Closed Bug 157989 Opened 18 years ago Closed 18 years ago

Possible heap corruption with 0-width GIF


(Core :: ImageLib, defect, P1, major)






(Reporter: security-bugs, Assigned: pavlov)



(Whiteboard: [adt1 RTM] [ETA 07/24])


(2 files)

Submitter name:                zen-parse
   Submitter email address:
   Acknowledgement checkbox:      on
   Product:                       Netscape 6.x
   Operating system:              Unix: x86 Linux
   OS version:                    redhat 7.0
   Issue summary:                 a "0 width" GIF image: exploitable heap overwrite.

Issue details:
A specially formed gif image with width 0 can be 
used to overwrite 'chunk' information and other 
heap based information.

By overwriting a function pointer with an 
address containing user supplied data, it is 
possible to execute arbitary code.

down to line 1310
and then line 1351 and 1354

This form was submitted from
with Opera/6.0 (Linux 2.2.19-7.0.16 i686; U)  [en].
We need to know if this is really exploitable.
Whiteboard: [adt1 RTM]
Paper?  Can you take a look at this?
Removing adt1.0.1 until there is a patch with reviews attached.
Keywords: adt1.0.1, nsbeta1nsbeta1+
Whiteboard: [adt1 RTM] → [adt1 RTM] [ETA Needed]
Firstly, gifcom is dead :)  It's been moved to /modules/libpr0n/decoders/gif/
which still contains a lot of the old gifcom code.

I talked it over with tor, and stopping the 0 width before the memory
allocation seemed to be the best solution.  And this is the patch.

The memory overwrite actually happens later on during do_lzw, which first
writes one byte, incrememnts the pointer, then checks whether it hit the end of
the buffer.  I looked into fixing the problem there, but it just caused
problems in other areas (namely in BeginGIF(), which will never initialize
mImageContainer if the width or height is 0.. which trickles down to assertion

I've also plugged the memory overwrite hole in do_lzw.	In theory, the width
check part of the patch should stop do_lzw from ever being called when the
buffer is 0 in length, but it's better to check anyway.
This doesn't do anything malicious, however, it does crash Mozilla.  I took a
14x14 GIF with one frame, and edited the image header and frame header so that
it says 0x0.
Blocks: 143047
Severity: normal → major
Priority: -- → P1
Target Milestone: --- → mozilla1.0.1
seems like we have a patch from Arron M. 

dan/mitch: can you review and super-review the patch, so we can try and get it
landed asap. thanks!
Whiteboard: [adt1 RTM] [ETA Needed] → [adt1 RTM] [ETA 07/22]
Comment on attachment 91974 [details] [diff] [review]
stop memory overwrite

Attachment #91974 - Flags: review+
Comment on attachment 91974 [details] [diff] [review]
stop memory overwrite

Attachment #91974 - Flags: superreview+
looks like we have the reviews.

pav: are you in a position to land this patch for aaron m?
Keywords: adt1.0.1
Blocks: 1.0.1
who's in a position to apply this patch to the trunk, so we can get some bake
time on this change?
Comment on attachment 91974 [details] [diff] [review]
stop memory overwrite

Mixed indentation (2 vs. 4 space), but that seems to pervade the file already. for 1.1beta trunk checkin.

Attachment #91974 - Flags: approval+
Whiteboard: [adt1 RTM] [ETA 07/22] → [adt1 RTM] [ETA 07/24]
Adding adt1.0.1+ on behalf of the adt.  Please get Drivers approval for the
branch and check the fix into the branch.
Keywords: adt1.0.1adt1.0.1+
a=chofmann for 1.0.1. add  the fixed1.0.1 keyword after getting this on the branch.
landed on trunk and branch.
Closed: 18 years ago
Keywords: fixed1.0.1
Resolution: --- → FIXED
thanks pav.

tpreston: teri, can you verify this fix on the branch, then replace "fixed1.0.1"
with "verified1.0.1"? thanks!
Verified fix checked into and
Group: security?
You need to log in before you can comment on or make changes to this bug.