fs/fat: In fs_stat(), when stat'ing the root directory, avoid calculating the address of the root directory entry.  The calculation is bogus (but not harmful) because the root directory does not have a directory entry.  Noted by Boris Astardzhiev.

    arch/arm/src/imxrt:  May eDMA channel linking a configuration option.  Add support to select the DMA channel priority and pre-emption controls.
    arch/arm/src/imxrt:  Update some HowTo comments in the eDMA header file.
    arch/arm/src/imxrt:  Fix a logic error in parmater passing.  Caller does not know actual channel number when setting up linked channel, only the channel handler.

    arch/arm/src/imxrt:  Fixes for clean eDMA driver build with Scatter/Gather enabled.
    arch/arm/src/imxrt:  Fixes for clean eDMA driver build with Scatter/Gather disabled.
    arch/arm/src/imxrt:  Add flags to DMA configuration to control transfer setup.  Remove some user interfaces that are inconsistent with modular design.
    arch/arm/src/imxrt:  Update DMA channel interrupt handler.
    arch/arm/src/imxrt:  Add implementation of eDMA imxrt_dmach_start().
    arch/arm/src/imxrt:  Add implementation of eDMA imxrt_dmach_setup().
    arch/arm/src/imxrt:  Add eDMA imxrt_tcd_chanlink().
    arch/arm/src/imxrt:  Add eDMA imxrt_dmach_getcount; free allocated TCDs automatically when the DMA completes or is aborted.

    arch/arm/src/imxrt:  Add structures to support list of TCDs for Scatter/Gather DMA.
    arch/arm/src/imxrt:  Add eDMA imxrt_dmach_initconfig().
    arch/arm/src/imxrt:  Add eDMA imxrt_tcd_instantiate().
    arch/arm/src/imxrt:  Replacing some of the logic cloned from SAMv7 XDMAC with eDMA logic from NXP sample code.  I am thinking that the eDMA is too complex to force into the same pattern as for other MCUs.
    arch/arms/src/imxrt/imxrt_edma.c:  Add support for in-memory TCDs.
    arch/arm/src/imxrt/chip:  Add an in-memory representation of the TCD in imxrt_edma.h

arch/arm/src/lpc17xx/lpc17_usbdev.c: Fix loss of RX initiatives.  USB bulk endpoints are double buffered on LPC17xx MCUs.  This means that up to two packets might be received on an OUT endpoint that can not be handled immediately if the receive request queue is empty.  Thus, rxpending must be a counter not a boolean flag.

arch/arm/src/imxrt:  IOMUXC logic now uses peripheral clock helpers (I think it was enabling the wrong clocks anyway).  Minor updates to eDMA logic.

arch/arm/src/imxrt: Centralilze and standardize control of peripheral clocking.  Add logic to initialize the DMAMUX.

drivers/lcd and configs/olimex-lpc1766stk:  Remove support for the Nokia6100 LCD.  That LCD uses a 9-bit SPI interface (the 9th bit being the command/data bit that is normally a discrete).  That communication was never successfully integrated.  I now believe that it would would require some special support from the low-level, MCU SPI driver to manage that 9th bit.  The deleted code is still available in the Obsoleted repository in the nuttx/drivers/lcd, nuttx/configs/olimex-lpc1766stk, and Patches directories.

configs/stm32f4discovery/include/board.h: Add SPI DMA pins definitions to get CS43L22 audio DAC working
configs/stm32f4discovery/audio: Add board configuration for CS43L22 audio example
configs/stm32f4discovery:  Update README

include/nuttx/video/fb.h:  No, I was right the first time.  sys/types.h is needed in some inclusion contexts.

include/nuttx/video/fb.h:  Make type used for overlay number consistent; Use standardized types (size_t and fb_coord_t) whenever possible.

As with the MCU temperature and VREFINT measurement, this patch requires user to enable the corresponding channel first. For Vbat channel the ioctl cmd is IO_ENABLE_DISABLE_VBAT_CH, and its arg should be a pointer to bool which must be true to enable and false to disable the Vbat channel.

Moreover, since Vbat input contains a built-in voltage divider, it is highly suggested to disable Vbat input channel after measurement is done in order to prevent battery drain through the divider.

Recent changes to UDP in order to correct problems with UDP connection mode introduced a new bug:  The 'to' address passed to sendto() must now be NULL in connection mode.  But there are errors because there was one place where the 'to' address was still being used.

This commit adds logic to first check for UDP connection mode.  If the UDP socket is connected, then the NULL 'to' address is ignored and saved connection address is used instead.

net/sixlowpan: TCP send should have PSH and ACK flags set when sending. This matches the standard network (Ethernet) TCP send logic.

arch/arm/src/imxrt:  Add framework for eDMA support.  Initial port is a rip off from the SAMA5Dx and is little more than the framework for the DMA support.

commit 7fd1f0d78546fa0315f4077b779efdd884e5bd53
Author: William Douglas <william@rocklandscientific.com>
Date:   Tue May 15 13:19:31 2018 -0700

    Add support for the second SDMMC device.

    The second SDMMC device was already supported but
    the clock was never enabled.  This fixes that.

sixlowpan: Preserve big-endian (network order) when uncompressing address.

This change is the counterpart to another recent endianness fix that occurred on compression of the ip address.

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