Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

By referring to our spec-sw subdirectory, we ensure versions are matched.
Also, we'll later need to include "spec.h", for carrier-specific DMA
code: we'd better not refer to ../spec-sw by default.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>
Acked-by: Michel Arruat <michel.arruat@cern.ch>

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

to not become crasy while grepping xx_exit or xx_init function (sorry...)

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

This is meant to fire the FMCADC_EDISABLED error when reconfiguring
and ongoing acquisition for which somebody is already waiting.

Every function that is sable to wait for a timeout now may return -1
with FMCADC_EDISABLED. The new zio commits support that by returning
POLLERR when using poll(2) over a control file descriptor and the
trigger is disabled.  A non-zio driver may support something similar.

Thus, an application may disable the trigger to unlock with error
any thread waiting for an input block.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

This adds a flushing option when the device is opened and when a new
acquisition is requested..  The code uses poll(2) and allows for
POLLERR to be returned, so it is prepared for the new zio semantics of
poll(2) for control files.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Release invalid blocks on DMA error
Signed-off-by: Federico Vaga <federico.vaga@gmail.com>

It is fair and necessary for peaceful coexistence of the driver
with other peers like FD and TDC.

This avoids page-split dma_items in a very dirty way

Debugging DMA error required to spot what next dma_item
adresses were being communicated to the DMA core; the
trace message has been deferred accordingly

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

This set of commits completely reviews and simplifies the API, fixing
the sequence of allocation/fill/deallocation for buffers.

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Signed-off-by: Federico Vaga <federico.vaga@gmail.com>

Signed-off-by: Federico Vaga <federico.vaga@gmail.com>

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Signed-off-by: Federico Vaga <federico.vaga@gmail.com>

At open time, users are expected to pass the total number of samples
they expect to need in their buffer, not the total number of bytes.
The name now matches current use in fald-simple-acq.
Signed-off-by: Federico Vaga <federico.vaga@gmail.com>
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

We need mapaddr and mapsize for munmap() at buffer_release time.
And we need the device pointer (not drivername) in order to route
the get_tstamp method. drivername, if needed, is in the device anyways.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

SQUASH proto
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>