Tracers want a correlated time between the kernel instrumentation and
user space. We really do not want to export sched_clock() to user
space, so we need to provide something sensible for this.

Using separate data structures with an non blocking sequence count
based update mechanism allows us to do that. The data structure
required for the readout has a sequence counter and two copies of the
timekeeping data.

On the update side:

  smp_wmb();
  tkf->seq++;
  smp_wmb();
  update(tkf->base[0], tk);
  smp_wmb();
  tkf->seq++;
  smp_wmb();
  update(tkf->base[1], tk);

On the reader side:

  do {
     seq = tkf->seq;
     smp_rmb();
     idx = seq & 0x01;
     now = now(tkf->base[idx]);
     smp_rmb();
  } while (seq != tkf->seq)

So if a NMI hits the update of base[0] it will use base[1] which is
still consistent, but this timestamp is not guaranteed to be monotonic
across an update.

The timestamp is calculated by:

	now = base_mono + clock_delta * slope

So if the update lowers the slope, readers who are forced to the
not yet updated second array are still using the old steeper slope.

 tmono
 ^
 |    o  n
 |   o n
 |  u
 | o
 |o
 |12345678---> reader order

 o = old slope
 u = update
 n = new slope

So reader 6 will observe time going backwards versus reader 5.

While other CPUs are likely to be able observe that, the only way
for a CPU local observation is when an NMI hits in the middle of
the update. Timestamps taken from that NMI context might be ahead
of the following timestamps. Callers need to be aware of that and
deal with it.

V2: Got rid of clock monotonic raw and reorganized the data
    structures. Folded in the barrier fix from Mathieu.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: John Stultz <john.stultz@linaro.org>