_ERROR_NOT_IMPLEMENTED from the document.setUserData() call. Is there some way to get the double unwrapping to happen on that branch? Or should we just not bother because that branch is missing so many wrappers anyway?

Comment 23

[moz_bug_r_a4]

It seems that this bug is a regression from bug 384632 and that bug is not
fixed on the 1.8 branch and thus the 1.8 branch is not affected by this bug.

Comment 24

[moz_bug_r_a4]

Attachment: testcase 2 - Arbitrary code execution

This testcase uses XSLTProcessor.setParameter().  This testcase works on
fx-3.5.3/fx-3.0.14 and is fixed on fx-3.5.4/fx-3.0.15.  On SeaMonkey-1.1.18, 
the same behavior as fx-3.5.4/fx-3.0.15.