Now, we use the offs parameter as the address
to be written by the protocol. This fixes the
protocol semantics, and makes it easier to support
new protocols.

Some chips have 8, 16 or 24-bit (and more )address
size definitions. So, allow the higher layer to
specify its size.

In this way, the only option to get the operations
table for a particular "protocol" is to instantiate
the appropriate class and use the *_get_ops ()
function.

With this, we can be safe that the user will always
pass the pointer ops that is inside the "protocol"
structure (along with its options).

Now, the same information needs to be read from
the appropriate structure through the access methods.

The idea here is to have SPI/I2C-obj objects
to customize the objects dynamically, instead
of having a "flags" parameter that needs knowledge
of the inner protocol.

Now, we pass the pointer to the operation
we want to use it, instead of the enum.

This will make it easier to declare a new operation
and use it. In the previous way, we had to manually
add the operation to the enumeration.

Now, we pass the pointer to the operation
we want to use it, instead of the enum.

This will make it easier to declare a new operation
and use it. In the previous way, we had to manually
add the operation to the enumeration.

As of now, devices that do not support SDB will fail,
as it does not implement a timeout mechanism.

For instance, Ethernet devices will wait forever to
receive the SDB magic number, but this will never happen.
So, avoid creating SDB in the first place until we have
a proper timeout/failover mechanism.

This HALT bit was unused. So, we cahnged its meaning
to "print without prefix/headers", which is useful
for more complex printing, like SDB records.

For BE devices we can use the new SDB initialization.
However, FE devices do not have SDb implemented.
So, they rely on the static spawning.

These addresses are not needed anymore,
as they are already read from SDB and passed
to the SMIO as part of the base address.

This is just to ensure that the FMC250M SMIO will be
initialized. This is temporary until we figure it out
a way to do this properly: signaling that the SMIO
has finished initializing.

All of the SMIOs are now spawned by the SDB
engine. So, if we want to avoid spawning a
known module or something else, we must
create a way to do so. Like a config file
supporting blacklisted modules.

It was missing the SDB prefix address (got from
LLIO) and the "first" offset from the current SDB device.

Now, these addresses are read from SDB

This will be used as a prefix address for read/writing
to the SDB memory space. If an endpoint does not implement
or support SDB, it should set SDB address to 0x0.

We need to pass the actual inst_id that was used to
generate the SMIO identifier. Previously, we were
passing the incorrect value.

We could have generated an invalid key. So, we
must check this before proceeding with spawning
an SMIO with this key.

This adds a first try to the long-waited
functionality of spawning the SMIOs based
entirely on the SDB parsing.

Now, we create/remove the SDBFS on devio_new ()
and devio_destroy ()

For now, only LLIO with PCIe endpoints are
supported.