935834 - CID 1123240: Resource leak on OOM in CodeGenerator.cpp as found by Coverity

Status: RESOLVED FIXED

(Core :: JavaScript Engine: JIT, defect)

Description

[Gary Kwong [:gkw] [:nth10sd] (NOT official MoCo now)]

Coverity analysis of Oct 31, 2013 source code has found a "resource leak" that may cause OOM.

http://hg.mozilla.org/mozilla-central/annotate/70de5e24d79b/js/src/jit/CodeGenerator.cpp#l2698 - count is not being freed here in CodeGenerator::maybeCreateScriptCounts?

bhackett, is this likely caused by bug 811349?

Comment 1

[Gary Kwong [:gkw] [:nth10sd] (NOT official MoCo now)]

(or might be a leak on OOM)

Comment 2

[Gary Kwong [:gkw] [:nth10sd] (NOT official MoCo now)]

This is CID 1123240.

Comment 3

[Dan Gohman [:sunfish]]

Attachment: cg-offsets-and-leak.patch

Attached is a patch which fixes the leak, and which fixes an oddity in the vicinity.

The "offset = script->pcToOffset(resume->pc());" restores a line of code which appears to have been accidentally made ineffective in change bc935b861f48, and accidentally buried in change 92a5a4a9b76b.

The rest of the patch is fixing the leak.

Comment 4

[Brian Hackett [Laid off!]]

Comment on attachment 8344631 [details] [diff] [review]
cg-offsets-and-leak.patch

Review of attachment 8344631 [details] [diff] [review]:
-----------------------------------------------------------------

Thanks!

Comment 5

[Brian Hackett [Laid off!]]

Comment on attachment 8344631 [details] [diff] [review]
cg-offsets-and-leak.patch

Review of attachment 8344631 [details] [diff] [review]:
-----------------------------------------------------------------

::: js/src/jit/CodeGenerator.cpp
@@ +2779,5 @@
>              // to associate information about the inlining with.
>              MResumePoint *resume = block->entryResumePoint();
>              while (resume->caller())
>                  resume = resume->caller();
>              JS_ASSERT(script->containsPC(resume->pc()));

Oh, also, this assert will be captured in the pcToOffset() call and can be removed.

Comment 6

[Dan Gohman [:sunfish]]

(In reply to Brian Hackett (:bhackett) from comment #5)
> Comment on attachment 8344631 [details] [diff] [review]
> cg-offsets-and-leak.patch
> 
> Review of attachment 8344631 [details] [diff] [review]:
> -----------------------------------------------------------------
> 
> ::: js/src/jit/CodeGenerator.cpp
> @@ +2779,5 @@
> >              // to associate information about the inlining with.
> >              MResumePoint *resume = block->entryResumePoint();
> >              while (resume->caller())
> >                  resume = resume->caller();
> >              JS_ASSERT(script->containsPC(resume->pc()));
> 
> Oh, also, this assert will be captured in the pcToOffset() call and can be
> removed.

Ok, I removed it.

https://hg.mozilla.org/integration/mozilla-inbound/rev/66bdd9219610

Comment 7

[Wes Kocher (:KWierso) (Not reading bugmail; email directly if needed)]

https://hg.mozilla.org/mozilla-central/rev/66bdd9219610