Re: [linux-next:master 8539/9537] include/linux/ftrace.h:126:16: error: implicit declaration of function 'arch_ftrace_get_regs'; did you mean 'ftrace_get_regs'?

[Mark Rutland <mark.rutland@arm.com>]

On Mon, Nov 21, 2022 at 12:40:15PM -0500, Steven Rostedt wrote:
> On Mon, 21 Nov 2022 17:26:15 +0000
> Mark Rutland <mark.rutland@arm.com> wrote:
> 
> > ... and since arm64 can be built with DYNAMIC_FTRACE=n we end up with:
> > 
> >   CONFIG_DYNAMIC_FTRACE=n
> 
> Is there any reason to have CONFIG_DYNAMIC_FTRACE=n and
> CONFIG_FUNCTION_TRACE=y?
>
> The only reason I keep the option to turn off DYNAMIC_FTRACE is for
> debugging purposes.
> 
> That is, if you have HAVE_DYNAMIC_FTRACE, then DYNAMIC_FTRACE should be
> enabled whenever it is needed by something else.
> 
> The reason for static ftrace is simply to help new architectures work their
> way to having DYNAMIC_FTRACE enabled. Perhaps the solution is to get rid of
> the DYNAMIC_FRACE prompt, and have it set whenever FUNCTION_TRACER is set
> and HAVE_DYNAMIC_FTRACE is set.

To be honest, we just didn't think to remove static ftrace; I'm happy to remove
that for arm64.

That said, I still think it makes sense to change the code to check 
CONFIG_DYNAMIC_FTRACE_WITH_ARGS rather than
CONFIG_HAVE_DYNAMIC_FTRACE_WITH_ARGS, as mentioned below, unless you intend for
that to have an effect on static ftrace on x86?

Everywhere else with DYNAMIC_FTRACE_WITH_ARGS forces DYNAMIC_FTRACE, so it ends
up equivalent; x86 and arm64 are the only exceptions today.

Thanks,
Mark.

> 
> For debugging purposes, I may add
> 
> HAVE_DEBUG_DYNAMIC_FTRACE
> 
> which brings back the prompt (for x86 only).
> 
> -- Steve
> 
> 
> >   CONFIG_HAVE_DYNAMIC_FTRACE_WITH_ARGS=y
> >   CONFIG_DYNAMIC_FTRACE_WITH_ARGS=n
> > 
> > Looking at this some more, ftrace_regs_has_args() should check for
> > CONFIG_DYNAMIC_FTRACE_WITH_ARGS rather than
> > CONFIG_HAVE_DYNAMIC_FTRACE_WITH_ARGS, or it'll bogusly report we have args when
> > CONFIG_DYNAMIC_FTRACE=n.
> > 
> > I'd prefer to change the core ftrace code to check for
> > CONFIG_DYNAMIC_FTRACE_WITH_ARGS for consistency with that, but I'll need to
> > check how that affects other architectures.