This driver supports boards that have 1 or 2 TIO chips with base
addresses 0x800 apart. Replace the static const array 'GPCT_OFFSET'
with a define and calculate the base address based on the chip index.
Signed-off-by: H Hartley Sweeten <hsweeten@visionengravers.com>
Reviewed-by: Ian Abbott <abbotti@mev.co.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

This spinlock is only used to protect changes to the private data 'dma_cfg'.

Before calling any function that would change the 'dma_cfg' the spinlock
'mite_channel_lock' is also locked. That spinlock is not unlocked until
after the 'dma_cfg' change.

Remove the redundant spinlock.
Signed-off-by: H Hartley Sweeten <hsweeten@visionengravers.com>
Reviewed-by: Ian Abbott <abbotti@mev.co.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

enum ni_gpct_register and enum ni_660x_register now have a 1:1
relationship for the NITIO_* registers. The static const array
is no longer necessary to find the proper NI660X_* register for
a given NITIO_*. Remove it and refactor the register read/write
functions.

Use the NITIO_* values to init the ni_660x_reg_data[] array and
remove the unnecessary NI660X_* enum values.
Signed-off-by: H Hartley Sweeten <hsweeten@visionengravers.com>
Reviewed-by: Ian Abbott <abbotti@mev.co.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Sort this enum so that it has a 1:1 relationship with the ni_tio.h
enum ni_gpct_register.
Signed-off-by: H Hartley Sweeten <hsweeten@visionengravers.com>
Reviewed-by: Ian Abbott <abbotti@mev.co.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

This fuction just returns a pointer from the private data. The name
might provide some confusion since it appears to be an exported
function from the mite driver.

Just remove it and get the pointer directly where needed.
Signed-off-by: H Hartley Sweeten <hsweeten@visionengravers.com>
Reviewed-by: Ian Abbott <abbotti@mev.co.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

For aesthetics, move the initialization of the default routing for the
DIO channels so it happens when the subdevice is initialized.
Signed-off-by: H Hartley Sweeten <hsweeten@visionengravers.com>
Reviewed-by: Ian Abbott <abbotti@mev.co.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Remove the extra for loop and just initialize the counter as the subdevices
are created.
Signed-off-by: H Hartley Sweeten <hsweeten@visionengravers.com>
Reviewed-by: Ian Abbott <abbotti@mev.co.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

The ni_gpct_device_construct() could fail allocating the memory for
device and its counters. For aesthetics, call the function before
initializing the subdevices.
Signed-off-by: H Hartley Sweeten <hsweeten@visionengravers.com>
Reviewed-by: Ian Abbott <abbotti@mev.co.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

For aesthetics, initialize the TIO chips before the subdevices are
allocated and initialized.

Refactor the function to initialize all the TIO chips and move it
to a better place in the driver.
Signed-off-by: H Hartley Sweeten <hsweeten@visionengravers.com>
Reviewed-by: Ian Abbott <abbotti@mev.co.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Make sure the interrupts are disabled before freeing the irq.
Signed-off-by: H Hartley Sweeten <hsweeten@visionengravers.com>
Reviewed-by: Ian Abbott <abbotti@mev.co.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Remove the unnecessary NULL pointer check.
Signed-off-by: H Hartley Sweeten <hsweeten@visionengravers.com>
Reviewed-by: Ian Abbott <abbotti@mev.co.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

For aesthetics, add some whitespace to the subdevice init and use
a couple local variables to make the code easier to follow.
Signed-off-by: H Hartley Sweeten <hsweeten@visionengravers.com>
Reviewed-by: Ian Abbott <abbotti@mev.co.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Remove enum ni_660x_constants and just #define the value.

Move all the constant #defines so they are in one place and rename
them so they are more conesistent.
Signed-off-by: H Hartley Sweeten <hsweeten@visionengravers.com>
Reviewed-by: Ian Abbott <abbotti@mev.co.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Fix the checkpatch.pl issue. Move the memory barrier to a better place.
Signed-off-by: H Hartley Sweeten <hsweeten@visionengravers.com>
Reviewed-by: Ian Abbott <abbotti@mev.co.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Fix the checkpatch.pl issues:

CHECK: spinlock_t definition without comment

For aesthetics, rename the 'soft_reg_copy_lock' to clarify what it's
used for.
Signed-off-by: H Hartley Sweeten <hsweeten@visionengravers.com>
Reviewed-by: Ian Abbott <abbotti@mev.co.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Convert this big switch into an array and refactor ni_660x_gpct_{write,read}()
functions to use the array to find the register offset.

