528300 - Crash [@ txExecutionState::~txExecutionState]

Status: RESOLVED FIXED

(Core :: XSLT, defect)

Description

[Paul Nickerson]

Attachment: testcase (don't load from bugzilla)

Load the testcase and refresh if it doesn't crash right away. EIP is sometimes different, especially when extra fluff is added to the transform and to the original document, so IMO this is exploitable.

Comment 1

[Paul Nickerson]

Attachment: stack

Comment 2

[Paul Nickerson]

WFM unless I save the file locally then open it.

Comment 3

[Peter Van der Beken [:peterv]]

Attachment: v1

Still need to add the crashtest.

Comment 4

[Peter Van der Beken [:peterv]]

Attachment: v1

Comment 6

[Peter Van der Beken [:peterv]]

I'll also add the crashtest from bug 528488.

Comment 7

[Peter Van der Beken [:peterv]]

Attachment: v1.1

Comment 8

[Paul Nickerson]

Comment on attachment 412490 [details] [diff] [review]
v1.1

I don't crash on this testcase or the one in bug 528488 with the patch applied.

Comment 9

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

Comment on attachment 412490 [details] [diff] [review]
v1.1

So why exactly are we crashing? Is the global-variable code ending up in a bad state or something?

While this patch seems like the right thing to do, I'm wondering if we could/should secure more code.

r=me anyhow

Comment 10

[Peter Van der Beken [:peterv]]

We bail out early because of an error when trying to get the variable in txExecutionState::getVariable. We don't pop the result handler from the stack. We have code to deal with that, not deleting the unknown handler if an error happens (since the result handler stack will delete it). But evaluateToNumber eats the error return and so we do delete the unknown handler, and then the stacks tries to do that again.

Comment 11

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

Ah, and we can't pop these as we go up the callstack because it was push by a txInstruction and so we can't pop these guys off as we unwind.

Would be nice to get this more stable, but I can't think of a way to do so for now.

Comment 12

[Peter Van der Beken [:peterv]]

Need this crash fix on branch (not yet landed on trunk, need blocking or approval first).

Comment 13

[Peter Van der Beken [:peterv]]

Attachment: v1.2

Turns out the testcase for bug 528488 exposed a small leak in txExecutionState::getVariable. This is what I landed.

Comment 14

[Peter Van der Beken [:peterv]]

http://hg.mozilla.org/mozilla-central/rev/a727327f9777
http://hg.mozilla.org/releases/mozilla-1.9.2/rev/74188d9e9faf

I'll nominate for other branches too.

Comment 15

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

Comment on attachment 413940 [details] [diff] [review]
v1.2

Would be nice if you could diff with -u8

r=me

Comment 16

[Peter Van der Beken [:peterv]]

Comment on attachment 413940 [details] [diff] [review]
v1.2

XSLT crash fix.

Comment 17

[Daniel Veditz [:dveditz]]

Comment on attachment 413940 [details] [diff] [review]
v1.2

Approved for 1.9.1.7 and 1.9.0.17, a=dveditz for release-drivers

Comment 18

[Peter Van der Beken [:peterv]]

http://hg.mozilla.org/releases/mozilla-1.9.1/rev/7bb736fa5f7a

Comment 19

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

Verified fixed in 1.9.1 with Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.9.1.8pre) Gecko/20100115 Shiretoko/3.5.8pre (.NET CLR 3.5.30729). Attached testcase crashes in 1.9.1.7 but just shows "Error during XSLT transformation: XSLT Stylesheet (possibly) contains a recursion." in the 1.9.1 current nightly.

Comment 20

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

Attachment: 1.8.0 version

Affects 1.8.0

Comment 21

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

Verified for 1.9.0.18 (Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.9.0.16pre) Gecko/2010020215 GranParadiso/3.0.18pre (.NET CLR 3.5.30729) with attached testcase. Results are the same as the 1.9.1.8 verification with a crash in 1.9.0.17 but an XSLT error in 1.9.0.18.