Introduce a generic method to write samples to the async buffer.

The number of samples is first checked against the number of samples that
would fill the async buffer. The size of each sample is determined using
the bytes_per_sample() helper. If all the samples will fit in the async
buffer they are written to the buffer using comedi_write_array_to_buffer().

This will allow converting all the comedi drivers to use a common method to
write data to the async buffer.

Since comedi_write_array_to_buffer() sets the COMEDI_CB_BLOCK event after
writing the data, those events can be removed from the drivers.

In addition, comedi_inc_scan_progress() will automatically detect the end of
scan and set the COMEDI_CB_EOS event. Those events can also be removed from
the drivers.
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 parameter is always passed as '0'. Remove the unnecessary parameter.

This allows removing the unnecessary check of the read_ptr overflow.
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 parameter is always passed as '0'. Remove the unnecessary parameter.

This allows removing the unnecessary check of the write_ptr overflow.
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 function is only used internally. Remove the export.
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 function is only used internally. Remove the export.
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 function is no longer 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>

Use comedi_buf_read_samples() to read the analog output samples from
the async buffer. That function will set the COMEDI_CB_BLOCK event
when samples are read from the async buffer.
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 comedi_buf_read_samples() to read the analog output samples from
the async buffer. That function will set the COMEDI_CB_BLOCK event
when samples are read from the async buffer.
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 comedi_buf_read_samples() to read the analog output samples from
the async buffer. That function will set the COMEDI_CB_BLOCK event
when samples are read from the async buffer. In addition, the core
will also handle the end-of-scan detection and set the COMEDI_CB_EOS
event.
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 comedi_buf_read_samples() to get the single samples from the async
buffer. The number of samples in the buffer was validated by the callers
so the error checking can be removed. In addition, the core will set the
COMEDI_CB_BLOCK event so it can also be removed.
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>

A number of drivers currently use comedi_buf_get() to read single samples
from the the async buffer. This works but the drivers have to handle the
COMEDI_CB_BLOCK and COMEDI_CB_EOS events. Converting those drivers to use
comedi_buf_read_samples() moves the handling of those events into the core.

Modify comedi_buf_read_samples() so that the async buffer is checked for
the number of full samples available. Use that to clamp the number of
samples that will be read. This makes sure that only full samples are
read from the async buffer making comedi_buf_read_samples() mimic the
action of comedi_buf_get() with the added benifit of handling the events.
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 function is only called by comedi_buf_read_samples(). Absorb 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 function is only used internally. Remove the export.
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 no longer 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>

Introduce a helper function to read the analog output samples from the
async buffer and prep the DMA.

Use comedi_buf_read_samples() to read the analog output samples from
the async buffer.
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>

Introduce a helper function to read the analog output samples from the
async buffer.

This fixes a bug in load_ao_dma_buffer(). That function was decrementing
the 'ao_count' by the number of bytes, not the number of samples, that
were read from the async buffer.

Use comedi_buf_read_samples() to read the analog output samples from
the async buffer.
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>

Introduce a helper function to load the analog output FIFO.

This fixes a bug in cb_pcidas_ao_inttrig(). That function was writing
'num_bytes' data values to the FIFO instead of 'num_points'. This results
in twice the number of data values getting written to the FIFO.

Use comedi_buf_read_samples() read the analog output samples from the async
buffer.
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 new generic method to read the analog output samples from the async
buffer.
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>

Introduce a generic method to read samples from the async buffer.

The number of requested samples is clampled to the number of samples that
would fill the async buffer. The size of each sample is determined using
the bytes_per_sample() helper. The number of bytes need are then read
from the async buffer using comedi_read_array_from_buffer().

This will allow converting all the comedi drivers to use a common method
to read data from the async buffer.

Since comedi_read_array_from_buffer() sets the COMEDI_CB_BLOCK event after
reading the data, those events can be removed from the drivers.

In addition, comedi_inc_scan_progress() will automatically detect the end of
scan and set the COMEDI_CB_EOS event. Those events can also be removed from
the drivers.
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 comedi core calls all the subdevice (*cancel) functions before it does
the driver (*detach). There is no reason to call the board reset function
in the (*detach).
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 analog input (*cancel) stops the async command. There is not reason to also
call the board reset 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>

Move the apci3120_reset() call in the (*auto_attach) to ensure that the board
is reset before hooking up the interrupt.
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>

It appears that this subdevice is actually broken. The subdevice functions
all violate the comedi API and, it they did work, it appears they would
break the analog input async commands.

