related to BUG: 1544 and 1810
Signed-off-by: Adam Wujek <adam.wujek@cern.ch>

It requires that ppsi stores port_ds and wr_port_ds in the shmem
Signed-off-by: Adam Wujek <adam.wujek@cern.ch>

This is needed to be able to dump these structures in the wrs_dump_shmem
Signed-off-by: Adam Wujek <adam.wujek@cern.ch>

Compilation of wrs_dump_shmem_ppsi.c is done from the wr-switch-sw repo
since it needs wrs_dump_shmem.h, which is in wr-switch-sw repo.
I wanted to avoid keeping two copies of .h file.
Signed-off-by: Adam Wujek <adam.wujek@cern.ch>

Disect WR frames in ptpdump and during printing debug info in PPSi.
Current implementation prints deltaTX and deltaRx in hex not in decimal.
Signed-off-by: Adam Wujek <adam.wujek@cern.ch>

This commit fixes dumping PTP packets in ptpdump and PPSi (via -d 002)
Signed-off-by: Adam Wujek <adam.wujek@cern.ch>

Signed-off-by: Adam Wujek <adam.wujek@cern.ch>

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

abscal: simplify code related to CONFIG_ABSCAL
Signed-off-by: Adam Wujek <adam.wujek@cern.ch>

Signed-off-by: Adam Wujek <adam.wujek@cern.ch>

kernel has two bugs regarding leap second handling: it need to have the
leap second flags cleared twice before setting them and it doesn't clear the leap second flags after the leap second happened. A workaround was implemented
accordingly.

Signed-off-by: Adam Wujek <adam.wujek@cern.ch>

Also in the loop for erbest the search was fixed if the first port
which is elected as ebest by default is not valid.

Since the portnumber is a 16bit value, memcp is not given to be correct
on all architectures depending on the endines

msg: reset ext_specific flags in the announce to zero when either no TLV is appended or the TLV is not a WR TLV

The flags was not set to any value when no TLV was appended or a wrong TLV was appended, in that
case the RAM content was wrongly propagated as WR flags which may caused that a node is going into
a wrong state.

For the dataset comparisment all known foreign masters have to be qualified
so the worst is found. Once aged out it will be removed so the qualification is
not relevant in this check.

Signed-off-by: Adam Wujek <adam.wujek@cern.ch>

The UTC leap flags are propagated to the system (unix) so the leap second is correctly handled at midnight
so the UTC time is always correct.

the variance is now set according to the servo state, the source and timescale
are set according to the clock classes set

Added clock accuracy constants and aligned the chaning of the clock class
with the accuracy

The clock class is now decreased and increased based on the servo state,
this only has an effect for wrs and wrpc since the contain a servo loop.

Signed-off-by: Adam Wujek <adam.wujek@cern.ch>

For links longer than ~13km, a value (uint32_t) returned from a function
__div64_32 shifted by 16 bits overflowed, introducing PPS offset.

Bug introduced in the commit:
35eef679 time: introduce a new unified data structure
Signed-off-by: Adam Wujek <adam.wujek@cern.ch>

For whatever reason we perform a normalization on an incorrect timestamp,
don't treat is as an negative value.
Signed-off-by: Adam Wujek <adam.wujek@cern.ch>

The clock class goes only to 6 if realy in gm mode, so not even passive
The variance was changed back to 0x8000 until computed

We don't compute the variance so the default value shall be 0xFFFF