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

This commit removes the part of wrpc.h that is not used.  It was there
when we tried to build a main function in ppsi itself, but now we
depend on the exteranl wrpc-sewpackage to provide drivers and headers,
so wrpc.h is now just for local material.

Then, wrc_ptp_ppsi.c is fixed with regard to globals and all the rest.
Moreover, I moved most of the initialization of static variables into
compile-time assignments instead of run-time, since there's nothing
dynamic here in those fields.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

I forgot to fix arch-wrpc when making the time_ops.
Code mainly by Tomasz (wrpc-sw/shell/cmd_time.c)
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

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

This new file needs some updates over what we imported from wrpc-sw,
because we now have pp_globals (in preparation for multi-link, thanks
Aurelio).

The commit also makes a little order in the inclusions of this file,
removing unneeded stuff and grouping hardware-related headers.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

These few lines are a copy of "wrc_ptp.h" as found in ptp-noposix.
They are added here because they are the prototypes of now-local
functions, and little more.

Including the file from wrpc-sw would have required another -I
directory, and I'd better not.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

This file comes from commit e00d99f of wrpc-sw.  It used to live there
as an alternative to wrc_ptp_noposix.c, as a top-level entry point to
the ptp engine.

However, ppsi is quite a moving target these months, and most
improvements in modularity of ppsi require a change in the top-level
file (like we change the main and startup files in the other
architectures).

Thus, this commit brings the file into ppsi (and wrpc-sw will remove
it when using this version of ppsi as submodule) so we can be work
on this repository and build-test everything with no effort, saving us
from always committing to wrpc-sw as well.

Note: this commit copies the file unchanged, and doesn't even try to
build it; later commits make the required adjustments to finally build
it.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>
Acked-by: Aurelio Colosimo <aurelio@aureliocolosimo.it>

While building under ubuntu-12.04 (64 bits, at least), bare builds
were failing because of undefined "__stack_chk_fail" and "__strcpy_chk".

The former is fixed by passing -ffreestanding to the build of pp_printf
(i.e., we now export CFLAGS in the toplevel Makefile), amd the latter
is fixing by preventing use of the inline printf function when building
for bare architectures.

This is from the manpage of gcc under ubuntu-12.04:

     NOTE: In Ubuntu 8.10 and later versions, -D_FORTIFY_SOURCE=2 is set
     by default, and is activated when -O is set to 2 or higher.  This
     enables additional compile-time and run-time checks for several
     libc functions.  To disable, specify either -U_FORTIFY_SOURCE or
     -D_FORTIFY_SOURCE=0.

Thanks Aurelio for reporting ./MAKEALL was failing.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

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

This was an error on my side. The result may be undefined in
"ntohs(++x)".  It had not bad effects so far, but newer compilers
correctly warn about it, and this commit fixes the error.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

This fixes warnings that either come from a newer compiler than what I
routinely use or from building in a 64-bit host.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Our "eval export CC=" needs an exta quoting level under dash,
to prevent an error like the following:

   export: 1: -m32: bad variable name

This setup still works under bash.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

When calculating the running average (and the average between two as
well), we must first add all items and then divide. Everyone does like
this, so we should as well, to avoid loosing the low bits

Example: 3 and 5, average is 4: (3+5) / 2 = 4, but 3/2 + 5/2 = 3.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

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

This only has effects on diagnostics: now all exit paths from
pp_update_clock() pass through the diagnostics stanza. I also removed
a duplicate diagnostic message and a now-unneeded else, so the good code
is at the first indentation level.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Signed-off-by: Aurelio Colosimo <aurelio@aureliocolosimo.it>

Signed-off-by: Aurelio Colosimo <aurelio@aureliocolosimo.it>

No functional change at all, just rename and make var name shorter.
Signed-off-by: Aurelio Colosimo <aurelio@aureliocolosimo.it>

Signed-off-by: Aurelio Colosimo <aurelio@aureliocolosimo.it>

This removes some needless local variables related to correction
field, but relying on a new "cField" entry in pp_instance, near
the t1,t2,t3,t4 status stamps.

