26.1 SSH and WebUI hangs up with Broadcom NIC, ifconfig with high CPU

[HD-Lux]

Hi,

I'm new to OPNsense and started my first setup with 26.1 as a fresh install on a Intel based desktop computer.
The system hangs up after a few minutes when I install a DualPort Broadcom BCM57416 Copper NIC.

OPNsense 26.1-amd64
FreeBSD 14.3-RELEASE-p7
OpenSSL 3.0.18
bnxt driver enabled

The machine has the NIC setup:
1 DualPort Intel x550
1 DualPort Broadcom BCM57416*
1 onboard Realtek

*
Part Number                         : BCM957416A4160
PCI Address                         : 0000:01:00.0
Chip Number                         : BCM57416
Chip Name                           : WHITNEY+
Description                         : Broadcom BCM57412 2x10GbT PCIe Ethernet NIC
Firmware Name                       : CHIMP_FW
Firmware Version                    : 236.1.126.0 (latest)
RoCE Firmware Version               : 236.1.126.0

The system is just installed, nothing configured except enabling the bnxt driver and SSH.
No WAN connection, only static LAN. No rules etc.
The Intel and BCM NICs are unconfigured just there, unconfigured. I use the onboard Realtek for management.

After Reboot all is fine. After some minutes (less than 15) SSH reacts slowly. The WebUI does not respond to navigation and live charts stop. "failed to load widget".
If I'm lucky I can see in top in my SSH session, that one or more ifconfig processes run with high CPU. Then at some point SSH will no longer respond. The LAN IP of the machine is still pingable though. I have not tested if the machine would route traffic because I have no networks attached except LAN.

root@OPNsense:~ # top

last pid: 26013;  load averages: 12.94,  3.41,  1.29 up 0+00:13:1811:54:25
101 processes: 11 running, 90 sleeping
CPU:  1.0% user,  0.0% nice,  8.1% system,  0.0% interrupt, 90.9% idle
Mem: 423M Active, 145M Inact, 560M Wired, 56K Buf, 6640M Free
ARC: 197M Total, 40M MFU, 120M MRU, 941K Anon, 1463K Header, 34M Other
     129M Compressed, 276M Uncompressed, 2.13:1 Ratio
Swap: 8192M Total, 8192M Free

  PID USERNAME    THR PRI NICE   SIZE    RES STATE    C   TIME    WCPU COMMAND
37452 root          1  23    0    17M  3188K iflib    2   0:07  38.09% ifconfig
46262 root          1  23    0    17M  3196K iflib    3   0:07  15.97% ifconfig
62435 root          1  23    0    17M  3168K iflib    3   0:03  15.87% ifconfig
66915 root          1  23    0    17M  3192K iflib    3   0:01   9.47% ifconfig
35358 root          1  23    0    17M  3188K iflib    2   0:09   7.76% ifconfig
48535 root          1  99    0    17M  3184K RUN      1   0:03   4.69% ifconfig
18758 root          3  20    0    45M    14M RUN      3   0:00   0.88% syslog-ng
  482 root         19  40    0   111M    40M accept   3   0:02   0.00% python3.11
97519 root          1  20    0    57M    31M select   3   0:01   0.00% php-cgi
97027 root          1  20    0    53M    29M select   3   0:00   0.00% php-cgi
 3008 hostd         6  20    0    51M    14M uwait    0   0:00   0.00% hostwatch
95586 root          1  20    0    53M    29M select   3   0:00   0.00% php-cgi
62546 unbound       4  20    0    90M    48M kqread   1   0:00   0.00% unbound

After a powercycle the behavior repeats.

With the broadcom NIC removed the machine is stable (at least for the time I'm testing with it).

[franco]

Is that because of ifconfig -v perhaps?


Cheers,
Franco

[HD-Lux]

I may have run the command in that SSH session:
ifconfig | grep -i flag

ix0: flags=8843<UP,BROADCAST,RUNNING,SIMPLEX,MULTICAST> metric 0 mtu 1500
ix1: flags=1008843<UP,BROADCAST,RUNNING,SIMPLEX,MULTICAST,LOWER_UP> metric 0 mtu 1500
re0: flags=8843<UP,BROADCAST,RUNNING,SIMPLEX,MULTICAST> metric 0 mtu 1500
lo0: flags=1008049<UP,LOOPBACK,RUNNING,MULTICAST,LOWER_UP> metric 0 mtu 16384
enc0: flags=0 metric 0 mtu 1536
pflog0: flags=0 metric 0 mtu 33152
pfsync0: flags=0 metric 0 mtu 1500
bnxt0: flags=8802<BROADCAST,SIMPLEX,MULTICAST> metric 0 mtu 1500
bnxt1: flags=8802<BROADCAST,SIMPLEX,MULTICAST> metric 0 mtu 1500

but I have seen that CPU on ifconfig also just running top after the reboot.

[franco]

Yes, try with "-v" to see if that's the issue. It's not "standard" but it's needed for SFP module information amongst other things which sometimes misbehaves in drivers because it's not really being tested.


Cheers,
Franco

[HD-Lux]

both NICs are RJ45, no SFP modules involved.
now I have the BCM NIC removed. If I run the -v command now, all is fine.
I have never used the -v parameter in my tests before.

[HD-Lux]

so to conclude: I put the BCM NIC into the box, let it run and it hangs in <15 min. This happens without any interaction or SSH / Webadmin connection.

[franco]

Ok, then the Broadcom driver has general quality issues. Not much we can do.


Cheers,
Franco