On Fri, Jan 07, 2022 at 12:01:17PM +0000, Catalin Marinas wrote:

> Regarding (1), I don't remember whether we decided to do it this way
> because it was more complicated to handle it in the kernel (like the 4
> more patches in this series) or because we wanted to leave the option to
> the dynamic loader. It would be good to clarify this and we may have a
> small window, as Jeremy said, where changing the ABI won't cause
> problems (well, hopefully, there's still a risk).

My understanding is that it was basically just a "let's defer everything
to userspace" thing.  It means that userspace is responsible for turning
on BTI and is therefore responsible for any workarounds which are needed
for problematic binaries, it's the absolute minimum the kernel can be
responsible for.  This all predates my involvement though.