As discussed in this email thread:
https://sourceforge.net/p/linuxptp/mailman/message/37047555/
there is a desire to synchronize multiple DSA switches in a
boundary_clock_jbod setup, using a PPS signal, and the ts2phc program
already offers a solid base.
This patch series extends the ts2phc program to cater for that use case
by introducing a new '-a' option and friends ('--transportSpecific').
This makes it quite similar to phc2sys which has the same ability, just
that ts2phc can give much better performance if the hardware can assist
with that.
The overall board design for my use case is that there's an SoC with an
embedded DSA switch, and hanging off of 3 ports of that embedded switch
are 3 external switches. Every networking device (the DSA master for the
embedded switch, the embedded switch, as well as each individual
external switch) has a PTP clock. The topology for PPS signal
distribution is fixed - that is given by hardware ability - the
/dev/ptp1 clock can emit a valid PPS, and all external switches can
timestamp that PPS (it is connected in a sort-of simplex "bus" design),
and it cannot be any other way around. It looks like this:
+---------------------------------------------------------------+
| LS1028A /dev/ptp0 (unused) |
| +------------------------------+ |
| | DSA master for Felix | |
| |(internal ENETC port 2: eno2))| |
| +------------+------------------------------+-------------+ |
| | Felix embedded L2 switch /dev/ptp1 | |
| | PPS source, sync target | |
| | +--------------+ +--------------+ +--------------+ | |
| | |DSA master for| |DSA master for| |DSA master for| | |
| | | SJA1105 1 | | SJA1105 2 | | SJA1105 3 | | |
| | |(Felix port 1)| |(Felix port 2)| |(Felix port 3)| | |
+--+-+--------------+---+--------------+---+--------------+--+--+
/dev/ptp2 /dev/ptp3 /dev/ptp4
PPS sink, sync reference PPS sink, sync target PPS sink, sync target
+-----------------------+ +-----------------------+ +-----------------------+
| SJA1105 switch 1 | | SJA1105 switch 2 | | SJA1105 switch 3 |
+-----+-----+-----+-----+ +-----+-----+-----+-----+ +-----+-----+-----+-----+
|sw1p0|sw1p1|sw1p2|sw1p3| |sw2p0|sw2p1|sw2p2|sw2p3| |sw3p0|sw3p1|sw3p2|sw3p3|
+-----+-----+-----+-----+ +-----+-----+-----+-----+ +-----+-----+-----+-----+
^
|
GM connected here
In text, it would be described as this:
cat ts2phc.cfg
[global]
first_step_threshold 0.00002
step_threshold 0.00002
ts2phc.pulsewidth 500000000
ts2phc.perout_phase 0
# Felix
[/dev/ptp1]
ts2phc.master 1
# SJA1105 switch 1
[/dev/ptp2]
ts2phc.channel 0
ts2phc.extts_polarity both
# SJA1105 switch 2
[/dev/ptp3]
ts2phc.channel 0
ts2phc.extts_polarity both
# SJA1105 switch 3
[/dev/ptp4]
ts2phc.channel 0
ts2phc.extts_polarity both
cat gPTP.cfg
[global]
gmCapable 1
priority1 248
priority2 248
logAnnounceInterval 0
logSyncInterval -3
syncReceiptTimeout 3
neighborPropDelayThresh 50000
min_neighbor_prop_delay -20000000
assume_two_step 1
path_trace_enabled 1
follow_up_info 1
transportSpecific 0x1
ptp_dst_mac 01:80:C2:00:00:0E
network_transport L2
delay_mechanism P2P
step_threshold 0.00002
tx_timestamp_timeout 20
boundary_clock_jbod 1
[sw1p0]
[sw1p1]
[sw1p2]
[sw1p3]
[sw2p0]
[sw2p1]
[sw2p2]
[sw2p3]
[sw3p0]
[sw3p1]
[sw3p2]
[sw3p3]
At a high level, what I have done is:
- I moved the PMC related code from phc2sys into pmc_common.c, for
ts2phc reuse
- I created an extra abstraction in ts2phc as "struct clock" that would
represent what's synchronizable. The "master" and "slave" concepts
retain their meaning, which is: "master" == the device that emits PPS,
and "slave" == the device that timestamps PPS.
