Use GPS data with ts2phc so time transfers from GPS to NIC PHY

[geerlingguy]

Currently the configuration in #1 is using generic for the ptp0 clock source:

sudo ts2phc -c /dev/ptp0 -s generic -l 7 -m \
  --ts2phc.pin_index 2 \
  --ts2phc.extts_polarity both --ts2phc.pulsewidth 100000000

From ts2phc's documentation:

ts2phc.nmea_serialport, ts2phc.nmea_baudrate
Specifies the serial port and baudrate in bps for character device providing ToD information when using the nmea PPS signal source. Note that if the options, ts2phc.nmea_remote_host and ts2phc.nmea_remote_port, are both specified, then the given remote connection will be used in preference to the configured serial port. The default serial port is /dev/ttyS0. The default baudrate is 9600 bps.

[geerlingguy]

If I attempt to override those options:

sudo ts2phc -c /dev/ptp0 -s generic -l 7 -m \
  --ts2phc.pin_index 2 \
  --ts2phc.extts_polarity both --ts2phc.pulsewidth 100000000 \
  --ts2phc.nmea_serialport /dev/ttyAMA0 --ts2phc.nmea_baudrate 115200

I get:

ts2phc: unrecognized option '--ts2phc.nmea_baudrate'

And if I drop nmea_baudrate, then it seems the time does not transfer regardless—probably because of the baud rate mismatch. (Should be 115200, default is 9600).

[geerlingguy]

From https://github.com/jclark/rpi-cm4-ptp-guide/blob/main/gps-config.md —

Baud rate. If ts2phc is used to interpret the GPS NMEA output, the the baud rate needs to be 9600, since the version of ts2phc in the Bullseye version of Raspberry Pi OS requires this. For many GPS modules, 9600 is default. But some more recent modules have a higher default baud rate (e.g. RCB-F9T defaults to 115200); in this case, then you may also need to disable some messages, to ensure that the amount of information transmitted by the GPS is not too much for 9600 baud.

@jclark seems to have a customized version of linuxptp including the baud rate option, see: jclark/rpi-cm4-ptp-guide#33 (comment)

[geerlingguy]

Attempting to use the latest master version of linuxptp (for some reason the Git repo doesn't include a 4.x branch, just up to v3.1.x, so I'm sticking with master for the time being).

# Remove Debian package for `linuxptp`
sudo systemctl stop ts2phc
sudo systemctl stop ptp4l
sudo systemctl stop phc2sys
sudo apt remove linuxptp

# Install linuxptp at `master`
cd ~
git clone https://github.com/richardcochran/linuxptp.git
cd linuxptp
make
sudo make install

# Verify I'm on latest
$ ptp4l -v
4.4-00022-ga0154d3

[geerlingguy]

Well... now running the option with the baud rate and serial port, I get:

...
ts2phc[26507.406]: /dev/ptp0 SKIP extts index 0 at 1744337873.144896382 src 1744337873.144921247
ts2phc[26507.507]: /dev/ptp0 SKIP extts index 0 at 1744337873.245995876 src 1744337873.246020630
ts2phc[26509.509]: poll returns zero, no events
ts2phc[26511.512]: poll returns zero, no events
ts2phc[26513.521]: poll returns zero, no events

It does the SKIP for a while, then starts spitting out the no events thereafter, every few seconds.

[jclark]

I did a backport of LinuxPTP 4.2 from testing to bookworm. https://github.com/jclark/rpi-cm4-ptp-guide/blob/main/files/linuxptp/linuxptp_4.2-1~bpo12%2B1_arm64.deb .

[geerlingguy]

@jclark I feel at some point if I could just clone your brain's knowledge into my own, I'd probably spent a few less hours on these things :)

Thank you for all your documentation! And the work on the linuxptp packages/packaging. I'll give that a try after messing with the git clone.

At this point, if I create a separate serial port in /dev as suggested in the Tindie docs for the TimeHAT:

sudo ln -s /dev/ttyAMA0 /dev/ttyS10

I am seeing some adding tstamp messages in the ts2phc logs:

...
ts2phc[26766.244]: adding tstamp 1744338134.469493988 to clock /dev/ptp0
ts2phc[26766.244]: /dev/ptp0 offset 2469493988 s2 freq +62499999
ts2phc[26766.335]: /dev/ptp0 SKIP extts index 0 at 1744338134.563244594 src 1744338132.74614439
ts2phc[26767.160]: /dev/ptp0 SKIP extts index 0 at 1744338135.407000109 src 1744338132.899608035
ts2phc[26767.252]: adding tstamp 1744338135.500750733 to clock /dev/ptp0
ts2phc[26767.252]: /dev/ptp0 offset 2500750733 s2 freq +62499999
...

And I'm seeing the Time offset in cgps slowly lowering back to 0... we'll see how quickly it steps down to that! I'll leave it for a bit.

Likely due to the rising/falling edge bug with the i226 driver, I'm still seeing tons of the SKIPs though.

@jclark - what NIC are you using in your testing? Just built-in on the Pi?

