Re: [PATCH] x86/alternative: Support relocations in alternatives

[Peter Zijlstra]

On Tue, Feb 07, 2023 at 05:05:27PM +0100, Peter Zijlstra wrote:

> Also, Masami, how difficuly would it be to do insn_is_nop() that matches
> most/all conventional NOP instructions?

Just looked at that, while NOP and NOPL are relatively easy to do, the
32bit NOPs are a pain to decode since they're actual instructions with
the side-effect of not actually doing anything, like 'mov %reg,%reg' and
'lea (%reg),%reg'. Esp. that latter is popular since it has all the
various displacement encodings to grow it into a very long instruction.

I'll put it on the todo list I suppose, not something for now.