The commit also adds a FIXME note: it is not clear to me how the
correction field can be managed in the back message (t3, t4).
Since the correctionField is updated by transparent clocks, I think
we need the one of the delay request (the one stamped in t3, t3),
*not* the delay response that I don't care about.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

For two-step masters, the code was adding the correction fields of
sync and follow-up.  This is wrong, since the follow-up message only
carries the sender's timestamp for sync.  So, the only correction to
consider is the one in sync (which accounts for transparent clocks
the frames passed through).

The approach taken is promoting the current correction to be a ppi
field (like t1..t4), while killing the other "temporary" field that
was used before this commit to keep the value that was going to be added.

Note, however, that how cField is used in the back path (t3,t4) is not
clear to me, and the specification is not helping: if the correction
is automatically updated by transparent clocks in my path, the master
should send back to me the value it received for my delay request.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

I initially looked at int64_to_TimeInternal to swap arguments, since
to_TimeInternal as timeinternal (the output) as first argument, and
they appeared in pairs.  However, int64_to_TimeInternal is only used
for correction fields, and they are scaled nanoseconds (page 126), not
nanoseconds.

Thus, this commit fixes the conversion and renames the function to
"cField_to_TimeInternal" to better tell the reader what it is.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

While getting mad with servo.c (which makes little sense to me), and
continuously following the long unseless indented stanza of
pp_update_delay() I finally decided to kill it.  This kills another
similar situation, but each of these changes makes no effect on the
program flow, only simplifies code as written.

You can diff with "-w" to verify I don't change code.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Now that it's clear pp_update_offset() receives t1 and t2, which are
fields of ppi itself, avoid passing them.

However now it's clear that this function might be seriously wrong,
and needs review (ofst_first_updated should die, I think, preferring
simply code to unneeded contortions).
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

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

This uses the "t1" field, introduced two commits ago, instead of local
variables in the functions handling sync/follow_up.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

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

This commit simply reorders the TimeInternal fields of pp_instance,
to make the structure more understandable.  It also adds t1 that
will be used by a later commit.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

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

This renames sync_receive_time to t2, delay_req_send_time to t3
and delay_req_receive_time to t4 (there is "t1" field in ppi yet).

Such naming is what the official documents use everywhere, so it's
better to respect that naming, that everybody in the field is familiar
with. Besides, they are shorter and they can be used unchanged for
pdelay operations (the previous pdelay implementation we inherited
from ptpd, used a separate set of fields that are not really needed).
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

This patch introduces struct pp_globals in include/ppsi/pp-instance.h
and starts using it in arch-gnu-linux. Actually, the behaviour does not
change yet, and only one interface is taken into account. In main-loop.c
there is a proposal for interfaces enumeration, based on runtime env
variables:
- one PPROTO_IF_NUM (number of interfaces);
- many PPROTO_IF_XX (each one defining an interface name).
Signed-off-by: Aurelio Colosimo <aurelio@aureliocolosimo.it>

This includes the functionality of tools/ptpdump inside the ppsi
executable. Each net_ops->send and net_ops->recv must call the
dumping functions (gnu-linux and bare-* already to in this branch;
arch-wrpc does not, and I'll all it later).

I don't want to declare globals in header files by error (see previous
commit).  So I'd better use this flag to request "multiple definition"
errors at build time.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

The new diagnostics for frame input/output, based on tools/dump-funcs.c,
reports all information, more orderly.   The diagnostic prints in
msg.c are thus duplicates (provided all architectures use the new
diagnostics in ->recv), and this commit removes them.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

In commit 7c14152c I removed some typedef, in my quest for getting rid
of them.  But I forgot to remove the type name, so we got "TLV" and
similar global variables in the object files.  This fixes the buglet,
then I add -fno-common to prevent this for happening againg.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Like what I did in two commits ago, now if debug level for "frames" is
2 or more, net_ops->send and net_ops->recv now dump the data to
stdout. The format is the same of tools/ptpdump, but the time is
not showing the local time, as we miss localtime() and I don't want
to waste time recoding it.

