to_TimeInternal and from_TimeInternal now return int (0 if no errors, -1 in
case of error
normalize_TimeInternal is now declared as static (was not used externally of
arith.c file)

previously commented "st_com_add_foreign" header is now removed (not used
externally of common-fun.c)

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

This change is needed to avoid requiring libgcc in freestanding
environments.

The div64 function I copied from the kernel only works with unsigned
numbers, but arth.h had them signed. I'm not sure at this point
whether PTP uses negative nanoseconds in this conversion, so I added a
diagnostic message. If it triggers, I can look for a solution, but I
suspect it doesn't happen.

Please note that a 64-bit arch may define the function as an inline
that does the calculation with C operators, but currently
arch-gnu-linux is used for both 32-bit and 64-bit hosts.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

This is the 64-bit unsigned division and remainder from the Linux
kernel, for 32-bit host systems.

The new lib/ directory can be used by archs to pick files, and that's
already exemplified by arch-gnu-linux and arch-bare-linux.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Maybe we should consider some generic lib, as most of this can be
shared across freestanding archs (we must check whether other
freestanding places are really so lib-less os they have these minimal
functions for example from newlib)
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>

This allows to uncomment the original DBGV in ptp code, until they are
turned to serious diagnostic messages. Then these two functions should
disappear, in my opinion.

Most DBGV use no arguments, so the default function is smaller; and
printf-like DBGV can be turned to DBGV_ARGS instead.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

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

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

I removed the trailing "interval" in all of them but announce, that
has both receipt and interval. This shortens some almost-80-cols
lines.  I'm not sure it's the right thing to do, but timer names are
unreadable anyways.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

the "ppi" object instance should consistently be the first argument of
functions that receive it. Besides being more consistent, it's smaller
and faster too (on processors with more than 4 registers, I mean).
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Boolean is braindead. We decided to leave them in the mandated
structures by now (until we discover they are not really mandated) but
not use them ourselves. This "found" variable wanted to be int, so be
it int.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>
Acked-by: Aurelio Colosimo <aurelio@aureliocolosimo.it>

This only applies checkpatch stuff, but I took the freedom to
add tabs to the stuctures in ieee1588_types.h, to make them more
readable.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

char may be signed or unsigned, so I'd better be safe here.
Meanwhile, I added some tabs to better line up the typedef stanza.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

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

this applies checkpatch-dictated fixes, with no functional change.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

These remove style warnings and errors, mostly mine. External files
(such as hw files in arch-spec and syscall stuff in bare-linux) have
not been fixed and should not be, in my opinion.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

All other state-* files refer to one state of the fsm, so rename common
functions accordingly
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

All files called state-*.c are individual states: rename accordingly
to preven confusion.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>
Acked-by: Aurelio Colosimo <aurelio@aureliocolosimo.it>

All other state-* files are states, so the table should use a
different name. "default" is unneeded, since this is the default protocol.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>
Acked-by: Aurelio Colosimo <aurelio@aureliocolosimo.it>

While the code layout comes from ptp-proposal, this is a different
package, now called pptp for "Portable PTP".
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>
Acked-by: Aurelio Colosimo <aurelio@aureliocolosimo.it>

<arch/arch.h> already defined the four network converters, so this
removes pp_htons and the other ones.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>
Acked-by: Aurelio Colosimo <aurelio@aureliocolosimo.it>

This fixes the "ppi->ch.fd" using the two descriptors
in the new net_path structure. It also istantiates a static
net_patch to avoid malloc in the freestanding environment.

The program is not linking because some functions called by
the protocol are still missing in this architecture.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Since we have two file descriptors (channels) in pp_instance, but current
code opens one socket only, use it in both places as a temporary measure.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>
Acked-by: Aurelio Colosimo <aurelio@aureliocolosimo.it>

This removes <dep/dep.h> by renaming it to <pproto/lib.h>, since those
functions are library ones, though pp_ prefixed.

<pproto/pproto.h> includes <pproto/lib.h>, since they are always needed,
and C files are fixed by not including <dep/dep.h>. Sometimes I included
<pproto/diag.h> to fix some compile warnings for implicit declaration of
pp_printf.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>
Acked-by: Aurelio Colosimo <aurelio@aureliocolosimo.it>

With this patch, the whole ptp protocol state machine has been analyzed
and mostly included in the ptp-wr project.

fixes a warning on return type

fix for coherency with ptpd: if announce timer expires and our clock can not
act as master, the state must be put back to LISTENING

state-passive.c and state-slave.c now use the same st_com_handle_pdelay_req
for PPM_PDELAY_REQ messages handling.

state-slave.c: only missing some servo functions and msg issue
pproto.h: shortened some variable names