On Tue, Feb 9, 2021 at 10:26 AM Andy Lutomirski <luto@xxxxxxxxxxxxxx> wrote:
> >
> > Anything else would just be insanely complicated, I feel.
>
> The other model is “don’t do that then.”
Hmm. I guess all the code that does int3 patching could just be taught
to always go to the next instruction instead.
I don't think advancing the rewriting is an option for the asm
alternative() logic or the static call infrastructure, but those
should never be about endbr anyway, so presumably that's not an issue.
So if it ends up being _only_ about kprobes, then the "don't do that
then" might work fine.
Linus