This is an example of a bare-i386 slave requesting its master:

  send: TIME: (1363278286 - 0x5141f9ce) 1363278286.52969000
  send: ETH: 88f7 (00:25:11:86:1f:d2 -> 01:1b:19:00:00:00)
  send: VERSION: 2 (type 1, len 44, domain 0)
  send: FLAGS: 0x0002 (correction 0x00000000)
  send: PORT: 00-25-11-ff-fe-86-1f-d2-00-01
  send: REST: seq 2, ctrl 1, log-interval 127
  send: MESSAGE: (E) DELAY_REQ
  send: MSG-DELAY_REQ: 1363278286.052689000
  send: DUMP: payload (size 44)
  send: DUMP: 81 02 00 2c  00 00 02 00  00 00 00 00  00 00 00 00
  send: DUMP: 00 00 00 00  00 25 11 ff  fe 86 1f d2  00 01 00 02
  send: DUMP: 01 7f 00 00  51 41 f9 ce  03 23 f8 68
  diag-frames-1-eth0: SENT 44 bytes at 1363278286.052969000 (delay_req)

This is a bare-i386 master receiving the same frame:

  recv: TIME: (1363278286 - 0x5141f9ce) 1363278286.92899000
  recv: ETH: 88f7 (00:25:11:86:1f:d2 -> 01:1b:19:00:00:00)
  recv: VERSION: 2 (type 1, len 44, domain 0)
  recv: FLAGS: 0x0002 (correction 0x00000000)
  recv: PORT: 00-25-11-ff-fe-86-1f-d2-00-01
  recv: REST: seq 2, ctrl 1, log-interval 127
  recv: MESSAGE: (E) DELAY_REQ
  recv: MSG-DELAY_REQ: 1363278286.052689000
  recv: TLV: too short (46 - 44 = 2)
  recv: DUMP: payload (size 46)
  recv: DUMP: 81 02 00 2c  00 00 02 00  00 00 00 00  00 00 00 00
  recv: DUMP: 00 00 00 00  00 25 11 ff  fe 86 1f d2  00 01 00 02
  recv: DUMP: 01 7f 00 00  51 41 f9 ce  03 23 f8 68  00 00
  diag-frames-1-eth0: RECV 46 bytes at 1363278286.092899000 (type 1)

And this is tools/ptprump running on the master, for the same frame:

  TIME: (1363278286 - 0x5141f9ce) 17:24:46.092814
  ETH: 88f7 (00:25:11:86:1f:d2 -> 01:1b:19:00:00:00)
  VERSION: 2 (type 1, len 44, domain 0)
  FLAGS: 0x0002 (correction 0x00000000)
  PORT: 00-25-11-ff-fe-86-1f-d2-00-01
  REST: seq 2, ctrl 1, log-interval 127
  MESSAGE: (E) DELAY_REQ
  MSG-DELAY_REQ: 1363278286.052689000
  TLV: too short (46 - 44 = 2)
  DUMP: payload (size 46)
  DUMP: 81 02 00 2c  00 00 02 00  00 00 00 00  00 00 00 00
  DUMP: 00 00 00 00  00 25 11 ff  fe 86 1f d2  00 01 00 02
  DUMP: 01 7f 00 00  51 41 f9 ce  03 23 f8 68  00 00
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

This file includes data structures that are needed by
tools/dump-funcs.c to be built in freestanding environments. It is
included by ptpdump.h, but only for freestanding builds.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