All the TIO (GPCT) registers are included in the array except for NITIO_G0_ABZ
and NITIO_G1_ABZ. These registers only exist on the ni_pcimio m-series boards
and this driver will never read/write them.

Just in case someone adds a new entry to the enum ni_gpct_register in ni_tio.h,
add a dev_warn() for any unhandled registers.
Signed-off-by: H Hartley Sweeten <hsweeten@visionengravers.com>
Reviewed-by: Ian Abbott <abbotti@mev.co.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

The (*auto_attach) initializes all the DIO channels to a default state.
Add a comment for clarity.
Signed-off-by: H Hartley Sweeten <hsweeten@visionengravers.com>
Reviewed-by: Ian Abbott <abbotti@mev.co.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Use the comedi.h provided constants (enum ni_660x_pfi_routing) instead
of defining new ones for the output sources.

Use a switch to clarify the channel/source validation.

For aesthetics, rename the private data members 'pfi_output_selects' and
'pfi_direction_bits'.

Remove the 'min_counter_pfi_chan' and 'max_dio_pfi_chan' from
enum ni_660x_constants. The open coded values make the code easier
to follow.
Signed-off-by: H Hartley Sweeten <hsweeten@visionengravers.com>
Reviewed-by: Ian Abbott <abbotti@mev.co.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Use some local variables to clarify this function.

This (*insn_bits) function is a bit different from most comedi drivers.
Add some comments to clarify why the shifts are used.
Signed-off-by: H Hartley Sweeten <hsweeten@visionengravers.com>
Reviewed-by: Ian Abbott <abbotti@mev.co.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Add some whitespace to the Digital I/O subdevice init and add a
comment about the channels. This driver is a bit goofy, only 32 of
the 40 channels can actually be used for Digital I/Os and 32 of
them can be routed to the counters for alternate use.
Signed-off-by: H Hartley Sweeten <hsweeten@visionengravers.com>
Reviewed-by: Ian Abbott <abbotti@mev.co.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Fix the checkpatch.pl issues.
Signed-off-by: H Hartley Sweeten <hsweeten@visionengravers.com>
Reviewed-by: Ian Abbott <abbotti@mev.co.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Fix the checkpatch.pl issues.
Signed-off-by: H Hartley Sweeten <hsweeten@visionengravers.com>
Reviewed-by: Ian Abbott <abbotti@mev.co.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

This inline function is only used by the (*auto_attach). Remove it
and just use a local variable for the calculation.
Signed-off-by: H Hartley Sweeten <hsweeten@visionengravers.com>
Reviewed-by: Ian Abbott <abbotti@mev.co.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Hard-coding the subdevice order is normally a bad idea. If a new subdevice
is added, or removed, it could potentially break pretty badly.

Remove the enum and associated NI_660X_GPCT_SUBDEV() helper that hard-code
the subdevice order.

Fix the (*auto_attach) so it initializes all the subdevices without depending
on the hard-coded order.

Change the interrupt handler so that all the counter subdevices are handled
without depending on the hard-coded order.
Signed-off-by: H Hartley Sweeten <hsweeten@visionengravers.com>
Reviewed-by: Ian Abbott <abbotti@mev.co.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Fix the checkpatch.pl issues about:

WARNING: Block comments use * on subsequent lines
WARNING: Block comments use a trailing */ on a separate line
Signed-off-by: H Hartley Sweeten <hsweeten@visionengravers.com>
Reviewed-by: Ian Abbott <abbotti@mev.co.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

This BUG_ON() happens if a mite DMA channel is already requested when an
ansynchronous command is started for one of the counter subdevices.

The comedi core will only call the (*do_cmd) if the subdevice is not busy.
In this driver, the (*cancel) for the subdevice will always release any
requested mite DMA channel.

Remove the BUG_ON() which can never occur.
Signed-off-by: H Hartley Sweeten <hsweeten@visionengravers.com>
Reviewed-by: Ian Abbott <abbotti@mev.co.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Tidy up this function to fix the checkpatch.pl issues:

WARNING: Prefer 'unsigned int' to bare use of 'unsigned'

For aesthetics, remove the static const local variables.
Signed-off-by: H Hartley Sweeten <hsweeten@visionengravers.com>
Reviewed-by: Ian Abbott <abbotti@mev.co.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Rename these functions so they have namespace associated with the
driver.

