Re: x86 entry perf unwinding failure (missing IRET_REGS annotation on stack switch?)

From: Josh Poimboeuf
Date: Tue Apr 28 2020 - 12:33:41 EST


On Tue, Apr 28, 2020 at 09:25:51AM -0700, Sean Christopherson wrote:
> On Tue, Apr 28, 2020 at 10:54:13AM -0500, Josh Poimboeuf wrote:
> > On Tue, Apr 28, 2020 at 10:49:09AM -0500, Josh Poimboeuf wrote:
> > > On Tue, Apr 28, 2020 at 05:25:52PM +0200, Peter Zijlstra wrote:
> > > > On Tue, Apr 28, 2020 at 09:31:57AM -0500, Josh Poimboeuf wrote:
> > > > > That's quite the monstrosity, and I still don't see the point. I
> > > > > thought we decided to just disallow CFI changes in alternatives anyway?
> > > > > That can be done much simpler.
> > > >
> > > > Something like so then ?
> > > >
> > > > ---
> > > > diff --git a/tools/objtool/check.c b/tools/objtool/check.c
> > > > index 8443ec690051..d14d83e6edb0 100644
> > > > --- a/tools/objtool/check.c
> > > > +++ b/tools/objtool/check.c
> > > > @@ -940,6 +940,7 @@ static int handle_group_alt(struct objtool_file *file,
> > > >
> > > > last_new_insn = insn;
> > > >
> > > > + insn->alt_group = true;
> > > > insn->ignore = orig_insn->ignore_alts;
> > > > insn->func = orig_insn->func;
> > > >
> > > > @@ -2242,6 +2243,11 @@ static int handle_insn_ops(struct instruction *insn, struct insn_state *state)
> > > > list_for_each_entry(op, &insn->stack_ops, list) {
> > > > int res;
> > > >
> > > > + if (insn->alt_group) {
> > > > + WARN_FUNC("alternative has CFI", insn->sec, insn->offset);
> > > > + return 1;
> > > > + }
> > > > +
> > >
> > > ACK (separate patch)
> >
> > BTW, since most people don't know what CFI is, how about something like
> >
> > "unsupported stack change in alternatives code"
>
> Would it be accurate to print
>
> "unsupported CFI stack change in alternatives code"?
>
> to give the developer something more explicit to plug into their search
> engine?

I don't have a strong opinion either way, though this warning is going
to be documented in stack-validation.txt anyway right Peter? :-)

--
Josh