Re: [PATCH 1/5] [PATCH 1/5] function-graph/x86: replace unbalanced ret with jmp

[Frederic Weisbecker <fweisbec@gmail.com>]

On Tue, Oct 13, 2009 at 05:12:46PM -0400, Steven Rostedt wrote:
> On Tue, 2009-10-13 at 23:02 +0200, Frederic Weisbecker wrote:
> > On Tue, Oct 13, 2009 at 04:33:50PM -0400, Steven Rostedt wrote:
> > > From: Steven Rostedt <srostedt@redhat.com>
> > > 
> > > The function graph tracer replaces the return address with a hook to
> > > trace the exit of the function call. This hook will finish by returning
> > > to the real location the function should return to.
> > > 
> > > But the current implementation uses a ret to jump to the real return
> > > location. This causes a imbalance between calls and ret. That is
> > > the original function does a call, the ret goes to the handler
> > > and then the handler does a ret without a matching call.
> > > 
> > > Although the function graph tracer itself still breaks the branch
> > > predictor by replacing the original ret, by using a second ret and
> > > causing an imbalance, it breaks the predictor even more.
> > 
> > 
> > 
> > I have troubles to understand by it breaks the predictor, especially
> > since there is not conditional branch in return_to_handler.
> > But still I don't understand why a ret would break more the branch
> > prediction than a jmp.
> 
> Calls are branch prediction jumps. Which associates the "ret" with the
> call. As it approaches the ret, it starts to receive the code after the
> call.
> 
> But this is stack order. Every call should hit one ret. But with the
> original code, we break this stack. We have one call and two rets. Which
> means that the branch prediction will also get messed up with the
> previous stored call.


Oh, ok I got it.
Thanks.