Latency Test OK, dennoch Fehlermeldung

/etc/network/interfaces has just

auto lo
iface lo inet loopback

everything else I did outcomment for the moment, but I need to have the better performance for the intel card

Ok, what adress to use - 10.10.10.2 or 10.10.10.1?

so, I tried both adresses, with both settings in etc/network/interfaces I have no networks afterwards.

Hiernach geht die Verbindung gar nicht mehr:
    auto eno1
    iface eno1 inet static
    address 10.10.10.2  (same with 10.10.10.1)
    hardware-irq-coalesce-rx-usecs 0

So I tried with a PC with Debian 12 and LinuxCNC 2.9.2 and
everything works as expected. I did setup the real time interface
in the network manager and just added the interfaces file settings
for IRQ coalescing.

OS version

Linux EliteDesk800G3 6.1.0-18-rt-amd64 #1 SMP PREEMPT_RT Debian 6.1.76-1 (2024-02-01) x86_64 GNU/Linux

Debian release:

 

interfaces:

 

ip a result:

 



 

THanks for your simulation check with a new installation, PCW!
OK, it  seems that my network configuration is magically faulty and can not be fixed to work with the lines in etc/network/interfaces, to reduce the ping times to my Mesa card in the normal way.

But, I don't want to install everything including Debian and LCNC completely new.
I think, I have to bite in the sour apple and fix it with the line
sudo ethtool -C eno1 rx-usecs 0   before each and every start of my LCNC Machine. I know that this works.
That wouldn't be comfortable but it works. And I can go further with my project.

My last step was to delete everything than the loop connection 127.0.0.1 from the /interfaces file,
configuring the Intel card with the network manager as before, and using the RTL8111 connection with Auto Internet.

The ip a with this connection is as follows:
1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN group default qlen 1000
    link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
    inet 127.0.0.1/8 scope host lo
       valid_lft forever preferred_lft forever
    inet6 ::1/128 scope host noprefixroute
       valid_lft forever preferred_lft forever
2: enp2s0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc fq_codel state UP group default qlen 1000
    link/ether 00:0a:cd:00:00:56 brd ff:ff:ff:ff:ff:ff
    inet 192.168.178.55/24 brd 192.168.178.255 scope global dynamic noprefixroute enp2s0
       valid_lft 831324sec preferred_lft 831324sec
    inet6 2003:ce:5f08:8600:d50e:a6ab:524:e6ae/64 scope global dynamic noprefixroute
       valid_lft 6923sec preferred_lft 1390sec
    inet6 fe80::8e97:e3df:28fe:e8cf/64 scope link noprefixroute
       valid_lft forever preferred_lft forever
3: eno1: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc fq_codel state UP group default qlen 1000
    link/ether 00:23:24:5b:b4:ac brd ff:ff:ff:ff:ff:ff
    altname enp0s25
    inet 10.10.10.2/24 brd 10.10.10.255 scope global noprefixroute eno1
       valid_lft forever preferred_lft forever



What do you think? Any other solution?

Thanks again for your thougts and hints!
Hanno/arno

Really odd, I did verify that the IRQ coalescing setting worked.
 The only thing I can think of of is that there's a bad character etc
in the interfaces file (you might try using my interfaces file)
This is possible since any errors parsing the file are well hidden

Another possibility is some strange interaction between the
network manager settings and the interfaces file. This could
be tested by deleting the network manager setup for the realtime
connection. I think something may be broken in this interaction
with newer Debian systems as older ones would not allow you to
change settings for an interface that was setup in the interfaces file
(it would be marked as "Unmanaged")

PCW, thanks for your ideas!
Well, I tried copy&Paste with your interfaces file, but no change.

After that, I believe that your idea of smth broken in the interaction of interfaces file and network manager is true. My last try today was to delete (again) any connection in the network manager and restart with the interfaces file. The result was as before - no eno1 connection possible. that's really crazy.

On the other hand, deleting the entries in interface file and managing the eno1 connection I got the connection to the mesa card but bad latency (not so bad but not sufficient). So, I am facing the decision to make a complete new Debian and LCNC installation or living with that.

Q: Is it possible to put the lines for ethtool in a textfile and start it from the desktop (like a batch in DOS/Windows) ? That would make it a little more comfortable. At the moment I don't want to take the lot of trouble with a full new installation.

BTW one more side effect - If I switch the mesa connection with the ethtool to -C eno1 rx-usecs 0 the auto interne connection does not work anymore. Connection is shown, but no function.
Q: Ok, I don't want to surf the internet while cnc'ing, but how can I stop/reset the ethtool enty?

Thanks for your answers!

So you are saying that changing eno1 coalesce settings affects the other interface?
That's extremely weird

Is it possible both interfaces are sharing an interrupt?

(lshw should show this)