329900 - Crash with evil testcase, using table-column-group, table-column, table-cell

Status: VERIFIED FIXED

(Core :: Layout: Tables, defect)

Description

[Martijn Wargers (dead)]

See upcoming testcase, which crashes Mozilla on load.

Talkback ID: TB16137409Q
HaveAutoWidth

(Talbkack has become really short, lately)

The testcase also crashes Mozilla1.7.12, so no (recent) regression.

Comment 1

[Martijn Wargers (dead)]

Attachment: testcase

Comment 2

[Bernd]

Arrgh, the stack is bogus even in a debug build. This is certainly exploitable, at least writing to 0x0000002 seems like exploitable to me. We  have a major error in the colgroup pseudo handling which yields a strange frame hierarchy cell - inner cell - cell. I need some time to come up with a fix. More than this I would like to move this bug after the OrderRowGroup removal which leads to the deleted object derefs.

Martijn

congratulations to your CVS privilege :-)
, so its fair to assume that you have a build. If you find these bugs so quick you need to learn to classify them. The bogus talkback should have warned you.

Comment 3

[Martijn Wargers (dead)]

(In reply to comment #2)
> congratulations to your CVS privilege :-)
Thanks :)

> , so its fair to assume that you have a build. If you find these bugs so quick
> you need to learn to classify them. The bogus talkback should have warned you.
Well, I asked Boris some time ago, and he explicitly said not to file bugs as security sensitive (something with way too much security bugs already).
I don't really know when a crash is exploitable. (normally, I don't fire up my debug build, maybe I should do that from now on with new crashes)

Comment 4

[Bernd]

No, I am not asking to file every bug as security sensitive. But this bug causes already random function calls, it's not the typical already  deleted stuff that I am used to. Launching a crash testcase in a debug build sounds like a good idea anyway, the stacktrace is more reliable than the talkback.

Comment 5

[Bernd]

Attachment: patch

The only question that this patch causes is why did I not close the door in bug 239294. Fear Uncertainty Doubt? Or did I not know Martijn at this time? Or did I try a minimal invasive patch?

Comment 6

[Bernd]

Attachment: next rev.

this is the next rev. still not *the* fix but closer to the issue.

Comment 7

[Bernd]

Attachment: next rev

Comment 8

[Boris Zbarsky [:bzbarsky]]

Comment on attachment 215297 [details] [diff] [review]
next rev

I _think_ I follow this.  :(  My eyes bleed, though.  :(

Comment 9

[Bernd]

fixed on trunk

Comment 10

[Martijn Wargers (dead)]

Verified fixed using 2006-03-28 trunk build.

Comment 11

[Martijn Wargers (dead)]

*** Bug 330015 has been marked as a duplicate of this bug. ***

Comment 12

[Bernd]

this regressed the testcase in bug 325984

Comment 13

[Bernd]

I will try to fix the issue in the open bug 325984

Comment 14

[Daniel Veditz [:dveditz]]

Is this exploitable? On a 1.5.0.4 windows debug build I crash in the nsMargin constructor reading from (0+offset) because mReflowState is null when nsBlockReflowState::BorderPadding() is called (the offset is the location of mComputedBorderPadding). Maybe some of these are uninitialized objects whose values would be non-null in a non-debug build, but in a debug Firefox 1.5.0.4 build this is pretty consistent.

If you think it's exploitable we need a patch for the 1.8 and 1.8.0 branches. If it's not exploitable we can live with this low frequency crash (and should un-hide the bug).

Comment 15

[Daniel Veditz [:dveditz]]

Never mind: this fixes bug 330015 which is definitely operating on a deleted frame, so we do need it on the old branches.

Comment 16

[Jay Patel [:jay]]

Comment on attachment 215297 [details] [diff] [review]
next rev

a=jay for 1.8.0 checkin

Comment 17

[Jay Patel [:jay]]

Comment on attachment 215297 [details] [diff] [review]
next rev

1.8.1 branch approval, a=jay for drivers

Comment 18

[Daniel Veditz [:dveditz]]

Comment on attachment 215297 [details] [diff] [review]
next rev

removing branch approvals from this patch in favor of the regression-fix cummulative patch in bug 333493

Comment 19

[Bernd]

fixed on 1.8 branhc by the cumulative patch in bug 333493

Comment 20

[Bernd]

fixed on 1.8.0.5

Comment 21

[Jay Patel [:jay]]

v.fixed on 1.8.0 branch with Mozilla/5.0 (Macintosh; U; Intel Mac OS X; en-US; rv:1.8.0.5) Gecko/20060626 Firefox/1.5.0.5, no crash with testcase.

Comment 22

[Alexander Sack]

Attachment: 1.0.x patch

Comment 23

[Bob Clary [:bc] (inactive)]

https://bugzilla.mozilla.org/attachment.cgi?id=214548&action=view
ff2b2 winxp, linux no crash
verified fixed 1.8

Comment 24

[Bob Clary [:bc] (inactive)]

crash test landed
http://hg.mozilla.org/mozilla-central/rev/fcb4bf8da2b8