"7i92+7i76 on P2: SSerial detects board but registration fails with EAGAIN"

Hi all,
I'm using Claude ai. And it suggested and composed this message, because I am no coder.
...
Good call, and yes — I can't actually post to the forum myself (no login, no ability to submit forms there), but I can draft a clear, complete post for you to paste in. Here's one:

---

**Subject: 7i92 + 7i76 on P2 — "Ignoring soft error during init" then "board fails HM2 registration" (EAGAIN -11)**


Working on a non-Tormach retrofit using PathPilot 2.14.4 (LinuxCNC 2.8.0~pre1) with a 7i92 and a single 7i76. Bitfile is `7i92_7i76x1D.bit`, confirmed via `mesaflash --readhmid` to match the official `.pin` file exactly (5 stepgens, 1 SSerial module, 1 QCount, all on P2; P1 is unused GPIO).

With the 7i76 cabled to **P1**, everything works fine — boots clean, X/Y/Z stepgens register, GPIO pins behave as expected. No SSerial devices show up (expected, since P1 has no SSerial module in this bitfile).

Per the `.pin` file, the 7i76 needs to be on **P2** to access its real field I/O via SSerial. After (removing the ribbon) moving the ribbon cable to P2 (direct pluging to db25) and confirming proper 24V field power + 5V logic power to the 7i76 (jumper on the 7i92 adjusted to supply 5V to P2 instead of P1 — two LEDs now lit on the 7i76 confirming both supplies are good), I get this on every boot attempt:

```
hm2/hm2_7i92.0: Smart Serial Firmware Version 43
Board hm2_7i92.0.7i76.0.0 Hardware Mode 0 = standard
Board hm2_7i92.0.7i76.0.0 Software Mode 0 = io_spin
Board hm2_7i92.0.7i76.0.0 Software Mode 1 = io_ana_spin
Board hm2_7i92.0.7i76.0.0 Software Mode 2 = io_enc_ana_spin_fv
hm2_eth: Ignoring soft error during init
board fails HM2 registration
hm2_eth: rtapi_app_main: Resource temporarily unavailable (-11)
```

So the board is clearly detected correctly (modes print fine) but registration fails right after with EAGAIN. I've tried:
- Removing `sserial_port_0=` entirely (letting it auto-detect per the hostmot2(9) manpage's note that no config should be needed)
- Explicitly setting `sserial_port_0=0xxxxxxx` (mode 0 for the one channel, rest disabled)
- Full power-cycle of both the PC and the 7i92/7i76 (not just a warm reboot)
- Enabling `debug_module_descriptors=1 debug_modules=1` on the hostmot2 load — descriptors parse fine, same failure point every time

`DRIVER_PARAMS`: `board_ip=10.10.10.10 config="num_encoders=1 num_pwmgens=2 num_3pwmgens=0 num_stepgens=5 enable_raw"`

Any idea what "soft error" is actually being ignored here, or what typically causes a valid-looking SSerial handshake to still fail registration with EAGAIN? Happy to post the full boot log if useful.

Thanks!

---

Feel free to tweak tone/details before posting — and if you want, mention you're happy to run any diagnostic commands PCW or others suggest, since that tends to get faster responses on that forum.
...
Thanks!

hm2_eth: rtapi_app_main: Resource temporarily unavailable (-11)

Suggests you have lost Ethernet communications during startup

Could be a network setup issue or perhaps marginal  5V power
that drops out with increased power drain at startup with the 7I76.

What does:

sudo chrt 99 ping -i .001 -c 60000 -q 10.10.10.10

report?    ( It should run for 1 minute then print timing statistics )

( replace 10.10.10.10 with 192.168.1.121 if your 7I92 is jumpered that way )


 

Yes, it's typically used for debugging.

It allows low level register access to FPGA registers.

Post this back to the thread — it's a genuinely useful negative result that narrows things down for PCW. Something like:

Ran the stress test as requested:

60000 packets transmitted, 60000 received, 0% packet loss, time 59998ms
rtt min/avg/max/mdev = 0.105/0.112/0.233/0.011 ms

Clean — no packet loss, tight timing. So it doesn't look like an Ethernet dropout issue. I also removed enable_raw from the config string per tommylight's question, but the same EAGAIN/"Ignoring soft error during init" failure persists.
For what it's worth: I've now reproduced this identical failure with two different bitfiles (7i92_7i76x1D.bit on P2, and 7i92_7i76x2D.bit on P1), on two different SSerial channel instances (.0.0 and .0.2). Both boards' modes print correctly (io_spin/io_ana_spin/io_enc_ana_spin_ff) before the same failure. This is on PathPilot 2.14.4 (LinuxCNC 2.8.0~pre1), if that Tormach fork detail matters.

That last paragraph is worth including since it's new information since your original post — the "reproduced across two different bitfiles/channels" detail might help PCW recognize this as something specific to the PathPilot fork rather than a generic hostmot2/hardware issue.

May be an introduced PathPilot bug since their hostmot2 driver has been
changed from LinuxCNC's version and Tormach doesn't use any sserial devices.

When the error happens, do you get a red LED illuminated on the 7I92?