Re: [PATCH 1/5] refcount_t: A special purpose refcount type

[Kees Cook]

On Fri, Feb 3, 2017 at 5:25 AM, Peter Zijlstra <peterz@xxxxxxxxxxxxx> wrote:
> Provide refcount_t, an atomic_t like primitive built just for
> refcounting.
>
> It provides saturation semantics such that overflow becomes impossible
> and thereby 'spurious' use-after-free is avoided.

Wheee :) Thanks for working on this

>
> Signed-off-by: Peter Zijlstra (Intel) <peterz@xxxxxxxxxxxxx>
> ---
> [...]
> --- a/lib/Kconfig.debug
> +++ b/lib/Kconfig.debug
> @@ -716,6 +716,19 @@ source "lib/Kconfig.kmemcheck"
>
>  source "lib/Kconfig.kasan"
>
> +config DEBUG_REFCOUNT
> +       bool "Verbose refcount checks"
> +       --help--

Quick feedback as I start playing with this: this isn't valid Kconfig
syntax (build breaks). It should either be "---help---" or just
"help", latter preferred.

> +         Say Y here if you want reference counters (refcount_t and kref) to
> +         generate WARNs on dubious usage. Without this refcount_t will still
> +         be a saturating counter and avoid Use-After-Free by turning it into
> +         a resource leak Denial-Of-Service.
> +
> +         Use of this option will increase kernel text size but will alert the
> +         admin of potential abuse.
> +
> +         If in doubt, say "N".
> +
>  endmenu # "Memory Debugging"
>
>  config ARCH_HAS_KCOV
>
>

-Kees

-- 
Kees Cook
Pixel Security