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>

Several changes to make it compile under the new abstraction

- probe and remove are related to an fmc_driver structure. Thus
the fd-spec.c is obsolete.  Still, the driver is bound to the spec
and as such it refuses to initialize under a different carrier.

- we don't have "spec" in the structure but "fmc"

- i2c referred to pdev, but has been cleaned up

- fd-zio unfortunately refers to pdev to build the local ID.

Still a lot of details are to be fixed.
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 describes the new way of the code, but code itself is only
modified by the later patches.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

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

This is a quick and dirty fix adding a delay of 500ms
after each call to fdelay_config_pulse, to avoid the
synch problem observed in zio whilst 2.6.24.7-rt27
clients wait for a fix.

In fdelay_config_pulse_ps, the pulse length is given in
picoseconds but, for some reason, computation of the
pulse end time by addition of that length was commented
out.

This requires fdelay_add_ps to be implemented. There is a
similar function in fine-delay.h, but it is only accessible
to kernel-space code and is particular to zio attributes.

New temperature attribute added for reading in fd_zio_info_get().

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

take it or drop it

At CERN controls installations, devices of the same class are
enumerated by a statically defined number known as LUN (logical
unit number), that maps into geographical information via
database entries. These are mapped through userspace programs
and stored as tags in device node names.

This patch adds a function to exploit this numbering

Under CERN CO current standard compiler
gcc (GCC) 4.1.2 20080704 (Red Hat 4.1.2-52)
a long, 64 bit division is generated even if the divisor is
a power of 2 (sigh). So we need yet another div_u64_rem.