- I added support for telling the PHC master to start emitting periodic
output using just phase information, and not an absolute time. This is
required for the kernel drivers I am working with.
- I added support for configuring the pulse width, not just hoping that
the value specified in ts2phc.cfg is representative of what the
hardware actually uses.
The changes should be backwards-compatible with the non-automatic mode.
I have only tested non-automatic mode with a PHC master (that's all I
have) and it still appears to work as before.
Changes in v6:
- eliminate use of offensive word "approximate"
- add back int fd to struct ts2phc_clock, and use it instead of
CLOCKID_TO_FD every time
- use generic port_state_normalize() rather than ts2phc logic duplication
- rename auto_init_ports() to ts2phc_auto_init_ports()
- adapt to newly added phc_index argument to pmc_agent_query_port_properties
- fixed bad patch splitting; add free(polling_array) to the patch where
it is needed
Changes in v5:
- Renamed "master" to "PPS source"
- Renamed "slave" to "PPS sink"
- Renamed "source clock" to "reference clock"
- Renamed "destination clock" to "target clock"
- Renamed "clock" to "ts2phc_clock"
- Renamed "port" to "ts2phc_port"
Changes in v4:
- Rebased and resolved minor conflicts
- Small coding style cleanups
- Added patch for perout and phase configuration
- Retested
Changes in v3:
- Split patch "phc2sys: break out pmc code into pmc_common.c" into more
trivial chunks. There is zero delta compared to previous monolithic
patch, otherwise.
- Replaced full license header in ts2phc.h with simple SPDX.
- Added usage message and manpage for '-a' option.
- Removed some useless curly braces.
- Moved ts2phc_master_get_clock() from ts2phc_phc_master.c to
ts2phc_master.c where it should be.
- Added justification in commit message for creating struct
ts2phc_private.
- Reordered headers alphabetically.
Changes in v2:
- Added Jacob's review tags, addressed some feedback.
- Dropped patch "pmc_common: fix potential memory leak in
run_pmc_events()"
- Reordered patches (put the tmv helpers first)
- Fixed memory leaks
Vladimir Oltean (10):
ts2phc: create a private data structure
ts2phc: instantiate a full clock structure for every PPS sink
ts2phc: instantiate a full clock structure for every PPS source of the
PHC kind
util: import port_state_normalize() logic from phc2sys
ts2phc: instantiate a pmc agent
ts2phc_slave: print offset to the source clock
ts2phc: split PPS sink poll from servo loop
ts2phc: reconfigure sync direction by subscribing to ptp4l port events
ts2phc: allow PHC PPS sources to be synchronized
ts2phc_phc_pps_source: make use of new kernel API for perout waveform
config.c | 1 +
makefile | 5 +-
missing.h | 52 +++
phc2sys.c | 18 +-
ts2phc.8 | 32 +-
ts2phc.c | 640 ++++++++++++++++++++++++++++++++++--
ts2phc.h | 65 ++++
ts2phc_generic_pps_source.c | 2 +-
ts2phc_generic_pps_source.h | 3 +-
ts2phc_nmea_pps_source.c | 6 +-
ts2phc_nmea_pps_source.h | 3 +-
ts2phc_phc_pps_source.c | 80 +++--
ts2phc_phc_pps_source.h | 3 +-
ts2phc_pps_sink.c | 381 +++++++++++----------
ts2phc_pps_sink.h | 12 +-
ts2phc_pps_source.c | 18 +-
ts2phc_pps_source.h | 8 +-
ts2phc_pps_source_private.h | 1 +
util.c | 13 +
util.h | 8 +
20 files changed, 1067 insertions(+), 284 deletions(-)
create mode 100644 ts2phc.h
--
2.34.1
_______________________________________________
Linuxptp-devel mailing list
Linuxptp-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linuxptp-devel
