317934 - CVE-2006-0294 [FIX]Crash with evil testcase, switching from position:relative to static

Status: VERIFIED FIXED

(Core :: Layout, defect, P1)

Description

[Martijn Wargers (dead)]

See upcoming testcase, when clicking on the button, current Mozilla builds crash.
This doesn't happen in Mozilla1.7.

Talkback ID: TB12309372W
0x00000000
nsIView::Destroy  [c:/builds/tinderbox/Fx-Trunk/WINNT_5.2_Depend/mozilla/view/src/nsView.cpp, line 307]
nsSplittableFrame::Destroy  [c:/builds/tinderbox/Fx-Trunk/WINNT_5.2_Depend/mozilla/layout/generic/nsSplittableFrame.cpp, line 71]
nsPositionedInlineFrame::Destroy  [c:/builds/tinderbox/Fx-Trunk/WINNT_5.2_Depend/mozilla/layout/generic/nsInlineFrame.cpp, line 1052]
nsFrameList::DestroyFrames  [c:/builds/tinderbox/Fx-Trunk/WINNT_5.2_Depend/mozilla/layout/generic/nsFrameList.cpp, line 138]
nsPositionedInlineFrame::Destroy  [c:/builds/tinderbox/Fx-Trunk/WINNT_5.2_Depend/mozilla/layout/generic/nsInlineFrame.cpp, line 1052]
nsBlockFrame::RemoveFrame  [c:/builds/tinderbox/Fx-Trunk/WINNT_5.2_Depend/mozilla/layout/generic/nsBlockFrame.cpp, line 5516]
nsFrameManager::RemoveFrame  [c:/builds/tinderbox/Fx-Trunk/WINNT_5.2_Depend/mozilla/layout/base/nsFrameManager.cpp, line 704]
nsCSSFrameConstructor::ContentRemoved  [c:/builds/tinderbox/Fx-Trunk/WINNT_5.2_Depend/mozilla/layout/base/nsCSSFrameConstructor.cpp, line 9946]
nsCSSFrameConstructor::RecreateFramesForContent  [c:/builds/tinderbox/Fx-Trunk/WINNT_5.2_Depend/mozilla/layout/base/nsCSSFrameConstructor.cpp, line 11495]
nsCSSFrameConstructor::RestyleElement  [c:/builds/tinderbox/Fx-Trunk/WINNT_5.2_Depend/mozilla/layout/base/nsCSSFrameConstructor.cpp, line 10377]
nsCSSFrameConstructor::ProcessOneRestyle  [c:/builds/tinderbox/Fx-Trunk/WINNT_5.2_Depend/mozilla/layout/base/nsCSSFrameConstructor.cpp, line 13211]
nsCSSFrameConstructor::ProcessPendingRestyles  [c:/builds/tinderbox/Fx-Trunk/WINNT_5.2_Depend/mozilla/layout/base/nsCSSFrameConstructor.cpp, line 13259]
nsCSSFrameConstructor::RestyleEvent::HandleEvent  [c:/builds/tinderbox/Fx-Trunk/WINNT_5.2_Depend/mozilla/layout/base/nsCSSFrameConstructor.cpp, line 13324]
SHELL32.dll + 0x520c24 (0x778b0c24)

Comment 1

[Martijn Wargers (dead)]

Attachment: testcase

Comment 2

[Martijn Wargers (dead)]

Well, this seems to have regressed between 2005-06-20 and 2005-06-22:
http://bonsai.mozilla.org/cvsquery.cgi?treeid=default&module=all&branch=HEAD&branchtype=match&dir=&file=&filetype=match&who=&whotype=match&sortby=Date&hours=2&date=explicit&mindate=2005-06-20+05%3A00%3A00&maxdate=2005-06-22+12%3A00%3A00&cvsroot=%2Fcvsroot

I don't have really an idea which bug could have caused the regression.

Comment 3

[Felix Miata]

testcase crashes 1.5rc3 on OS/2

Comment 4

[Andrew Schultz]

Attachment: valgrind log

I crashed with builds back to 2005-06-10 but then couldn't crash with any build.  valgrind caught it trying to use already-freed memory.

Comment 5

[Boris Zbarsky [:bzbarsky]]

I see this only when we process the restyle for the inner <q> first.  Then when we're processing the restyle for the outer <q> we get to destroying the brand-new frame the inner restyle created for the positioned <span> and crash when trying to delete its view...

Comment 6

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

The inner frame shouldn't have a view anymore ... did we somehow fail to remove the view from its parent when we reframed the inner element?

Comment 7

[Boris Zbarsky [:bzbarsky]]

Attachment: This should fix it...

This looks like a regression from bug 292116 -- changing the parent means we get a mismatch between the frame and view trees....

Comment 8

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

Comment on attachment 204322 [details] [diff] [review]
This should fix it...

don't you need to initialize appendAfterFrame to nsnull to avoid a 'possible use of uninitialized' warning?

Comment 9

[Boris Zbarsky [:bzbarsky]]

I didn't get those warnings when I compiled, so I assumed that gcc is finally sorta-smart enough...

I'll init them if warnings appear when I check in.

Comment 10

[Boris Zbarsky [:bzbarsky]]

Fixed.

Comment 11

[Boris Zbarsky [:bzbarsky]]

Comment on attachment 204322 [details] [diff] [review]
This should fix it...

This is a double-delete, hence potentially exploitable... I think we should take this on the 1.8 branch, after it bakes on the trunk a tad.

Comment 12

[Stephen Donner [:stephend] Not actively reading bugmail]

I can't speak to any *potential* regressions, but the crash itself is certainly fixed in trunk build 2005-11-29-09 of SeaMonkey trunk on Windows XP.

Verified (trunk)

Comment 13

[Mike Schroepfer]

Comment on attachment 204322 [details] [diff] [review]
This should fix it...

Please land on 1.8 and 1.8.1 branches.

Comment 14

[Boris Zbarsky [:bzbarsky]]

Could someone please land this for me?  If not, I'll do it in early January...

Comment 15

[Boris Zbarsky [:bzbarsky]]

*** Committing to MOZILLA_1_8_BRANCH... 
nsCSSFrameConstructor.h
new revision: 1.187.6.4; previous revision: 1.187.6.3
nsCSSFrameConstructor.cpp
new revision: 1.1110.6.13; previous revision: 1.1110.6.12

*** Committing layout/base/nsCSSFrameConstructor.cpp on MOZILLA_1_8_0_BRANCH... 
new revision: 1.1110.6.12.2.1; previous revision: 1.1110.6.12
*** Committing layout/base/nsCSSFrameConstructor.h on MOZILLA_1_8_0_BRANCH... 
new revision: 1.187.6.3.2.1; previous revision: 1.187.6.3

Comment 16

[Jay Patel [:jay]]

v.fixed on 1.8.0.1 with Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.8.0.1) Gecko/20060111 Firefox/1.5.0.1, no crash with testcase in comment #1

Comment 17

[Bob Clary [:bc] (inactive)]

v test case does not crash 1.8.0.1, 1.8.1, 1.9a1 windows/linux

Comment 18

[Bob Clary [:bc] (inactive)]

crash test landed
http://hg.mozilla.org/mozilla-central/rev/876c3596f749