Re: Curious about Copy Fail Linux bug and any relation to FreeBSD

This is a logic bug in algif_aead on linux.
This bug will only ever apply to linux.

linux.ko does not emulate algif_aead.

Furthermore, this bug is a combo of arch decisions implemented in
linux that simply do not exist in FBSD.

Firstly, there's no AF_ALG on FBSD.  Nothing is exposed to userland.
FBSD's cryptodev is an ioctl based approach and has much tighter
scoping.
Secondly, there's no 'splice into pipe' surface area in FBSD.
Lastly, the OCF (OpenCrypto Framework) never made the same
optimization that was in authencesn.

The above isn't to say that one day a bug like this can't make it into
FBSD by implementing looser controls but as it stands now, this
specific type of 'splice into pipe' attack simply does not exist.

Thanks,
Paul Procacci

On Fri, May 1, 2026 at 2:17 PM Nathan <[email protected]> wrote:
>
> Hi, I just wanted to ask about the recently talked-about Copy-Fail bug in Linux.
> I was wondering if the FreeBSD community has tried to replicate it in FBSD , or if that even makes sense. Why does it even happen in Linux, and maybe why it doesn't here?
>
> And I'm just sorta curious if there is some fundamental structural decision that might be protecting FBSD from that type of exploit. Or if it was simply a fluke bug.
>
> I know I'm asking large questions, just thought I'd reach out to this list to prompt the topic and maybe get pointed in some direction.
> If it hasn't been tested I'd love to try to test it myself, but I'm in a bit over my head.
>


-- 
__________________

:(){ :|:& };: