Re: [PATCH 2/3] x86/insn-eval: Introduce insn_decode_mmio()

[Kirill A. Shutemov]

On Tue, Oct 19, 2021 at 10:34:45AM +0200, Peter Zijlstra wrote:
> On Mon, Oct 18, 2021 at 07:49:41PM +0300, Kirill A. Shutemov wrote:
> > On Mon, Oct 18, 2021 at 05:53:49PM +0200, Peter Zijlstra wrote:
> > > On Mon, Oct 18, 2021 at 06:33:32PM +0300, Kirill A. Shutemov wrote:
> > > 
> > > > diff --git a/arch/x86/lib/insn-eval.c b/arch/x86/lib/insn-eval.c
> > > > index fbaa3fa24bde..2ab29d8d6731 100644
> > > > --- a/arch/x86/lib/insn-eval.c
> > > > +++ b/arch/x86/lib/insn-eval.c
> > > > @@ -1559,3 +1559,85 @@ bool insn_decode_from_regs(struct insn *insn, struct pt_regs *regs,
> > > >  
> > > >  	return true;
> > > >  }
> > > > +
> > > > +/**
> > > > + * insn_decode_mmio() - Decode a MMIO instruction
> > > > + * @insn:	Structure to store decoded instruction
> > > > + * @bytes:	Returns size of memory operand
> > > > + *
> > > > + * Decodes instruction that used for Memory-mapped I/O.
> > > > + *
> > > > + * Returns:
> > > > + *
> > > > + * Type of the instruction. Size of the memory operand is stored in
> > > > + * @bytes. If decode failed, MMIO_DECODE_FAILED returned.
> > > > + */
> > > > +enum mmio_type insn_decode_mmio(struct insn *insn, int *bytes)
> > > > +{
> > > > +	int type = MMIO_DECODE_FAILED;
> > > > +
> > > > +	*bytes = 0;
> > > > +
> > > > +	insn_get_opcode(insn);
> > > 
> > > insn_get_opcode() can fail. Either you assume it's already called and
> > > don't call it, or you can't assume anything and get to do error
> > > handling.
> > 
> > Fair enough. I will return MMIO_DECODE_FAILED if insn_get_opcode() fails.
> > 
> > BTW, looks like is_string_insn() suffers from the same issue. Not sure how
> > to fix it though.
> 
> AFAICT all callers of insn_get_addr_ref() (which is what
> is_string_insn() seems to be part of) do a insn_decode_*() call with
> error checking before.
> 
> So it looks like that insn_get_opcode() in there is superfluous.
> probably same for insn_has_rep_prefix() / get_seg_reg_override_idx() /
> get_eff_addr_modrm_*(). That all wants cleaning up.
> 
> The esaiest way is probably to push those things up the callchains into
> the !static function and have it fail early there.

Looks like insn_get_addr_ref() is the only non-static callsite:

	is_string_insn
	  check_seg_overrides
	    resolve_seg_reg
	      get_seg_base_limit
		get_addr_ref_XX
		  insn_get_addr_ref
	  resolve_default_seg
	    resolve_seg_reg (as above)

What about this: