For aesthetics, move this function out of the async command support
code.

For safety, the INT request (end-of-conversion flag) should be cleared
before doing each conversion and after the final data sample is read.
The driver currently does not do this.

Refactor the function a bit so it's more like the pcl818 and pcl816
drivers and use common code to clear the flag for a timeout and after
the last sample.

Do a bit of other tidying up during the move.
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 this function out of the async command support
code.

For safety, the INT request (end-of-conversion flag) should be cleared
before doing each conversion and after the final data sample is read.
This driver does that but it's a bit awkward with the initial clear being
outside the for loop that reads the samples.

Refactor the function a bit so it's more like the pcl818 driver and we
can use common code to clear the flag for a timeout and after the last
sample.

Do a bit of other tidying up during the move.
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 this function out of the async command support
code.

For safety, the INT request (end-of-conversion flag) should be cleared
before doing each conversion and after the final data sample is read.
This driver already clears the flag before starting a conversion but it
does not clear the flag after the final sample.

Refactor the function a bit so that the flag is cleared for a conversion
timeout and after the last sample.

Do a bit of other tidying up during the move.
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, tidy up the analog output subdevice code.

Change the regsiter map defines to simplify the code.

Move the analog subdevice support functions out of the analog input
support functions. Tidy them up a bit during the move.

Reverse the logic of the subdevice init and add some whitespace.
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, tidy up the analog output subdevice code.

Change the regsiter map defines to simplify the code.

Move the analog subdevice support functions out of the analog input
support functions. Tidy them up a bit during the move.

Add some whitespace to the subdevice init.
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 this function to follow the pcl812 and pcl818 drivers
better.

Remove the commented out cut-and-paste code from the pcl818 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>

For aesthetics, tidy up the digtial subdevice code.

Rename the register map defines for the digital input and output ports.

Move the two digital subdevice (*insn_bits) functions out of the analog
support functions. Tidy them up a bit during the move.
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 PCL-816 boards have 16 digital inputs and 16 digtial outputs. Add the
subdevice support for these channels.

Allocate additional subdevice space to the analog outputs. This code is
not currently in the driver so mark the subdevice as COMEDI_SUBD_UNUSED.
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, tidy up the digtial subdevice code.

rename the register map defines for the digital input and output ports.

Move the two digital subdevice (*insn_bits) functions out of the analog
support functions. Tidy them up a bit during the move.

Add some whitespace to the subdevice init.
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 DMA, FIFO, and EOC interrupt handler functions, that are called by
the _real_ interrupt function, always return IRQ_HANDLED. Change the
return type for these functions to void and move the final return to
the real interrupt function.

Change the parameters to the handler functions to the comedi_device and
comedi_subdevice pointers.

At some point in the handler functions the interrupt request is cleared.
Move this to the real interrupt function.

Also at some point in the handlers, comedi_event() is called to pass any
events to the comedi subsystem. Move this to the real interrupt function
also.

For aesthetics, and to clarify the code, rename the interrupt function and
the handler functions.
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 driver only supports interrupt driven DMA transfers for async
command support. Absorb the interrupt_pcl816_ai_mode13_dma() helper
into the interrupt handler to clarify the code.

At some point during the interrupt, comedi_event() is called to pass
any events to the comedi subsystem. Move this to the interrupt handler.
Add a comedi_event() call to pcl816_ai_poll() due to the removal of
the call from transfer_from_dma_buf().

For aesthetics, rename the interrupt function so it has namespace
associated with 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>

The DMA and EOC interrupt handler functions, that are called by the
_real_ interrupt function, always return IRQ_HANDLED. Change the
return type for these functions to void and move the final return to
the real interrupt function.

Change the parameters to the handler functions to the comedi_device and
comedi_subdevice pointers.

At some point in the handler functions the interrupt request is cleared.
Move this to the real interrupt function.

Also at some point in the handlers, comedi_event() is called to pass any
events to the comedi subsystem. Move this to the real interrupt function
also.

