This is old misc stuff that I rebased yesterday. Actually, I had more
moves to fsm-lib, but some of it already moved to common-fun (which I
dislike, but let's ignore the thing by now).

update:
include/libwr/sfp_lib.h
include/libwr/util.h
shmem.c
util.c
Signed-off-by: Adam Wujek <adam.wujek@cern.ch>

This is another step in a general cleanup of the state machine.  To
help me knowning what is audited and what is not, I rename the "common
functions" to fsm-lib. This is the management of announce.

- master and slave functions were the same, so they are unified now

- hooks->handle_annunce returns int (it may fail)
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

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

We check the frame length for the various message type in a single place,
thus removing all checks around (that didn't cover all cases anyways).

Also, we check for protocol version 2, something that we never did.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

This removes a number of explicit checks in the states themselves.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

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

update:
include/libwr/sfp_lib.h
include/libwr/util.h
shmem.c
util.c
Signed-off-by: Adam Wujek <adam.wujek@cern.ch>

On the switch, there is no need for the ppsi to use config file in
the /wr/etc/ppsi.conf.
Signed-off-by: Adam Wujek <adam.wujek@cern.ch>

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

As a side effect, we get a reported "enter initializing" message
when wrs has a link-up event (because in that case state-change is
performed externally, not by a state that switches to another).
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Commit d2b2c636 introduced a caching of the current state structure.
The "current" structure, as long as not null, is considered valid.
arch-wrs, just sets ppi->state to initializing on link-up, without
clearing the cached pointer, so fsm.c was using the old pointer,
resulting in use of closed file descriptors, as no init was performed.

This doesn't apply to change from withing the states themselves, as
fsm.c manages the state change.

As a side effect, now that we check the state numberm there's no need
to zero the pointer on state leave. (And while a it another trivial detail...).
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>

Some arguments are not needed any more. Removed.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Most arguments of net->send are redundant, as the ppi includes all
info. In particular, chtype and "is_pdelay_addr" derive from the
message type.  And the timestamp is always internal.

I chose not to remove message, len and msgtype because the message
being sent includes the header (ppi->tx_offset), so there's some
calculation to extract them.  This may happen in a later commit, if I
find that is a size improvement.

Bugs fixed as a side effect:
   - the "if" for "is_pelay" missed PDELAY_RESP_F_UP (now it's in msgtype.c)
   - wrpc pdelay was always using the e2e mac address

This commit decreases the size of all archs by 60 bytes, but increases
wrpc by 20 bytes because of the fix above
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

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

Most stuff in the header can be factorized. The commit does so.
Some fixme's are removed (about correctionField for example) but
we know they still apply.
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 was my fault. The initial time jump was based on "T4 + MPD". This
on the assumption that T4 is sharper then time->get().  At least this
is true for White Rabbit systems, where time->get() was not even
correct when I started the project.

As a result, the initial jump for a p2p slave was always wrong,
because it referred to an old timestamp. We ended up a few hundreds ms
off the master, to the start a long steering.

We now add ofm to the current time, and the result is good in both case.
WR systems have their own servo anyways, but even a slave to non-wr
masters should behave well, because time->set now works.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

The idea that somebody wants to steer time for >1s offsets is
perverse.  Let's simplify (if any, we may want to jump for <1s offsets
that are bigger than a few milliseconds).
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>

The reshaping for pdelay broke the idea that if the offset is more
than one second we don't run the PI at all (to avoid effects of the I
part after the jump).

This restores the early-exit from the servo when a jump in time happens.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Also, two functions that were needlessly public are now static.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

This commit assumes that "transportSpefic" in header byte 0 and
"reserved" in header byte 1 are always 0.  Thus, it removes unneeded
bitmasking and while at it uses names for message types.  It also
removes redundant assignments and adds a few missing ones.

The commit simplifies all the "pack" functions; it makes them
consistent by setting all header fields in transmit order, and marking
unclean spots (i.e. correctionField assignment).

Buglets fixed in the process;
  - correctionField was not cleared in follow_up
  - pdelay_resp_f_up was using the wrong length name
  - pdelay_resp_f_up was not setting controlField and logMessageInterval
  - delay_req was not clearing flags
  - pdelay_req was not blanking the trailing bytes
  - pdelay_resp was not setting controlField and logMessageInterval
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

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

This commit is required to uplift a wr-switch kernel version

Since the kernel commit
850a545bd net: validate HWTSTAMP ioctl parameters
Follows: v3.1-rc9
Precedes: v3.2-rc1
SIOCSHWTSTAMP ioctl check whether field "flags" of struct hwtstamp_config
is empty, otherwise ioctl returns -EINVAL
Signed-off-by: Adam Wujek <adam.wujek@cern.ch>

Update sfp_lib.h and hal_shmem.h to support diagnostics (DOM) in SFPs.
Changes not affecting PPSI code.
Signed-off-by: Adam Wujek <adam.wujek@cern.ch>

There were many changes to the ppsi structures in few previous commits.
Signed-off-by: Adam Wujek <adam.wujek@cern.ch>

bug introduced in 4126868c, we were incrementing the delay-request
sequence number instead of ours.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

The function has gone big over time, it's better not inline any more.
The change saves 230 byte in wrpc builds.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

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

While it may make sense to setup constant delay somewhere (for WR we
define them, for instance), the legacy implementation was unused and
wrong.

Unused because nobody sets it (even if we could do it in the config
file).  Wrong because inbound is never used, and both are global and
not per-port.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Now this is the behaviour for the choice:

    arch-bare-*

         if you select P2P, it's built and selected. Default is E2E_ONLY.

    arch-sim
    arch-unix
    arch-wrs

         The P2P is built by default configuration, and the mechanism
         is selected at configuration time. Default is E2E. If you
         choose E2E_ONLY and configure pdelay, it will ignore the config.
         (Suboptimal but I'm lazy: do *not* use E2E_ONLY for those archs.

    arch-wrpc

        The default is e2e until changed at run time.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

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

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