[jclark]

The NICs I test on are:

• Raspberry Pi Compute Module - CM4/CM5
• Intel - i210-T1, i225-T1

I haven't tried the Intel NICs on arm64. I haven't tried i226-T1 yet, but I don't expect it to be any different. I have one, but I haven't soldered on the header yet. (I'm not so hot at soldering.)

Last time I tried ts2phc with nmea and Intel NICs, it didn't work at all.

[geerlingguy]

Testing with:

sudo ts2phc -c /dev/ptp0 -s nmea -l 7 -m \
  --ts2phc.pin_index 2 \
  --ts2phc.extts_polarity both --ts2phc.pulsewidth 1000000 \
  --ts2phc.nmea_serialport /dev/ttyS10 --ts2phc.nmea_baudrate 115200

Might have to give up for the evening... hitting my head against the wall. Sometimes it seems to start syncing, other times it starts grabbing timestamps... but now I've kinda broken something so I'll come back to it later when I have a fresh mind :)

If I switch from nmea to generic and try again without the nmea options, I wind up with:

Apr 10 22:57:36 mini-time ts2phc[5042]: [2455.372] /dev/ptp0 SKIP extts index 0 at 1744343819.900690580 src 1744343893.900717423
Apr 10 22:57:37 mini-time ts2phc[5042]: [2456.272] /dev/ptp0 SKIP extts index 0 at 1744343820.800697275 src 1744343894.800727493
Apr 10 22:57:37 mini-time ts2phc[5042]: [2456.372] /dev/ptp0 SKIP extts index 0 at 1744343820.900698011 src 1744343894.900729203
Apr 10 22:57:38 mini-time ts2phc[5042]: [2457.272] /dev/ptp0 SKIP extts index 0 at 1744343821.800704699 src 1744343895.800730224
Apr 10 22:57:38 mini-time ts2phc[5042]: [2457.372] /dev/ptp0 SKIP extts index 0 at 1744343821.900705435 src 1744343895.900732470
Apr 10 22:57:39 mini-time ts2phc[5042]: [2458.272] /dev/ptp0 SKIP extts index 0 at 1744343822.800712123 src 1744343896.800746429
Apr 10 22:57:39 mini-time ts2phc[5042]: [2458.372] /dev/ptp0 SKIP extts index 0 at 1744343822.900712859 src 1744343896.900742472
Apr 10 22:57:40 mini-time ts2phc[5042]: [2459.272] /dev/ptp0 SKIP extts index 0 at 1744343823.800719547 src 1744343897.800748452
Apr 10 22:57:40 mini-time ts2phc[5042]: [2459.372] /dev/ptp0 SKIP extts index 0 at 1744343823.900720283 src 1744343897.900756124
Apr 10 22:57:41 mini-time ts2phc[5042]: [2460.272] /dev/ptp0 SKIP extts index 0 at 1744343824.800726971 src 1744343898.800756418
Apr 10 22:57:41 mini-time ts2phc[5042]: [2460.372] /dev/ptp0 SKIP extts index 0 at 1744343824.900727707 src 1744343898.900757740

It seems like it works for a little bit, but ultimately reaches a state where it just does the SKIP for both edges, forever.

[geerlingguy]

Getting the clock back on track:

# Stop phc-affecting utilities
sudo systemctl stop ts2phc
sudo systemctl stop phc2sys

# Force a time update using chrony/NTP
sudo chronyc -a makestep  # This can break stuff, use at your own risk

# Set the clock on the NIC
sudo phc_ctl eth1 "set;" adj 37

[geerlingguy]

Testing the kernel patch for igc:

cd ~ ; mkdir kernel; cd kernel
git clone --depth=1 https://github.com/raspberrypi/linux
sudo apt install bc bison flex libssl-dev make

cd linux ; KERNEL=kernel_2712 ; make bcm2712_defconfig;
make -j6 Image.gz modules dtbs
sudo make -j6 modules_install

sudo cp /boot/firmware/$KERNEL.img /boot/firmware/$KERNEL-backup.img
sudo cp arch/arm64/boot/Image.gz /boot/firmware/$KERNEL.img
sudo cp arch/arm64/boot/dts/broadcom/*.dtb /boot/firmware/
sudo cp arch/arm64/boot/dts/overlays/*.dtb* /boot/firmware/overlays/
sudo cp arch/arm64/boot/dts/overlays/README /boot/firmware/overlays/

Then confirm a reboot shows latest kernel. Then apply patch, recompile, reboot, and profit!

cd ~ ; cd kernel ; cd linux
wget https://raw.githubusercontent.com/Time-Appliances-Project/Incubation-Projects/refs/heads/master/Software/I225_ppsin_fix/i225_pps_rising_edge_only_1-31-2025.patch
patch -p1 < i225_pps_rising_edge_only_1-31-2025.patch

KERNEL=kernel_2712 ; make bcm2712_defconfig;
make -j6 Image.gz modules dtbs
sudo make -j6 modules_install

sudo cp /boot/firmware/$KERNEL.img /boot/firmware/$KERNEL-backup.img
sudo cp arch/arm64/boot/Image.gz /boot/firmware/$KERNEL.img
sudo cp arch/arm64/boot/dts/broadcom/*.dtb /boot/firmware/
sudo cp arch/arm64/boot/dts/overlays/*.dtb* /boot/firmware/overlays/
sudo cp arch/arm64/boot/dts/overlays/README /boot/firmware/overlays/