For aesthetics, and to clarify the code, rename the interrupt function and
the handler functions.
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 private data is not needed. We can determine the
interrupt mode based on the 'dma' and 'usefifo' members.

Refactor the interrupt handler and the (*cancel) function to not use
the 'ai_mode' and remove the setting of this member in the (*do_cmd)
heler functions.
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 private data is not needed.

Refactor the interrupt handler and the (*cancel) function to not use
the 'int816_mode' and remove the setting of this member in the (*do_cmd)
heler function.
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>

If an async ai command is not running or the ai_mode is 0 the interrupt
routine doesn't do anything other than spew some noise and clear the
interrupt request in the hardware.

Because this driver is manually attached, the "premature interrupt"
check in the interrupt handler should never happen. The interrupt is
only hooked up during the attach and it's released during the detach.

Combine these checks so that the interrupt function exits quick if it
has nothing to handle. Remove the noise.
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>

If an async ai command is not running or the int816_mode is 0 the
interrupt routine doesn't do anything other than spew some noise and
clear the interrupt request in the hardware.

Because this driver is manually attached, the "premature interrupt"
check in the interrupt handler should never happen. The interrupt is
only hooked up during the attach and it's released during the detach.

Combine these checks so that the interrupt function exits quick if it
has nothing to handle. Remove the noise.
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>

Because this driver is manually attached, the "spurious interrupt"
check in the interrupt handler should never happen. The interrupt is
only hooked up during the attach and it's released during the detach.

Leave the check but remove the noise. Also make sure to clear the
interrupt request in the hardware.
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>

When using DMA, the async command (*cancel) operation is delayed until the
current DMA transfer is complete. When the DMA transfer finishes the interrupt
routine detects this and calls the (*cancel) again which should then cancel the
async command.

The current logic does not work when cmd->stop_src == TRIG_NONE. In this case
the (*cancel) function keeps delaying the cancel. The actual DMA does stop
because the DMA handler is not called to setup the next transfer but the
(*cancel) code is never executed.

Rename the 'irq_was_now_closed' flag in the private data to 'ai_cmd_canceled'
to clarify what it is.

Only set the flag in the (*cancel) when a DMA transfer is running and the
async command has not already been canceled. The interrupt routine then does
not need all the extra checks.
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 flag in the private data is set when an async command is canceled with
the ai (*cancel) operation. Rename the flag to 'ai_cmd_canceled' to clarify
its use.

Move the check for the flag in the interrupt handler. If the async command
was canceled there is no reason to handle the interrupt. Just clear interrupt
and return.

Also, make sure to clear the interrupt when the device is not attached.
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>

If an async command is not running the (*cancel) function doesn't do
anything. Exit the function early if this is the case.

This allows reducing the indent level in the rest of the function.

Also, move the setting of the 'irq_was_now_closed' to the only place
wher it actually gets left set to 1 and remove the goto.
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>

If an async command is not running the (*cancel) function doesn't do
anything. Exit the function early if this is the case.

This allows reducing the indent level in the rest of the function.
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 DMA, FIFO, and EOC interrupt handlers all have common code that
checks for channel dropout when running an async command.

Factor this common code into a helper function.

Only return the sample if the channel is valid. The EOC handler was
previously returning the value then checking for channel dropout.
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>

To clarify the code, introduce a helper function to read the analog
input data sample from the FIFO and optionally return the channel
that the sample was for. The channel is used to check for dropped
samples.
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 A/D FIFO uses two registers to get each analog data sample.
PCL818_FI_DATALO is the LSB of the data and PCL818_FI_DATAHI is
the MSB of the data. The current define for PCL818_FI_DATAHI is
incorrect and results in the LSB getting read twice.
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 DMA, FIFO, and EOC interrupt handlers all have common code that
bumps the counters in the private data that keep track of what channel
is being sampled next and when all the data has been sampled.

Factor this common code into a helper function.

