678362 - Assertion failure: length < (uint32(1) << 23), at jshashtable.h:368

Status: RESOLVED FIXED

(Core :: JavaScript Engine, defect)

Description

[Christian Holler (:decoder)]

The following code asserts on mozilla-central (revision 4b4b359e77e4, no options required):

var replacer = [0, 1, 2, 3];
Object.defineProperty(replacer, 3.e7, {});
JSON.stringify({ 0: 0, 1: 1, 2: 2, 3: 3 }, replacer)


Using 3.e7 this passes after the assert, using 3.e8 fails with InternalError due to allocation failure. I haven't seen any crash or other suspicious behavior but according to Luke it's directly not clear if this is a security problem or not. Remove S-s when this is confirmed to be non-harmful.

Comment 1

[Christian Holler (:decoder)]

Sorry, the revision was for mozilla-inbound, mozilla-central is also affected though.

Comment 2

[Luke Wagner [:luke]]

Attachment: fix

So, this (1<<23) bound is real since it guards the subsequent multiply.  On overflow, this will allocate less capacity then requested.  That would be a security problem if json code expected that, if init() succeeded, that it could insert infallibly, but json doesn't.

With a bit of rejiggering, init() can still do only a single test that covers all overflow.

Comment 3

[Luke Wagner [:luke]]

Attachment: fix v.2

This patch adds static asserts to explain the relaxing of >= to >.

Comment 4

[Jeff Walden [:Waldo]]

Comment on attachment 552565 [details] [diff] [review]
fix v.2

Review of attachment 552565 [details] [diff] [review]:
-----------------------------------------------------------------

::: js/src/tests/ecma_5/JSON/stringify-replacer.js
@@ +161,5 @@
> +}
> +catch (e)
> +{
> +  assertEq(""+e, "InternalError: allocation size overflow");
> +}

1. New tests should go in new files, for ease of debugging when we break them during development (or after, if it happens that way).

2. InternalError grunge is totally not standard at all, so these two tests should be in an extensions/ directory.

Comment 5

[Luke Wagner [:luke]]

Done and done.
http://hg.mozilla.org/integration/mozilla-inbound/rev/796b8466d549

Comment 6

[Marco Bonardo [:mak] (Away Apr 25 - May 5)]

http://hg.mozilla.org/mozilla-central/rev/796b8466d549

Comment 7

[Christian Holler (:decoder)]

Automatically extracted testcase for this bug was committed:

https://hg.mozilla.org/mozilla-central/rev/efaf8960a929