Re: [PATCH v5 02/17] objtool: Better handle IRET

[Miroslav Benes <mbenes@suse.cz>]

On Fri, 17 Apr 2020, Peter Zijlstra wrote:

> On Fri, Apr 17, 2020 at 01:29:32PM +0200, Miroslav Benes wrote:
> > On Thu, 16 Apr 2020, Peter Zijlstra wrote:
> 
> > > +		case INSN_EXCEPTION_RETURN:
> > > +			if (handle_insn_ops(insn, &state))
> > > +				return 1;
> > > +
> > > +			/*
> > > +			 * This handles x86's sync_core() case, where we use an
> > > +			 * IRET to self. All 'normal' IRET instructions are in
> > > +			 * STT_NOTYPE entry symbols.
> > > +			 */
> > > +			if (func)
> > > +				break;
> > > +
> > > +			return 0;
> > > +
> > >  		case INSN_CONTEXT_SWITCH:
> > >  			if (func && (!next_insn || !next_insn->hint)) {
> > >  				WARN_FUNC("unsupported instruction in callable function",
> > 
> > It looks really simple.
> > 
> > Have you tried Julien's proposal about removing INSN_STACK altogether, 
> > move the x86 to arch/x86/ and call handle_insn_ops() unconditionally, or 
> > have you just postponed it? As I said, I think it could be better in the 
> > long term, but the above looks good for now as well.
> 
> If you look at this other set I send yesterday:
> 
>   https://lkml.kernel.org/r/20200416150752.569029800@infradead.org
> 
> (also, sorry for not adding you to the Cc; also best look at the gitweb
> version, the patches I send out are missing a hunk and lacking some
> back-merges.. clearly I wasn't having a good day yesterday).
> 
> it has this intra_function_calls crud that needs explicit conditional
> handle_insn_ops().

Ah, ok. Thanks for letting me know. There are so many patches for objtool 
flying around now that it is easy to miss something.

Miroslav