Re: [RFC PATCH v2 00/17] objtool: Function validation tracing

[Peter Zijlstra <peterz@infradead.org>]

On Wed, Sep 24, 2025 at 12:58:13PM +0200, Peter Zijlstra wrote:
> On Wed, Sep 24, 2025 at 12:52:35PM +0200, Alexandre Chartre wrote:
> > 
> > On 9/24/25 12:10, Peter Zijlstra wrote:
> > > On Wed, Sep 24, 2025 at 12:08:47PM +0200, Peter Zijlstra wrote:
> > > > On Wed, Sep 24, 2025 at 09:42:06AM +0200, Peter Zijlstra wrote:
> > > > 
> > > > > > >    d051:  perf_get_x86_pmu_capability+0x51    | xchg   %ax,%ax
> > > > 
> > > > That libopcode is 'funny', isn't that typically spelled "nop" ?
> > > 
> > > Ooh, I see, it is "osp nop" and yeah binutils also seems to do that as
> > > "xchg %ax,%ax".
> > 
> > Yes, "xchg %ax,%ax" is NOP2 (opcodes: 0x66 0x90), "nop" is NOP1 (0x90).
> > 
> > That's one more improvement we can do: identify NOP instructions and
> > display them as NOP<n>
> 
> Right, I have just the function for that. Let me do a patch for you.

Bah, I'm having trouble with the makefiles, but we sorta already detect
nops, and the below should sorta work. I'll improve decode.c a bit.

--- a/tools/objtool/disas.c
+++ b/tools/objtool/disas.c
@@ -273,6 +273,9 @@ static size_t disas_insn(struct disas_co
 	dctx->insn = insn;
 	dctx->result[0] = '\0';
 
+	if (insn->type == INSN_NOP)
+		return snprintf(dctx->result, DISAS_RESULT_SIZE, "NOP%d", insn->len);
+
 	/*
 	 * Set the disassembler buffer to read data from the section
 	 * containing the instruction to disassemble.