Re: [PATCH v2 0/2] x86, refcount: Implement fast refcount overflow

[Kees Cook]

On Tue, Apr 25, 2017 at 7:01 PM, PaX Team <pageexec@xxxxxxxxxxx> wrote:
> On 25 Apr 2017 at 15:56, Kees Cook wrote:
>
>> This protection is a modified version of the x86 PAX_REFCOUNT
>> implementation from PaX/grsecurity. This speeds up the refcount_t API by
>> duplicating the existing atomic_t implementation with a single instruction
>> added to detect if the refcount has wrapped past INT_MAX (or below 0)
>> resulting in a signed value.
>
> 'signed value' sounds somewhat ambiguous given that in C a signed type (such
> as the one beneath refcount_t) can have both negative and positive values yet
> you didn't mean the latter here i guess.

Yeah, the language for the CPU "sign flag" confuses this. I will
attempt to clarify for future versions.

>> Various differences from PaX:
>> - uses "js" instead of "jo" to trap all signed results instead of just
>>   under/overflow transitions
>
> there're differences in my 4.11 port but this isn't one of them.

Any changes you'd suggest for upstreaming?

-Kees

-- 
Kees Cook
Pixel Security