Re: [PATCH 2/9] objtool: Fix ORC unwinding in non-JIT BPF generated code

[Peter Zijlstra]

On Fri, Jun 14, 2019 at 12:35:38AM -0700, Alexei Starovoitov wrote:
> On Fri, Jun 14, 2019 at 09:08:52AM +0200, Peter Zijlstra wrote:
> > On Thu, Jun 13, 2019 at 08:20:30PM -0500, Josh Poimboeuf wrote:
> > > On Thu, Jun 13, 2019 at 01:57:11PM -0700, Alexei Starovoitov wrote:
> > 
> > > > and to patches 8 and 9.
> > > 
> > > Well, it's your code, but ... can I ask why?  AT&T syntax is the
> > > standard for Linux, which is in fact the OS we are developing for.
> > 
> > I agree, all assembly in Linux is AT&T, adding Intel notation only
> > serves to cause confusion.
> 
> It's not assembly. It's C code that generates binary and here
> we're talking about comments.

And comments are useless if they don't clarify. Intel syntax confuses.

> I'm sure you're not proposing to do:
> /* mov src, dst */
> #define EMIT_mov(DST, SRC)                                                               \
> right?

Which is why Josh reversed both of them. The current Intel order is just
terribly confusing. And I don't see any of the other JITs having
confusing comments like this.

> bpf_jit_comp.c stays as-is. Enough of it.

I think you're forgetting this is also arch/x86 code, and no, it needs
changes because its broken wrt unwinding.