Re: [RFC PATCH v2] ftrace: support very early function tracing

[Steven Rostedt <rostedt@goodmis.org>]

On Wed, 8 Nov 2017 09:10:58 +0100
Peter Zijlstra <peterz@infradead.org> wrote:

> On Tue, Nov 07, 2017 at 09:17:06PM -0500, Steven Rostedt wrote:
> > > +#ifdef CONFIG_X86_TSC
> > > +	entry->clock = rdtsc();
> > > +#else
> > > +	entry->clock = trace_clock_local();
> > > +#endif  
> 
> > > +#ifdef CONFIG_X86_TSC
> > > +	cpu_khz = native_calibrate_cpu();
> > > +#endif  
> 
> > > +#ifdef CONFIG_X86_TSC
> > > +		ns = cycles_to_ns(entry->clock, cpu_khz);
> > > +#else
> > > +		ns = entry->clock;
> > > +#endif  
> 
> Yeah, no, not going to happen...

Agreed.

What you can do is have an:

	arch_early_trace_clock();

and in trace.c have:

u64 __weak arch_early_trace_clock(void)
{
	return trace_clock_local();
}

u64 __weak arch_early_trace_clock_to_ns(u64 clock)
{
	return clock;
}

and in x86 have:

u64 arch_early_trace_clock(void)
{
	return rdtsc();
}

u64 arch_early_trace_clock_to_ns(u64 clock)
{
	return cycles_to_ns(clock, cpu_khz);
}

Is that acceptable?

-- Steve