With this "make modules_install" at top level works. It is not really
used much, but it makes sense to make it work as advertised.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Signed-off-by: Federico Vaga <federico.vaga@cern.ch>

Signed-off-by: Federico Vaga <federico.vaga@cern.ch>

Signed-off-by: Federico Vaga <federico.vaga@cern.ch>

in_atomic() function should not be used inside a driver, and it does not
guarantee atomic context detection. Following the comment from the
kernel source

/*
 * Are we running in atomic context?  WARNING: this macro cannot
 * always detect atomic context; in particular, it cannot know about
 * held spinlocks in non-preemptible kernels.  Thus it should not be
 * used in the general case to determine whether sleeping is possible.
 * Do not use in_atomic() in driver code.
 */
Signed-off-by: Federico Vaga <federico.vaga@cern.ch>

Signed-off-by: Federico Vaga <federico.vaga@cern.ch>

Signed-off-by: Federico Vaga <federico.vaga@cern.ch>

Signed-off-by: Federico Vaga <federico.vaga@cern.ch>

This patch mask the user value before write it on the hardware register.
If the value does not fit the mask it will be printed a warning message.
Signed-off-by: Federico Vaga <federico.vaga@cern.ch>

Blocks were already released by doing ZFA_STOP, which calls
trigger_abort_disable(), which calls our trigger abort handler, which
releases the blocks
Signed-off-by: Federico Vaga <federico.vaga@cern.ch>

Signed-off-by: Federico Vaga <federico.vaga@cern.ch>

Signed-off-by: Federico Vaga <federico.vaga@cern.ch>

With the new code organization the auto-start options must be moved
after release the ZIO_CSET_BUSY flags. Otherwise, the ZIO core will loop
forever tring to arm a trigger.
Signed-off-by: Federico Vaga <federico.vaga@cern.ch>

Trigger and channel signals are not going through the
same path on the board and trigger is faster.
Trying to sample the trigger itself by connecting
it to a channel, one can see a delay of 20ns between trigger and
its sampling. The compensation is done by delaying the trigger by 2
sample tick (2*10ns)

The new firmware (3.0) increase decimation register width
from 16 to 32 bits

Signed-off-by: Federico Vaga <federico.vaga@cern.ch>

This patch is temporary at it should be reverted in the future when a
better alternative is ready
Signed-off-by: Federico Vaga <federico.vaga@cern.ch>

Signed-off-by: Federico Vaga <federico.vaga@cern.ch>

Signed-off-by: Federico Vaga <federico.vaga@cern.ch>

Signed-off-by: Federico Vaga <federico.vaga@cern.ch>

Checking trigger parameter at each single change may lead
to detect error on intermediate configuration which is not the final
one. The check is done when the user starts an acquisition

Application using the ADC driver may run with real-time scheduling and
high priority.
The default shared queue is running with non real-time scheduling
policy.
ADC crates its own workqueue, that one can adjust in terms
of scheduling policy to fulfill application's real-time constraints.

The IRQ handler defers to a workqueue the execution of the DMA.
The workqueue can be delayed (depending of workqueue scheduling policy)
and it may happen that meanwhile a trigger start/stop sequence occurs
to serve a trigger configuration change. Data blocks exist but are
inconsistent with the current acquisition.
To avoid this concurrency issue CSET_BUSY flag is raised
in the IRQ Handler.

This is work of Michel Arruat, recommitted by Alessandro.  These
svec-specific files match what the previous commit did for spec code.
We are thus using the new register table, sdb support, and
carrier-provided VIC support for interrupts.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

This is a major change, work of Michel Arruat, then re-committed as a
single commit by Alessandro.  The original development happened mainly
on the SVEC carrier, and support for SPEC (PCI) was added back at a
later time.  I (Alessandro) chose to pick individual changes one at
a time, for better documentation and later "git blame", up to this point.

Now, this is a big commit that completely changes the driver,
mainly in the following places:

  - split carrier-specific from generic code, like the TDC driver is now
    (this is mainly the dma-specific and irq-specific code)
  - use of SDB to get base addresses (with new fmc-util file)
  - use the VIC core to manage interrupts, relying on carrier support
  - change the field-desc table for the better
  - add carrier-specific init, reset, exit
  - split irq and adc-specific stuff out of zio-drv.c
  - DMA data and interrupt is now different: the timestamp is at end-of-data,
    thus we don't have the "trigger-happened" interrupt any more
  - use a work queue for interrupt management, instead of a tasklet, this
    allows sleeping, which is needed by this code base
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>

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>

While the headers are not used yet the next commits will include them.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

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