realtek-re-kmod 1102.01 regression: RTL8125 does not recover after physical link

Started by MSchmidt1973, July 14, 2026, 07:59:18 PM

Previous topic - Next topic

Hello,

I have found a reproducible regression in the current "realtek-re-kmod" package when using an RTL8125 2.5 GbE network adapter.

Environment

* OPNsense "26.1.11_6-amd64"
* Fujitsu Futro S520
* Additional Realtek RTL8125 2.5 GbE adapter
* Interface name: "re1"
* "os-realtek-re-1.0"
* Current driver package: "realtek-re-kmod-1102.01.1403000"

The RTL8125 adapter is not detected without the "os-realtek-re" plugin and its "realtek-re-kmod" dependency.

Problem description

After booting OPNsense with the Ethernet cable already connected, the interface works normally.

The problem occurs after physically disconnecting and reconnecting the cable:

1. Boot OPNsense with "re1" connected to the switch.
2. Confirm that the interface and LAN connectivity are working.
3. Disconnect the Ethernet cable from "re1".
4. Reconnect the cable.

After reconnecting:

* "ifconfig re1" shows the "LOWER_UP" flag.
* "ifconfig re1" reports "status: active".
* The "RUNNING" flag does not return.
* No corresponding "re1: link state changed to UP" message is logged by "dmesg".
* The interface remains unusable and cannot be pinged.

Running the following command is sufficient to restore the interface:

ifconfig re1 up

No preceding "ifconfig re1 down" is required.

Immediately afterwards:

* the "RUNNING" flag returns,
* a link-state message is logged,
* and LAN connectivity is restored.

Version comparison

Downgrading only the driver package with:

opnsense-revert -r 26.1.10 realtek-re-kmod

installs:

realtek-re-kmod-1101.00.1403000

After rebooting with this version, the problem disappears completely.

With "1101.00.1403000", reconnecting the cable causes the interface to recover automatically after approximately one or two seconds. The "RUNNING" flag returns and connectivity is restored without any manual command.

I also removed all previously added tunables and a temporary "devd" workaround before repeating the tests.

The issue is reproducible every time on this hardware.

This therefore appears to be a regression introduced between "realtek-re-kmod" versions "1101.00" and "1102.01".

There appear to be a number of issues with the latest updates.  Feel free to report them to https://bugs.freebsd.org/


Cheers,
Franco