Added BOARD_MCG_C2_FCFTRIM and BOARD_MCG_C2_LOCRE0 to
  configure the MCG_C2 register
  cleanup of some comments

  Original BOARD_FR_DIV was never used - that is a good thing because
  the value ws definec shifted and the code also shifited it.

  We define the bits as a common set of names. This means that
  an index may be added to a name i.e. LOCK is LOCK0 as that is
  the superset name.

  This includes arch/arm/include/kinetis/kinetis_mcg.h
  to bring in the MCG versioning and defines the KINETIS_K66
  family for the added SoCs:

   --------------- ------- --- ------- ------- ------ ------ ------ -----
   PART NUMBER     CPU     PIN PACKAGE  TOTAL  PROGRAM EEPROM SRAM  GPIO
                   FREQ    CNT          FLASH  FLASH
   --------------- ------- --- ------- ------- ------ ------ ------ -----
   MK66FN2M0VMD18  180 MHz 144 MAPBGA   2   MB    —    — KB  260 KB 100
   MK66FX1M0VMD18  180 MHz 144 MAPBGA  1.25 MB  1 MB   4 KB  256 KB 100
   MK66FN2M0VLQ18  180 MHz 144 LQFP     2   MB    —    — KB  260 KB 100
   MK66FX1M0VLQ18  180 MHz 144 LQFP    1.25 MB  1 MB   4 KB  256 KB 100

   The motvations is to version the IP blocks of the Kinetis
   K series family of parts.

   This added versioning and configuration features for the
   Kinetis MCG IP block.

   It is envisioned that in the long term as a chip is added.
   The author of the new chip definitions will either find
   the exact configuration in an existing  chip define and
   add the new chip to it Or add the MCG feature configuration
   #defines to the chip ifdef list in
   arch/arm/include/kinetis/kinetis_mcg.h  In either case the
   author should mark it as "Verified to Document Number:"
   taken from the reference manual.

   The version KINETIS_MCG_VERSION_UKN has been applied to
   most all the SoCs in the kinetis arch prior to this commit.

   The exceptions are the CONFIG_ARCH_CHIP_MK60FN1M0VLQ12,
   All K64 and K66 which not have Verified MCG configurations.

Merged in david_s5/nuttx-5/david_s5/there-is-a-return-path-with-node-not-set-1486503652666 (pull request #204)

There is a return path with node not set

Approved-by: Gregory Nutt

inode/fs_inoderemove.c: In function 'inode_unlink':
inode/fs_inoderemove.c:74:17: warning: 'node' may be used uninitialized in this function [-Wmaybe-uninitialized]
   FAR struct inode *node;

up_timer_initialize() is named incorrectly.  The prefix should be the architecture name, not up_ since it is private to the architecture.  up_timerisr() is similarly misnamed and should also be private since it is used only with the xyz_timerisr.c files.  Also updat TODO list.

Soft link: Change how a value is returned so that _inode_linktarget will function in all of its use contexts.

Soft links:  Eliminate in-stack allocation of path working buffer needed for traversal fo paths with soft links.

Add macros support that will eventually allow dynamic allocation of strings need to support soft links.

tools/noteinfo.c: A hack tool that I use to analyze some sched_note output.  Needs a home and may be useful to others.

inode_find:  Now takes struct inode_desc_s type as input.  This was necessary before that structure includes some data storage.  It was used within inode_find(), but that means that the life of the data was the life of inode_find().  That data must persist longer.  It is now provided by the caller so that the life of the data persists for the entire life of the caller.

FS: Remove inode_search_nofollow().  That logic is no handled with a new field in the input argument structure.

Soft links: Fix logic and remove kludge of last commit.  Rename inode_dereference() as inode_linktarget() and make global.

FS: Separate inode_search() and inode_free() from fs_inode.c and put in separate files.  Flesh out symbolic link logic in stat() and readdir().  There are still some issues with stat().