529087 - Crash [@ nsXBLBinding::AllowScripts]

Status: RESOLVED FIXED

(Core :: XBL, defect)

Description

[Mats Palmgren (inactive)]

It's ranked #276 in the 3.5.5 (past 7 days) top crash list:
http://crash-stats.mozilla.com/topcrasher/byversion/Firefox/3.5.5/7
with 389 crashes, 380 on Win, 9 on OSX.

stack:
nsXBLBinding::AllowScripts               content/xbl/src/nsXBLBinding.cpp:1380
nsXBLBinding::ExecuteAttachedHandler     content/xbl/src/nsXBLBinding.cpp:977
nsXBLBinding::ExecuteAttachedHandler     content/xbl/src/nsXBLBinding.cpp:975
nsRunnableMethod<nsFileUploadContentStream>::Run   xpcom/nsThreadUtils.h:278
nsContentUtils::RemoveScriptBlocker   content/base/src/nsContentUtils.cpp:4344
PresShell::DoFlushPendingNotifications	layout/base/nsPresShell.cpp:4842
PresShell::FlushPendingNotifications	layout/base/nsPresShell.cpp:4798
nsDocument::FlushPendingNotifications	content/base/src/nsDocument.cpp:6375
nsQueryReferent::operator	nsWeakReference.cpp:88
nsComputedDOMStyle::GetPropertyCSSValue layout/style/nsComputedDOMStyle.cpp:349
xul.dll@0x8c4807	

There are also 2 crashes on trunk within the past week, so it appears
the bug hasn't been fixed yet.

Comment 1

[Boris Zbarsky [:bzbarsky]]

Line 1380 is:

1380   nsIDocument* doc = mBoundElement->GetOwnerDoc();

right?

Comment 2

[Mats Palmgren (inactive)]

Yes.

This crash might be what bug 307562 was about?

Comment 3

[Boris Zbarsky [:bzbarsky]]

Hmm.  mBoundElement is a weak ref, and is never unset, as far as I can tell.  How's that supposed to work?

Comment 4

[Jonas Sicking (:sicking) No longer reading bugmail consistently]

Taking this. ETA end of next week.

Comment 5

[Jonas Sicking (:sicking) No longer reading bugmail consistently]

Attachment: Patch to fix

I think this should do it.

Usually we're protected by the fact that nodes are always UnbindFromTree'ed from their document before they die. And unbinding removes any xbl bindings.

However nodes with NODE_FORCE_XBL_BINDINGS set can have xbl bindings without being in a document, and thus can die while still having bindings.

The change in this patch to nsNodeUtils takes care of this. This should ensure that mBoundElement never becomes dangling. This should be enough to take care of the security issues here.

However we also need to deal with that we can get a null mBoundElement while running nested constructors. Otherwise we would crash with a nullpointer deref and still have the topcrash reported in comment 0. The change to nsXBLBinding should take care of that.

Comment 6

[Jonas Sicking (:sicking) No longer reading bugmail consistently]

I think I can come up with tests for this. Asking for review in the meantime though.

Comment 7

[Johnny Stenback (:jst)]

Olli, can you review this security bug fix?

Comment 8

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

Yes, I will later this week. Sorry for the delay.

Comment 9

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

Comment on attachment 441671 [details] [diff] [review]
Patch to fix

Ok, makes sense.

Though, if BindingManager() may return null, we should rename it to GetBindingManager().
In theory it may return null, but I think we need to change ownership back to what it used to be (so that we can have strong parent pointer - which is a different bug.).

Comment 10

[Jonas Sicking (:sicking) No longer reading bugmail consistently]

I think it can only return null during cycle collection. But yes, ideally we should improve the situation there.

Comment 11

[Damon Sicore (:damons)]

Can we get this landed?

Comment 12

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

Security bugs need also sr.

Comment 13

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

Comment on attachment 441671 [details] [diff] [review]
Patch to fix

Or if peterv is away for some time, perhaps jst could sr.

Comment 14

[Mats Palmgren (inactive)]

> Security bugs need also sr.

Not anymore. That rule was removed from the policy.
http://www.mozilla.org/hacking/reviewers.html
(the policy change was announced in mozilla.governance)

Comment 15

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

Comment on attachment 441671 [details] [diff] [review]
Patch to fix

Huh, we're changing rules faster than people learn them.

Comment 16

[Damon Sicore (:damons)]

Can we land this?

Comment 17

[Jonas Sicking (:sicking) No longer reading bugmail consistently]

Checked in. Thanks for the review!

http://hg.mozilla.org/mozilla-central/rev/60fda31d4012

Comment 18

[Jonas Sicking (:sicking) No longer reading bugmail consistently]

Hrm.. also checked in bustage fix due to merge.

http://hg.mozilla.org/mozilla-central/rev/adf3e43ffdd9

Comment 19

[Johnny Stenback (:jst)]

This should be marked fixed now, no?

Comment 20

[Jonas Sicking (:sicking) No longer reading bugmail consistently]

Indeed! Somehow missed that in all the flag-setting.

Comment 21

[Daniel Veditz [:dveditz]]

Comment on attachment 441671 [details] [diff] [review]
Patch to fix

Approved for 1.9.2.5 and 1.9.1.11, a=dveditz for release-drivers

Comment 23

[Jonas Sicking (:sicking) No longer reading bugmail consistently]

Fixed on 1.9.2 and 1.9.1

http://hg.mozilla.org/releases/mozilla-1.9.1/rev/446e17503f87
http://hg.mozilla.org/releases/mozilla-1.9.2/rev/15dccae78d7e

Comment 24

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

Are there STR that can be used for verification of this fix on 1.9.1 and 1.9.2?

Comment 25

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

Jonas? Mats? Any reply?