Bug 156780 – bmalloc: fix up overflow checks

• Status: RESOLVED
• Resolution: FIXED
• Component: bmalloc

[Geoffrey Garen]

bmalloc: fix up overflow checks

[Geoffrey Garen]

Created attachment 276795 [details]
Patch

[Mark Lam]

Comment on attachment 276795 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=276795&action=review

r=me

> Source/bmalloc/ChangeLog:11
> +        we were missing smoe cases.

typo: /smoe/some/.

[Mark Lam]

Comment on attachment 276795 [details]
Patch

Is it possible to add some tests that exercise the overflow cases?

[Geoffrey Garen]

> Is it possible to add some tests that exercise the overflow cases?

Good question. This requires some changes to the MallocBench infrastructure to allow for failure. I'll tackle this in a follow-up patch.

[Geoffrey Garen]

Committed r199759: <http://trac.webkit.org/changeset/199759>

[Oliver Hunt]

Comment on attachment 276795 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=276795&action=review

> Source/bmalloc/bmalloc/Heap.cpp:347
> +    size_t roundedSize = size ? roundUpToMultipleOf(largeAlignment, size) : largeAlignment;

I wonder if we should make an override for roundUp... to take Checked<T> - then switch to Checked<> for bmalloc internal arithmetic?

[Geoffrey Garen]

> > Source/bmalloc/bmalloc/Heap.cpp:347
> > +    size_t roundedSize = size ? roundUpToMultipleOf(largeAlignment, size) : largeAlignment;
> 
> I wonder if we should make an override for roundUp... to take Checked<T> -
> then switch to Checked<> for bmalloc internal arithmetic?

I'd like to come up with a good idiom, maybe using builtin checked arithmetic. Not sure what the best option is.