Re: [PATCH v3 06/10] x86/traps: Decode LOCK Jcc.d8 #UD

[Peter Zijlstra]

On Wed, Feb 19, 2025 at 10:20:25AM -0800, Kees Cook wrote:

> I realize these are misplaced chunks, but passing ud_type into the
> handler feels like a layering violation to me. I struggled with this
> when making recommendations for the UBSAN handler too, so I'm not sure
> I have any better idea. It feels like there should be a way to separate
> this logic more cleanly. The handlers are all doing very similar things:
> 
> 1- find the address where a bad thing happened
> 2- report about it
> 3- whether to continue execution
> 4- where to continue execution
> 
> The variability happens with 1 and 4, where it depends on the instruction
> sequences. Meh, I dunno. I can't see anything cleaner, so passing down
> ud_type does seem best.

Yeah, agreed. I couldn't get rid of relying on ud_type entirely (it was
worse), I'll see if I can come up something.