Re: [PATCH 2/2] rust: time: Use wrapping_sub() for Ktime::sub()

[Kees Cook]

On Tue, Apr 23, 2024 at 02:11:22PM -0700, Boqun Feng wrote:
> On Thu, Apr 11, 2024 at 04:08:01PM -0700, Boqun Feng wrote:
> > Currently since Rust code is compiled with "-Coverflow-checks=y", so a
> > normal substraction may be compiled as an overflow checking and panic
> > if overflow happens:
> > 
> >         subq    %rsi, %rdi
> >         jo      .LBB0_2
> >         movq    %rdi, %rax
> >         retq
> > .LBB0_2:
> >         pushq   %rax
> >         leaq    str.0(%rip), %rdi
> >         leaq    .L__unnamed_1(%rip), %rdx
> >         movl    $33, %esi
> >         callq   *core::panicking::panic::h59297120e85ea178@GOTPCREL(%rip)
> > 
> > although overflow detection is nice to have, however this makes
> > `Ktime::sub()` behave differently than `ktime_sub()`, moreover it's not
> > clear that the overflow checking is helpful, since for example, the
> > current binder usage[1] doesn't have the checking.
> > 
> 
> Ping. Thomas, John and Stepthen. Could you take a look at this, and the
> discussion between Miguel and me? The key question is the behavior when
> ktime_sub() hits a overflow, I think. Thanks!
> 
> (Cc Kees as well)

While working on the signed (and unsigned) integer overflow sanitizer
support on the C side for the kernel, I've also run into timekeeping
being a questionable area[1]. I *think* from what I can tell, it's always
expected to have wrapping behavior.

Can we define the type itself to be wrapping? (This has been my plan on
the C side, but we're still waiting on a finalized implementation of the
"wraps" attribute.[2])

-Kees

[1] This is strictly WIP, as I think fixing the _types_ is going to be
    the more sustainable solution, but you can see some of what I was
    poking at:
    https://git.kernel.org/pub/scm/linux/kernel/git/kees/linux.git/commit/?h=devel/overflow/enable-unsigned-sanitizer&id=284464817a59b14f00d397bfbf1bf05683ed2f58
[2] https://github.com/llvm/llvm-project/pull/86618

-- 
Kees Cook