Figure out easier way to get igc i226-V driver working for rising-edge detection

[geerlingguy]

One of the more annoying aspects of using an i226-V is having to deal with the rising/falling edge detection with ts2phc. This requires overriding --ts2phc.extts_polarity both and setting --ts2phc.pulsewidth 1000000, but even there, many of the interrupts seem to be skipped entirely, especially after ts2phc is running for a while.

For example:

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

See #13 and #14 for more context.

It would be interesting if we could either:

1. Make it easier to manage a custom kernel with this patch that fixes the bug. Here's what I'm currently testing, it's not fun and the initial kernel compile takes 45 min or so.
2. Maintain a kernel DKMS module version of the igc driver, with the fix applied, so it is easier to keep the kernel up to date automatically, and not have to take on the entire kernel build process over time.

For the second, there's an older igc DKMS project that has not been maintained for 4 years. I can't find anyone maintaining any others out of the main Linux source tree though.

Intel also offers a Downloadable driver pack, but I don't see an easy way to mess with patching that and using it separately.

[geerlingguy]

Testing a DKMS module install... modifying instructions from @julianstj1

intel-igc-ppsfix-v2.zip

# Install prerequisites
sudo apt install gcc dkms raspberrypi-kernel-headers  # For Raspberry Pi OS
# sudo apt install gcc dkms linux-headers-$(uname -r)  # For Ubuntu

# TODO: SEE LATER COMMENTS FOR LATEST INSTRUCTIONS
wget https://github.com/user-attachments/files/19711659/intel-igc-ppsfix-v2.zip
unzip intel-igc-ppsfix-v2.zip
cd intel-igc-ppsfix

# use dkms to build my patched igc
sudo dkms add .
sudo dkms build --force igc -v 5.4.0-7642.46
sudo dkms install --force igc -v 5.4.0-7642.46

# back up current igc module
sudo cp /lib/modules/$(uname -r)/kernel/drivers/net/ethernet/intel/igc/igc.ko.xz /lib/modules/$(uname -r)/kernel/drivers/net/ethernet/intel/igc/igc.ko.xz.bak

# copy over updated one from where dkms installed itself
sudo cp /lib/modules/$(uname -r)/updates/dkms/igc.ko.xz /lib/modules/$(uname -r)/kernel/drivers/net/ethernet/intel/igc/igc.ko.xz

sudo depmod -a
sudo update-initramfs -u

sudo reboot

After a reboot, I see:

$ dmesg | grep igc
[    2.224403] igc: loading out-of-tree module taints kernel.
[    2.231509] igc 0001:01:00.0: enabling device (0000 -> 0002)
[    2.231533] igc 0001:01:00.0: Julian dev info debug
[    2.231535] igc 0001:01:00.0: PCIe PTM not supported by PCIe bus/controller
[    2.369754] igc 0001:01:00.0 (unnamed net_device) (uninitialized): PHC added
[    2.400924] igc 0001:01:00.0: 4.000 Gb/s available PCIe bandwidth (5.0 GT/s PCIe x1 link)
[    2.400939] igc 0001:01:00.0 eth1: Julian test with pps fix MAC: 00:a0:c9:00:00:02

[geerlingguy]

However, when I try using ts2phc, it fails:

 $ sudo ts2phc -c /dev/ptp0 -s generic --ts2phc.pin_index 2 -m -l 7
ts2phc[4107.479]: config item (null).message_tag is '(null)'
ts2phc[4107.479]: config item (null).verbose is 1
...
ts2phc[4107.479]: PI servo: sync interval 1.000 kp 0.700 ki 0.300000
ts2phc[4107.479]: config item (null).free_running is 0
ts2phc[4107.479]: PPS sink /dev/ptp0 has ptp index 0
ts2phc[4107.479]: PTP_EXTTS_REQUEST2 failed: Operation not supported
failed to initialize PPS sinks

[geerlingguy]

@JulianStj is working on a canonical place for these files to live over in https://github.com/Time-Appliances-Project/Products/tree/main/TimeHAT — I will wait until he's had a chance to test on the Pi before proceeding further!

[geerlingguy]

Some progress—using those instructions (above), I can at least get rising edge working for Pi OS with 6.6.y (but not yet 6.12.y). I will try the following next:

• Install desktop environment + rpi-connect so I can get pygpsclient access
• See if the GPS module is actually outputting NMEA data, force it if not
• See if I can get NMEA working so I can force the system clock to an actual timestamp from GPS!