Make sure the COMEDI_CB_BLOCK event is set when advancing to the next
channel so that when the comedi_event() is finally called the core will
wake the async_queue to return the analog input data.

Also, make sure the COMEDI_CB_EOS event is set at the completion of each
scan.
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>

Factor out the code that bumps the counters in the private data that keep
track of what channel is being samples next and when all the data has been
sampled.

Don't clear the events in the driver. The comedi core will clear the events
at the end of the comedi_event().

Make sure the COMEDI_CB_BLOCK event is set when advancing to the next
channel so that when the comedi_event() is finally called the core will
wake the async_queue to return the analog input data.

Also, make sure the COMEDI_CB_EOS event is set at the completion of each
scan.
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 DMA and EOC interrupt handlers all have common code that bumps the
counters in the private data that keep track of what channel is being
sampled next and when all the data has been sampled.

Factor this common code into a helper function.

Don't clear the events in the driver. The comedi core will clear the events
at the end of the comedi_event().

Make sure the COMEDI_CB_BLOCK event is set when advancing to the next
channel so that when the comedi_event() is finally called the core will
wake the async_queue to return the analog input data.

Also, make sure the COMEDI_CB_EOS event is set at the completion of each
scan.
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 driver only supports async commands if a DMA channel is available. Modify
the subdevice init so that the command support only gets hooked up if we can
do DMA.
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>

According to the users manual, when using interrupt (non-DMA) transfers
this hardware generates an interrupt at the start of a conversion. This
requires the interrupt routine to busywait until the end-of-conversion.
It appears this was unreliable and interrupt only async command support
was removed at some time. Async command support is still available when
DMA is used.

Remove the unreachable interrupt only handler code.
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 interrupt is only generated by the hardware at the completion of
an A/D conversion. Because of this the sanity check to make sure that
the A/D conversion is complete and data is available is probably
unnecessary but it doesn't hurt anything.

The busywait loop is a different issue. Interrupt routines should not
busywait. That's just mean...

Remove the bustwait and use pcl818_ai_eoc() to check for the end-of-
conversion.
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 interrupt is only generated by the hardware at the completion of
an A/D conversion. Because of this the sanity check to make sure that
the A/D conversion is complete and data is available is probably
unnecessary but it doesn't hurt anything.

The busywait loop is a different issue. Interrupt routines should not
busywait. That's just mean...

Remove the bustwait and use pcl812_ai_eoc() to check for the end-of-
conversion.
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 private data is not used. Remove it.
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 private data is only used in the initial dma setup.
Refactor that code to not need it and remove the member.
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>

To assist in cleaning up the the rest of the driver, factor the initial
dma setup and the next dma buffer setup into helper functions.
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>

To assist in cleaning up the the rest of the driver, factor the initial
dma setup and the next dma buffer setup into helper functions.
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>

To assist in cleaning up the the rest of the driver, factor the initial
dma setup and the next dma buffer setup into helper functions.
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 private data is not needed. We can check the stop_src
to determine this information:

  neverending_ai -> cmd->stop_src == TRIG_NONE
  !neverending_ai -> cmd->stop_src == TRIG_COUNT
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 private data is not needed. We can check the stop_src
to determine this information:

  ai_neverending -> cmd->stop_src == TRIG_NONE
  !ai_neverending -> cmd->stop_src == TRIG_COUNT
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 private data is not needed. We can check the stop_src
to determine this information:

  ai_neverending -> cmd->stop_src == TRIG_NONE
  !ai_neverending -> cmd->stop_src == TRIG_COUNT
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>

Some of the PCL-818 compatible boards have a FIFO that can be used
when running analog input async commands. This FIFO increased the
resource range used by the driver.

Request the correct resources used by the board even if the FIFO is
not used. This prevents another driver from trying to use the I/O
space.

Modify the attach so that the 'usefifo' flag is only set if we have
and IRQ, the board has a FIFO, and the user wahts to use it.

Fix pcl818_reset() so that the FIFO is always flushed and disabled
if it exists.
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>