Remove unused minipc call get_timing_state.
Remove functions functions halexp_get_timing_state, export_get_timing_state
and struct hexp_timing_state_t.
Update developer manual
Signed-off-by: Adam Wujek <adam.wujek@cern.ch>

Update developer manual after removal of get_port_state.
Signed-off-by: Adam Wujek <adam.wujek@cern.ch>

hal_port_lookup was redundant with hal_lookup_port.
File hal_shmem.c is not needed anymore.
add const qualifier to name parameter of hal_lookup_port function
Signed-off-by: Adam Wujek <adam.wujek@cern.ch>

also removed typedefs
hexp_port_state_t
hexp_port_list_t
Signed-off-by: Adam Wujek <adam.wujek@cern.ch>

Minipc call to get port state will not be used anymore, since HAL export port
state through shmem.
Signed-off-by: Adam Wujek <adam.wujek@cern.ch>

Function hal_port_get_exported_state, fills structure hal_port_state with data
that are not available in SHM. It just assigns defines.
These defines are private to HAL, and assignment should be done in HAL.

Similar problem is for wrpc. There is a wrapper function that pretends to be
HAL. All HAL info should be filled in that function.

Fields added:
--uint32_t clock_period;
--uint32_t t2_phase_transition;
--uint32_t t4_phase_transition;

Additionally updated wrs_dump_shmem.c to dump new fields and the version number
of HAL SHM.
Signed-off-by: Adam Wujek <adam.wujek@cern.ch>

No technical change
Signed-off-by: Adam Wujek <adam.wujek@cern.ch>

It can be regenerated if needed (and added to .gitignore I hope).
At this point, it's not useful. It actually hits us in "git grep"
when we look for code.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

-DDEBUG in libwr forces LOG_DEBUG level in libwr/wrs-msg.c. This is
the desired behaviour, but we want it off by default now.

All other DEBUG symbols in Makefiles have no effect, so they are
removed to avoid tricking our developers in setting/unsetting them
and see no effect at all.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

This was a build-time conditional. It is now a run-time conditional
using wrs-msg.  As a result, we have a verbose rtud. Fixed in next
commit.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

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

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

Previous commits made a number of local variables and cast assignment
into identity operations. So this removes them.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

typedefs are bad; and typedefs to void pointers more so (there is
no checking by the compiler any more).  Use an opaque structure instead.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

There is no need to use packed structures unless they match hardware
and or stored-structures of legacy processors.  etherpacket here may
look like an issue, but it's naturally aligned by itself. Nobody does it
for a reason.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Meanwhile, remove one of the typedefs and avoid "my_socket":
"wr_socket" is more correct.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

meanwhile, turn "6" to ETH_ALEN. It doesn't make sense to typedef
if it doesn't hide details; so let's use uint8_t and it's defined size.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

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

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

We returned PTPD_NETIF_ERROR or PTPD_NETIF_OK but the caller was
cheking against 0 (or not checking). Use 0 and -1 removing the names.
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: Adam Wujek <adam.wujek@cern.ch>

Use "wr_date -n" to get host time (which might come from NTP
information) instead of WR time.

save_kconfig() new algorithm.
It supports reading and saving previous commented lines and
enabling/disabling binary options (i.e. CONFIG_PTP_WR_DEFAULT=y or "#
CONFIG_PTP_WR_DEFAULT is not set").
It's not done by hand anymore.

header ports were not loaded correctly because of format changes in
tmp/ports.conf output. Fixed now.

delta_tx and delta_rx in the SFP information are differences from
the values in the "calibration" SFP model. Thus they are signed
not uint32_t.  Fortunately nobody used negative values so far.

I also renamed the fields, to ensure the compiler would find all uses.

The commit also syncs headers to ppsi, including the changes that
happened here but were irrelevant to there.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

When two free-running switches were turned on at the same time, the
slave's ppsi exited with a fatal error because it was setting system
time to a negative UTC value (small - 35).  We now don't set system
time to anything older than today -- if WR is that wrong, anything
else is better for sure.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>
Acked-by: Adam Wujek <adam.wujek@cern.ch>

All clients using query_ports moved to use shm for communication with HAL.
Signed-off-by: Adam Wujek <adam.wujek@cern.ch>

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

wrsw_hal does not support "-c" parameter anymore.
Signed-off-by: Adam Wujek <adam.wujek@cern.ch>

Arrays uses WRS_N_PORTS as maximum number of ports.
Signed-off-by: Adam Wujek <adam.wujek@cern.ch>

Iussue log entry after every 100 unsuccessful tries instead of
just one entry when more than 100 unsuccessful tries.
Signed-off-by: Adam Wujek <adam.wujek@cern.ch>

read_ports should return -1 when it cannot get consistent
data within 100 tries
Signed-off-by: Adam Wujek <adam.wujek@cern.ch>

Additionally it looks like rtud entries where wrong previously from some moment
because entries had wrong destination port (reported by rtu_stat).
However no switch missbehaviour was observed.
As a side effect rtud removes entries after cable unplug.

Requires commits:
-- "userspace/libwr: shmem init-time lock"
-- "userspace/wrsw_hal: add support for shmem init-time lock"
Signed-off-by: Adam Wujek <adam.wujek@cern.ch>

Release write lock when HAL's shm is populated with correct data.
Requires commit: "userspace/libwr: shmem init-time lock"
Signed-off-by: Adam Wujek <adam.wujek@cern.ch>

Needed to ensure that writer is available before reader.
Note that data might not be available, when lock is released.
Read data with wrs_shm_seqbegin and wrs_shm_seqend to ensure proper data!
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Even function halexp_query_ports was called results were not used.
Signed-off-by: Adam Wujek <adam.wujek@cern.ch>