The current status with ts2phc trying nmea is showing:

Apr 11 16:57:08 mini-time ts2phc[6665]: [678.834] source ts not valid
Apr 11 16:57:09 mini-time ts2phc[6665]: [679.897] nmea: no valid rmc fix
Apr 11 16:57:09 mini-time ts2phc[6665]: [679.897] source ts not valid
Apr 11 16:57:09 mini-time ts2phc[6665]: [679.897] nmea: no valid rmc fix
Apr 11 16:57:09 mini-time ts2phc[6665]: [679.897] source ts not valid
Apr 11 16:57:10 mini-time ts2phc[6665]: [680.663] nmea source timed out
Apr 11 16:57:10 mini-time ts2phc[6665]: [680.959] nmea: no valid rmc fix
Apr 11 16:57:10 mini-time ts2phc[6665]: [680.959] source ts not valid
Apr 11 16:57:10 mini-time ts2phc[6665]: [680.959] nmea: no valid rmc fix

[geerlingguy]

Installed desktop with sudo apt install raspberrypi-ui-mods.

I ran raspi-config and switched to boot to Desktop/GUI, set desktop to auto login, set the compositor to labwc, and rebooted. I also installed rpi-connect and registered my Pi with Pi Connect so I could remotely view it.

Then installed pygpsclient:

sudo apt install python3-pip python3-tk python3-pil python3-pil.imagetk libjpeg-dev zlib1g-dev
# Maybe `python3 -m pip install rasterio` or `sudo apt install python3-rasterio` for automatic extent detection
sudo rm -rf /usr/lib/python3.11/EXTERNALLY-MANAGED
python3 -m pip install --upgrade pygpsclient

Run it:

sudo systemctl stop gpsd
sudo rm /run/gpsd.sock
~/.local/bin/pygpsclient --userport /dev/ttyAMA0

Need to figure out how to verify NMEA output... seems like it's outputting in a UBX format?

[geerlingguy]

Well, after selecting this option from the UBX Config settings, I'm seeing NMEA output:

But restarting ts2phc, it seems like it's not getting the valid timestamps it wants...?

Apr 18 22:35:59 mini-time ts2phc[3853]: [5798.429] nmea sentence: GNRMC,XXXXXX.00,A,XXXX.XXXXX,N,XXXXX.XXXXX,W,0.XXX,,XXXXXX,,,A,V
Apr 18 22:35:59 mini-time ts2phc[3853]: [5798.443] nmea sentence: GNGSA,A,3,80,81,79,,,,,,,,,,4.67,2.31,4.06,2
Apr 18 22:35:59 mini-time ts2phc[3853]: [5798.459] nmea delay: 37227343783588 ns
Apr 18 22:35:59 mini-time ts2phc[3853]: [5798.459] nmea: utc time is past leap second table expiry date
Apr 18 22:35:59 mini-time ts2phc[3853]: [5798.459] source ts not valid
Apr 18 22:35:59 mini-time ts2phc[3853]: [5798.459] nmea delay: 37227343783588 ns
Apr 18 22:35:59 mini-time ts2phc[3853]: [5798.459] nmea: utc time is past leap second table expiry date
Apr 18 22:35:59 mini-time ts2phc[3853]: [5798.459] source ts not valid

Searching online, I found jclark/rpi-cm4-ptp-guide#1 — but I'm not sure if that's actually causing the time sync to fail.

[geerlingguy]

I also tried setting it to CFG-MSG - Enable NMEA, Suppress UBX, and Turn ON all NMEA msgs, but neither of those made a difference (just showed more nmea sentence output than with minimum msgs enabled).

I still get:

ts2phc[7350.847]: nmea delay: 37318687068221 ns
ts2phc[7350.847]: nmea: utc time is past leap second table expiry date
ts2phc[7350.847]: source ts not valid
ts2phc[7350.847]: nmea delay: 37318687068221 ns
ts2phc[7350.847]: nmea: utc time is past leap second table expiry date
ts2phc[7350.847]: source ts not valid

It seems like gpsctl --nmea does the same thing as the GUI app does in terms of setting the UBX into nmea mode.

[geerlingguy]

See update in earlier issue: #14 (comment)

I may be able to use gpsd + gpspipe -r + socat to pipe the data ts2phc requires, in a format it can ingest.