[geerlingguy]

During the kernel build with the patch, I get:

$ make -j6 Image.gz modules dtbs
  CALL    scripts/checksyscalls.sh
  CC [M]  drivers/net/ethernet/intel/igc/igc_main.o
  CC [M]  drivers/net/ethernet/intel/igc/igc_mac.o
  CC [M]  drivers/net/ethernet/intel/igc/igc_i225.o
  CC [M]  drivers/net/ethernet/intel/igc/igc_base.o
  CC [M]  drivers/net/ethernet/intel/igc/igc_nvm.o
drivers/net/ethernet/intel/igc/igc_main.c: In function ‘igc_tsync_interrupt’:
drivers/net/ethernet/intel/igc/igc_main.c:5461:43: error: ‘struct igc_adapter’ has no member named ‘ts0_pin’
 5461 |                 unsigned int pin = adapter->ts0_pin;
      |                                           ^~
drivers/net/ethernet/intel/igc/igc_main.c:5465:32: error: ‘struct igc_adapter’ has no member named ‘ts0_flags’; did you mean ‘ptp_flags’?
 5465 |                 if ( (adapter->ts0_flags & PTP_RISING_EDGE) &&
      |                                ^~~~~~~~~
      |                                ptp_flags
drivers/net/ethernet/intel/igc/igc_main.c:5466:37: error: ‘struct igc_adapter’ has no member named ‘ts0_flags’; did you mean ‘ptp_flags’?
 5466 |                         (!(adapter->ts0_flags & PTP_FALLING_EDGE) ) ) {
      |                                     ^~~~~~~~~
      |                                     ptp_flags
drivers/net/ethernet/intel/igc/igc_main.c:5486:32: error: ‘struct igc_adapter’ has no member named ‘ts0_flags’; did you mean ‘ptp_flags’?
 5486 |                 if ( (adapter->ts0_flags & PTP_FALLING_EDGE) &&
      |                                ^~~~~~~~~
      |                                ptp_flags
drivers/net/ethernet/intel/igc/igc_main.c:5487:37: error: ‘struct igc_adapter’ has no member named ‘ts0_flags’; did you mean ‘ptp_flags’?
 5487 |                         (!(adapter->ts0_flags & PTP_RISING_EDGE) ) ) {
      |                                     ^~~~~~~~~
      |                                     ptp_flags
drivers/net/ethernet/intel/igc/igc_main.c:5501:43: error: ‘struct igc_adapter’ has no member named ‘ts1_pin’
 5501 |                 unsigned int pin = adapter->ts1_pin;
      |                                           ^~
drivers/net/ethernet/intel/igc/igc_main.c:5505:32: error: ‘struct igc_adapter’ has no member named ‘ts1_flags’; did you mean ‘ptp_flags’?
 5505 |                 if ( (adapter->ts1_flags & PTP_RISING_EDGE) &&
      |                                ^~~~~~~~~
      |                                ptp_flags
drivers/net/ethernet/intel/igc/igc_main.c:5506:37: error: ‘struct igc_adapter’ has no member named ‘ts1_flags’; did you mean ‘ptp_flags’?
 5506 |                         (!(adapter->ts1_flags & PTP_FALLING_EDGE) ) ) {
      |                                     ^~~~~~~~~
      |                                     ptp_flags
drivers/net/ethernet/intel/igc/igc_main.c:5512:32: error: ‘struct igc_adapter’ has no member named ‘ts1_flags’; did you mean ‘ptp_flags’?
 5512 |                 if ( (adapter->ts1_flags & PTP_FALLING_EDGE) &&
      |                                ^~~~~~~~~
      |                                ptp_flags
drivers/net/ethernet/intel/igc/igc_main.c:5513:37: error: ‘struct igc_adapter’ has no member named ‘ts1_flags’; did you mean ‘ptp_flags’?
 5513 |                         (!(adapter->ts1_flags & PTP_RISING_EDGE) ) ) {
      |                                     ^~~~~~~~~
      |                                     ptp_flags
drivers/net/ethernet/intel/igc/igc_main.c: In function ‘igc_probe’:
drivers/net/ethernet/intel/igc/igc_main.c:7222:16: error: ‘struct igc_adapter’ has no member named ‘ts0_pin’
 7222 |         adapter->ts0_pin = 1000;
      |                ^~
drivers/net/ethernet/intel/igc/igc_main.c:7223:16: error: ‘struct igc_adapter’ has no member named ‘ts1_pin’
 7223 |         adapter->ts1_pin = 1000;
      |                ^~

[geerlingguy]

Putting a pin in the kernel compile for now — see #15

[github-actions]

This issue has been marked 'stale' due to lack of recent activity. If there is no further activity, the issue will be closed in another 30 days. Thank you for your contribution!

Please read this blog post to see the reasons why I mark issues as stale.