Fix C++ __guard implementation for ARM.  The standard C++ ABI that most platforms follow defines __guard to be 64 bits.  The existing implementation of libxx_cxa_guard.cxx follows this.  However, the 32-bit ARM C++ ABI defines it as 32 bits instead, and changes the meaning slightly so only the lowest bit is used. This matters because GCC creates guard symbols without regards to what libxx_cxa_guard.cxx says.  So on ARM, gcc allocates 4 bytes, but __cxa_guard_release writes 8 bytes, zeroing out another unlucky variable nearby. Fix it by special-casing 32-bit ARM in libxx_cxa_guard.

configs/:  Back out many of the changes in two recent commits:  Too many changes from SPIDEV_WIRELESS to SPIDEV_CONTACTLESS.  Specifically NRF24L01 should still be SPIDEV_WIRELESS.

stm32f103-minimum: dd GPIO device driver example for STM32F103-Minimum.  This chang adds the initialization needed by stm32f103-minimum board to support the NuttX GPIO Subsystem.

STM32L4: Add STM32L475 pinmap.  Initial cut is just the the L476 pinmap with unsupported devices removed.

STM32L4:  Ad support for the STM32L475 family.  Incomplete -- still needs pinmap, rcc, otgfs, syscfg

Kinetis:Fixed waning for kinetis_mpudisable

Approved-by: Gregory Nutt <gnutt@nuttx.org>

  Missing header file added

Upstream kinetis usb

Approved-by: Gregory Nutt <gnutt@nuttx.org>

   Removed the notion of attached. The khci_usbattach is call early in
   the init either in board_initalize or in board_app_initalize. In
   either case it is always done prior to the the class register.
   Therefore the khci_usbattach call only set a flag, and that
   flag is only tested in the class register. The class register will
   enable the soft connect pull up.

  The call is not need by the driver if CONFIG_USBDEV_BUSPOWERED=y
  The class register will enable the soft connect pull up. The
  khci_usbattach call only set a flag, and that  flag is only tested
  in the class register. On a USB powered device if we are running we
  are attached.

   This fixes a bug were a SoC does not have a clockdivN register
   and passes a 0 for the init value. This prevents overflow of
   the 0 decremented to -1 (0xffffffff) spilling over to other
   clockdivN feilds.

  Remove magic numbers from code, documented the use of
  undocumented bits.

  Remove comments and code that were not appropriate for this
  hardware.

  Removed ifdef that's that were always compiled and removed code
  blocks that were never compiled.

  Ensure proper access order to hardware.
  Per the reference manual: disable endpoints prior to configuring buffer
  descriptor, then enable endpoints
  Reorganize interrupt processing order to offload data after processing
  errors.

  Reorganize initialization so that there is a clear initialization phase,
  reset phase for both the hardware and software structures.

  By breaking the initialization into smaller pieces, the reset interrupt
  only resets the resources within the controller that should be reset.

  Rework suspend and resume logic so they perform properly

  Made attach and detach functions optional. As they do not make sense for
  a bus powered device.

  Ensured the calls to up_usbinitalize up_usbuninitalize do not violate the
  USB spec.

   Allow the board config to define the USBOTG-FS to have Read access to
   FLASH.

   The hardware reset state of the the MPU precludes any bus
   masters other then DMA access to memory. Unfortunately
   USB and SDHC have there own DMA and will not have access to
   memory in the default reset state.

   This change disabled the MPU if present on system startup.

MTD FLASH drivers:  The byte write method of several drivers had a cloned error:  It was not locking the bus while performing byte write operations.

libfflush remove extra semicolon

Approved-by: Gregory Nutt <gnutt@nuttx.org>

Comment out most of 430060e9.  Clean-up some dependencies in top-level Makefile

This reverts commit 430060e9.

That commit has some unexpected side-effects