This is needed because another project builds this file for the lm32,
which is big-endian.  This makes no effect on the PC.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

In order to avoid spurious interrupts from previous running, disable all
interrupts before registering the IRQ handler.
Signed-off-by: Federico Vaga <federico.vaga@cern.ch>

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

This prevent confusion while reading the code because, for example, a
question: why initialize both timer and tasklet if we can use only one, and
we cannot change the configuration at run-time?

fd_irq_init:
    setup_timer only if the user requires timer mode, otherwise initialize tasklet
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>

Explain why the handler doesn't clear the interrupt in EIC_ISR even if
the documentation regarding this register is saying something different.
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>
Acked-by: Alessandro Rubini <rubini@gnudd.com>

Signed-off-by: Federico Vaga <federico.vaga@cern.ch>
Acked-by: Alessandro Rubini <rubini@gnudd.com>

The Kbuild writesin 'src' the correct path to the source code automatically.
This avoid us to explicitly write 'M' variable
Signed-off-by: Federico Vaga <federico.vaga@cern.ch>
Acked-by: Alessandro Rubini <rubini@gnudd.com>

Signed-off-by: Federico Vaga <federico.vaga@cern.ch>
Acked-by: Alessandro Rubini <rubini@gnudd.com>

these are not a kernel tools, so we do not need use kbuild enviroment. Use a custom
environment to avoid confusion. Moreover, add EXTRACFLAGS to allow people to cross
compile between 32bit and 64bit on x86 architecture

    make EXTRACFLAGS="-m32"
    make EXTRACFLAGS="-m64"
Signed-off-by: Federico Vaga <federico.vaga@cern.ch>
Acked-by: Alessandro Rubini <rubini@gnudd.com>

Without the 'gitmodules' rule you cannot automatically clone the submodules
Signed-off-by: Federico Vaga <federico.vaga@cern.ch>
Acked-by: Alessandro Rubini <rubini@gnudd.com>

Actually, it reverts c92af50f and
pick the files directly from the fmc-bus project
Signed-off-by: Federico Vaga <federico.vaga@cern.ch>
Acked-by: Alessandro Rubini <rubini@gnudd.com>

Reason: zio block the channel if the raw_io function returns an error (possibly
a bug in zio user trigger handling).

This fixes a potentially very dangerous behaviour: previously the output was defaulted to channel 1 and
running fmc-fdelay-pulse with no parameters by unaware user would result with generation of a 10 Hz
waveform on output 1.

By default, outputs are disabled and programmed to 600 ns delay, 250 ns width, 1x repeat and 1 us period. This patch
allows safe read-modify-write operations.

This is a hotfix for the COHAL/FESA guys: the current driver/lib does not decouple
enabling/disabling the output and setting its mode (pulse/delay). Previously, disabling the output
was leading to loss of its configuration. This patch allows to retain it under the condition that the
output is only switched between DISABLED and DELAY modes.

kernel/fd-zio.c: fix incorrect readout of repeat count if output configured in infinite repetition mode

submodules: updated zio to include the __devinit patch (fixes build issues on newer kernels with CONFIG_SPI enabled)

The should pick other repositories from ../ only if both use and them
are submodules of ../ .

Only checking that ../zio is a directory is not enough, because it
can be a separate clone (old version, for example) or worse an
intermediate directory that doesn't host headers.

A user reported painful build errors for this reason.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

This directly affects tools/fmc-fdelay-status, although the tool is
not changed.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

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

TDC user offset should be applied to all timestamps. The code, however,
didn't do that properly, and fast_tdc mode was missing it.

Now user_offset is applied where the hardware offset is, like it should
have since time zero.

This should have no effect, as user_offset is an option nobody used
so far, as far as we know.
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 should in a way fix the missing error checks in parse_time,
without forcing us to rewrite parsing with more stringent checks.

Times are reported before activating rather than after each parse
action, so the user can see what 1mhz or pps turn out to be, as
well as the default values.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>