For now just tidy up the subdevice init.

Add the SDF_READABLE subdev_flag, this subdevice provides an (*insn_read).

Fix the subdevice 'maxdata' initialization. The timer is 24-bits.

Remove the 'len_chanlist' initialization. This member is only used by
subdevices that support async commands.

Also remove the 'range_table' initialization. The core will set it to
range_unknown which is appropriate.
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 white space to the analog input subdevice
initialization.

The async command support in this driver is pretty messed up. I'm not
really sure if it actually works yet. Luckly this driver was not
setting the SDF_CMD_READ subdev_flag so the command support was not
actually getting enabled.

The command support only works if an interrupt is available. Refactor
the subdevice init so that the command support is only hooked up if
we have an interrupt. For now always disable the command support. Once
the driver has been cleaned up the command will be enabled.
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 white space to the digital output subdevice
initialization.

Remove the subdev_flags SDF_GROUND and SDF_COMMON. These only apply to
analog subdevices. Also remove the unnecessary SDF_READABLE subdev_flag.
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 white space to the digital input subdevice
initialization.

Remove the subdev_flags SDF_GROUND and SDF_COMMON. These only apply to
analog 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, add some white space to the analog output subdevice
initialization.
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_subdevice 'readback' member and the core provided (*insn_read)
function to provide radback of the analog output channels.
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 comedi core expects (*insn_write) functions to write insn->n values and
return the number of values written or an errno. This function currently
returns insn->n but it only writes a single data value.

Fix the function to work like the core expects.

There are two registers used to update the analog outputs. Offset 0x08 is
used to update channels 0-3 and offset 0x0a to update channels 4-7. Bits
14 and 15 in each register set the mux to select which channel to update.
The lower 14 bits are the value used to set the DAC.

For aesthetics, tidy up the defines used for the register offsets and bits
in the 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 apci3120_reset() function is called at the end of the (*auto_attach) and
with the (*detach) is called to unload the driver. Part of this function
resets all the analog output channels to 0V.

There are two problems with this.
1) Only the APCI-3120 has analog outputs, the APCI-3001 does not.
2) The DA_READY bit in the status register needs to be checked before each
   write to update the analog outputs. It's unknown what the DA_READY bit
   does on the APCI-3001 board.

Just remove the analog output reset to avoid any problems.

Also, remove the unnecessary udelay() in apci3120_reset().
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 DA_READY bit (0x0001) in the status register needs to be set before the
analog outputs can be updated.

Use comedi_timeout() to prevent a deadlock condition while waiting for the
DA_READY bit to get set.
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 holds the start address of PCI BAR 1 that is
used to access the board registers. The start address is also stored in the
comedi_device 'iobase' member. Use that instead and remove the private data
member.

Remove some unnecessary casts when reading the board 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>

This member of the private data holds the start address of PCI BAR 0 that is
used to access the AMCC registers.

Rename this CamelCase member and fix its type.

Remove an unnecessary local variable, 'ui_Tmp', in apci3120_cyclic_ai() and
tidy up the rest of the local variable declarations.
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 holds the start address of PCI BAR 2 that is
used to access the AMCC Add-On registers.

Rename this CamelCase member and fix its type.
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 board registers are accessed using the bse address devpriv->iobase not
devpriv->i_IobaseAmcc. Fix a read in apci3120_interrupt() that is using the
wrong base address.
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 AMCC registers are accessed using the base address devpriv->i_IobaseAmcc
not devpriv->i_IobaseAddon. Fix a couple writes in apci3120_interrupt_dma()
that are using the wrong base address.
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 by the driver. 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>

The analog outputs of the APCI-3120 have a 14-bit bipolar range and use straight
binary values (0x0000 = -10V to 0x3fff = +10V) to set the outputs. This driver
tries to fake a unipolar range by munging the comedi data values to only output
0V to +10V signals (0x0000 to 0x1fff values are munged to 0x2000 - 0x3fff).

This causes problems with the comedilib API functions that convert between comedi
values and physical values. It's also possible for the user to pass a data value
that would get munged and actually change the wrong channel.

Fix this by change the subdevice range_table to range_bipolar10 and remove the
munging in apci3120_ao_insn_write().
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 struct str_AnalogReadInformation is not used in this driver.
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>

For aesthetics, wrap the DMA buffer information in a struct.
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>