If debug level for "frames" is 2 or more, net_ops->send and
net_ops->recv now dump the data to stdout. The format is the same
of tools/ptpdump, with "send:" and "recv:" in front. This is an example:

  send: TIME: (1363277414 - 0x5141f666) 17:10:14.748281
  send: VERSION: 2 (type 11, len 64, domain 0)
  send: FLAGS: 0x0002 (correction 0x00000000)
  send: PORT: 00-22-15-ff-fe-d7-c0-ef-00-01
  send: REST: seq 9, ctrl 5, log-interval 1
  send: MESSAGE: (G) ANNOUNCE
  send: MSG-ANNOUNCE: stamp 0.000000000
  send: MSG-ANNOUNCE: grandmaster-quality f8-fe-60f0
  send: MSG-ANNOUNCE: grandmaster-prio 248 248
  send: MSG-ANNOUNCE: grandmaster-id 00-22-15-ff-fe-d7-c0-ef
  send: DUMP: payload (size 64)
  send: DUMP: 8b 02 00 40  00 00 02 00  00 00 00 00  00 00 00 00
  send: DUMP: 00 00 00 00  00 22 15 ff  fe d7 c0 ef  00 01 00 09
  send: DUMP: 05 01 00 00  00 00 00 00  00 00 00 00  00 00 11 f8
  send: DUMP: f8 fe f0 60  f8 00 22 15  ff fe d7 c0  ef 00 00 a0
  diag-frames-1-eth0: SENT 64 bytes at 1363277414.748281517 (announce)

And this is the same frame in the other host:

  recv: TIME: (1363277414 - 0x5141f666) 17:10:14.691032
  recv: VERSION: 2 (type 11, len 64, domain 0)
  recv: FLAGS: 0x0002 (correction 0x00000000)
  recv: PORT: 00-22-15-ff-fe-d7-c0-ef-00-01
  recv: REST: seq 9, ctrl 5, log-interval 1
  recv: MESSAGE: (G) ANNOUNCE
  recv: MSG-ANNOUNCE: stamp 0.000000000
  recv: MSG-ANNOUNCE: grandmaster-quality f8-fe-60f0
  recv: MSG-ANNOUNCE: grandmaster-prio 248 248
  recv: MSG-ANNOUNCE: grandmaster-id 00-22-15-ff-fe-d7-c0-ef
  recv: DUMP: payload (size 64)
  recv: DUMP: 8b 02 00 40  00 00 02 00  00 00 00 00  00 00 00 00
  recv: DUMP: 00 00 00 00  00 22 15 ff  fe d7 c0 ef  00 01 00 09
  recv: DUMP: 05 01 00 00  00 00 00 00  00 00 00 00  00 00 11 f8
  recv: DUMP: f8 fe f0 60  f8 00 22 15  ff fe d7 c0  ef 00 00 a0
  diag-frames-1-eth0: RECV 64 bytes at 1363277414.691032726 (type b)

This is what I got for the same frame in a ptpdump running in the
master.

  TIME: (1363277414 - 0x5141f666) 17:10:14.748376
  ETH: 0800 (00:22:15:d7:c0:ef -> 01:00:5e:00:01:81)
  IP: 17 (192.168.16.1 -> 224.0.1.129) len 92
  UDP: (320 -> 320) len 72
  VERSION: 2 (type 11, len 64, domain 0)
  FLAGS: 0x0002 (correction 0x00000000)
  PORT: 00-22-15-ff-fe-d7-c0-ef-00-01
  REST: seq 9, ctrl 5, log-interval 1
  MESSAGE: (G) ANNOUNCE
  MSG-ANNOUNCE: stamp 0.000000000
  MSG-ANNOUNCE: grandmaster-quality f8-fe-60f0
  MSG-ANNOUNCE: grandmaster-prio 248 248
  MSG-ANNOUNCE: grandmaster-id 00-22-15-ff-fe-d7-c0-ef
  DUMP: payload (size 64)
  DUMP: 8b 02 00 40  00 00 02 00  00 00 00 00  00 00 00 00
  DUMP: 00 00 00 00  00 22 15 ff  fe d7 c0 ef  00 01 00 09
  DUMP: 05 01 00 00  00 00 00 00  00 00 00 00  00 00 11 f8
  DUMP: f8 fe f0 60  f8 00 22 15  ff fe d7 c0  ef 00 00 a0
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

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

This is oh so boring, but I need to differentiate send from receive
when dumping from withing ppsi itself.  This is a long patch that has
no effect on tools/ptpdump, since it pases the empty string as prefix.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>