This is an updated version with input from H. Peter Anvin and Peter Zijlstra.
It may still not be up to par to actually pull. I would like to hear
more comments from people.

Yes, the NMI/INT3/Lockdep is complex. But I believe this is the nature
of the beast. If someone can come up with a simpler solution, I'm all ears.
Right now, the upstream tree has a bug when enabling function tracer
when lockdep is enabled that may crash the system.

Thoughts on this?

-- Steve

These are in the latest tip/perf/urgent-2 tree, which can be found at:

  git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace.git
tip/perf/urgent-2

Head SHA1: 5963e317b1e9d2a4511503916d8fd664bb8fa8fb


Steven Rostedt (5):
      ftrace: Synchronize variable setting with breakpoints
      ftrace: Use breakpoint method to update ftrace caller
      x86: Reset the debug_stack update counter
      x86: Allow nesting of the debug stack IDT setting
      ftrace/x86: Do not change stacks in DEBUG when calling lockdep

----
 arch/x86/include/asm/ftrace.h |    2 +-
 arch/x86/kernel/cpu/common.c  |    8 +++-
 arch/x86/kernel/entry_64.S    |   44 ++++++++++++++++--
 arch/x86/kernel/ftrace.c      |  102 ++++++++++++++++++++++++++++++++++++++---
 arch/x86/kernel/nmi.c         |    6 ++-
 arch/x86/kernel/traps.c       |    8 +++-
 6 files changed, 154 insertions(+), 16 deletions(-)