Re: [PATCH v2 1/3] fortify: Fix __compiletime_strlen() under UBSAN_BOUNDS_LOCAL

[Kees Cook]

On Tue, Sep 06, 2022 at 07:36:46PM -0700, Nick Desaulniers wrote:
> On Fri, Sep 2, 2022 at 1:43 PM Kees Cook <keescook@xxxxxxxxxxxx> wrote:
> >
> > Co-developed-by: Nick Desaulniers <ndesaulniers@xxxxxxxxxx>
> 
> That's overly generous of you!

Well, it was a lot of work to track down, and you wrote it up that way,
I just moved things around a little bit. :)

> Anyways, the disassembly LGTM and the bot also came back green.
> 
> Reviewed-by: Nick Desaulniers <ndesaulniers@xxxxxxxxxx>
> Tested-by: Android Treehugger Robot
> Link: https://android-review.googlesource.com/c/kernel/common/+/2206839

Thank you!

> Another thought, Nikita suggested that you could also compare mode 1 vs mode 3:
> https://github.com/llvm/llvm-project/issues/57510#issuecomment-1235126343

Yeah, it could work (I tried this as well), but I think the better
approach is checking index 0.

> That said, since mode 3 returns 0 for "unknown" I'd imagine that
> wouldn't be pretty since it wouldn't be a direct comparison against
> __p_size.

Yeah -- it is a little weird. I might come back to this if we get more
glitches like this in the future.

-- 
Kees Cook