Fix the checkpatch.pl issues:

WARNING: Prefer 'unsigned int' to bare use of 'unsigned'
Signed-off-by: H Hartley Sweeten <hsweeten@visionengravers.com>
Reviewed-by: Ian Abbott <abbotti@mev.co.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Rename this function to help shorten some of the long lines.

Remove the inline, let the compiler figure it out.

Change the 'unsigned' parameters to 'unsigned int' to fix the
checkpatch.pl issues:

WARNING: Prefer 'unsigned int' to bare use of 'unsigned'
Signed-off-by: H Hartley Sweeten <hsweeten@visionengravers.com>
Reviewed-by: Ian Abbott <abbotti@mev.co.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Rename this function to help shorten some of the long lines.

Remove the inline, let the compiler figure it out.

Change the 'unsigned' parameters to 'unsigned int' to fix the
checkpatch.pl issues:

WARNING: Prefer 'unsigned int' to bare use of 'unsigned'
Signed-off-by: H Hartley Sweeten <hsweeten@visionengravers.com>
Reviewed-by: Ian Abbott <abbotti@mev.co.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Remove the enums global_interrupt_{status,config}_register_bits and
add defines for the CamelCase values.
Signed-off-by: H Hartley Sweeten <hsweeten@visionengravers.com>
Reviewed-by: Ian Abbott <abbotti@mev.co.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

The BUG_ON() checks in the helper functions are not necessary. The mite
driver quiries the PCI chip to determine the number of DMA channels.
This is then used when a DMA channel is requested so the channel will
always be in range.

Convert the inline functions used to set the bits in the NI600X_DMA_CFG
register into macros. Also convert the associated enum dma_selection.
This clarifies the association with the register.

Rename the associated 'dma_configuration_soft_copies' member of the
private data to allow shorting some of the ugly long lines in the
driver.

This also fixes a number of checkpatch.pl issues about:

WARNING: Prefer 'unsigned int' to bare use of 'unsigned'
Signed-off-by: H Hartley Sweeten <hsweeten@visionengravers.com>
Reviewed-by: Ian Abbott <abbotti@mev.co.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Remove this enum and add a define for the bit.
Signed-off-by: H Hartley Sweeten <hsweeten@visionengravers.com>
Reviewed-by: Ian Abbott <abbotti@mev.co.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Reformat the multi-line comment in the kernel CodingStyle.
Signed-off-by: H Hartley Sweeten <hsweeten@visionengravers.com>
Reviewed-by: Ian Abbott <abbotti@mev.co.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Convert the inline functions used to set the bits in the NI600X_IO_CFG
registers into macros. Also convert the enum ni_660x_pfi_output_select
into defines. This clarifies the association with the register.

This also fixes a number of checkpatch.pl issues about:

WARNING: Prefer 'unsigned int' to bare use of 'unsigned'
Signed-off-by: H Hartley Sweeten <hsweeten@visionengravers.com>
Reviewed-by: Ian Abbott <abbotti@mev.co.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Rename this CamelCase struct and the associated 'registerData' variable.
Signed-off-by: H Hartley Sweeten <hsweeten@visionengravers.com>
Reviewed-by: Ian Abbott <abbotti@mev.co.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

This enum is used to define the, unused, 'direction' of each register
in struct NI_660xRegisterData. Remove the unused member, as well as
the enum.
Signed-off-by: H Hartley Sweeten <hsweeten@visionengravers.com>
Reviewed-by: Ian Abbott <abbotti@mev.co.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

All the registers are defined struct NI_660xRegisterData and they are
either 2 or 4 bytes in size. Remove the enum and just use a char member
to define the size as 2 or 4 bytes.

Simplify the ni_660x_{write,read}_register() functions and remove the
unnecessary BUG() in each.
Signed-off-by: H Hartley Sweeten <hsweeten@visionengravers.com>
Reviewed-by: Ian Abbott <abbotti@mev.co.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

This member of the struct is not used, and just takes up space. Remove it.
Instead, add the enum ni_660x_register indexes to the table to clarify, and
document, the entries.
Signed-off-by: H Hartley Sweeten <hsweeten@visionengravers.com>
Reviewed-by: Ian Abbott <abbotti@mev.co.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

The BUG_ON() in this function is unnecessary. The 'pfi_channel' will
always be in range of the subdevice 'n_chan' (NUM_PFI_CHANNELS) which
will return a valid 'reg'.

Convert the inline function into a simple macro.
Signed-off-by: H Hartley Sweeten <hsweeten@visionengravers.com>
Reviewed-by: Ian Abbott <abbotti@mev.co.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>