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

The serdes does not lock if the ADC chip is not configured before. We
did not need this before because we were not changing the ADC chip
operational mode, but now we ask the chip to use a different number of
bits.
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 goes against what described in the ZIO framework, but it is
actually a source of problems because the framework is largely
unmaintained.

With this patch the driver play only with raw values. It will be part
of the library duty to propose a different convention.
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>

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

The data coming out from the FMC-ADC-110M FPGA is always using the
little-endian byte order. This means that we need to fix the byte
order only on big-endian CPUs.

This allows to improve performances on SVEC-based designs since there
is no need anymore to fix the endianness in software becuase it is
already done in hardware.
Signed-off-by: Federico Vaga <federico.vaga@cern.ch>

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

`get_seconds` disappeared in kernel 5.11 and onwards,
it is now required to call `ktime_get_real_seconds`.
Signed-off-by: Gwenhael GOAVEC <gwenhael.goavec@femto-st.fr>
Signed-off-by: Mathis MARION <mathis.marion@grenoble-inp.org>

`setup_timer` is no longer available in kernel 4.15 and onwards,
in favor of `timer_setup`.
Before, we could pass an argument to the callback function as an
unsigned int (generally used with a pointer cast), but now the timer
itself is passed and we retrieve the container struct with a new
macro `from_timer` which is uses `container_of`.
The flags parameter can be set as 0 for the same behavior.
Signed-off-by: Gwenhael GOAVEC <gwenhael.goavec@femto-st.fr>
Signed-off-by: Mathis MARION <mathis.marion@grenoble-inp.org>
Signed-off-by: Federico Vaga <federico.vaga@cern.ch>

The signature for `__sg_alloc_table_from_pages` changed in
kernel 5.10 and onwards.
Two agruments are added which can be passed as NULL and 0
for the same behavior as before, except for error checking,
which was addressed in this patch.
Signed-off-by: Gwenhael GOAVEC <gwenhael.goavec@femto-st.fr>
Signed-off-by: Mathis MARION <mathis.marion@grenoble-inp.org>

The return value of debugfs_create_regset32() was not really used, if
not to inform the user that a debugfs file was missing due to an
error.

After Linux 5.8, the return value disappeared, so instead of keeping
pre-processor conditions, just align with the best practices of the
latest versions.
Signed-off-by: Federico Vaga <federico.vaga@cern.ch>

The signature for `debugfs_create_regset32` changed from
returning a `struct dentry *` to `void`.
Error checking is no longer possible.
Signed-off-by: Gwenhael GOAVEC <gwenhael.goavec@femto-st.fr>
Signed-off-by: Mathis MARION <mathis.marion@grenoble-inp.org>

`unsigned int ioread32(void __iomem *)` became
`unsigned int ioread32(const void __iomem *)` in
kernel 5.8 and above.
Changed struct fa_memory_ops accordingly.

Added missing include for linux/io.h
Signed-off-by: Gwenhael GOAVEC <gwenhael.goavec@femto-st.fr>
Signed-off-by: Mathis MARION <mathis.marion@grenoble-inp.org>

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>

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 goes against what described in the ZIO framework, but it is
actually a source of problems because the framework is largely
unmaintained.

With this patch the driver play only with raw values. It will be part
of the library duty to propose a different convention.
Signed-off-by: Federico Vaga <federico.vaga@cern.ch>

This new bit starts and enable the serdes delay mechanism