Mesa i76e error finishing read. Another Time with LinuxCNC 2.9.1

More
04 Jan 2024 18:23 #289799 by N3pu
Hi,

I read a lot of this issue so sorry for opening another topic but i don´t get it working.

- Using a mesa 7i76e
- Error occurs after 30 minutes to about 2 hours runtime.
- I allready disabled the CPU power safe options (i read in other topics here), see attached
- I allready increased the servo thread time to 2 ms - same issue
- My Ethernetcard is an Intel I219-LM
- CPU is i5 6500T
- Kernel version PREEMPT_RT Debian 6.1.55-1 (2023-09-23)

What I read in some other threads is that I have to disable something in IRQ coalsion Setting for an Intel network card. But I have absolutely no idea how I should do this? This seems to be the only thing I do not did for now. Can somebody introduce me step by step?

BR
Jannik
Attachments:

Please Log in or Create an account to join the conversation.

More
04 Jan 2024 18:35 #289803 by PCW
Is the computer setup for UEFI mode?

(there is another thread using a HP EliteDesk Mini G2 with a I5-6500T that had
network latency issues in legacy BIOS mode but was OK in UEFI mode)

To try disabling coalescing, in a terminal window, type

sudo ethtool -C [your_ethernet_device_name] rx-usecs 0

This will only work until you reboot. to fix it permanently, you edit the interfaces
file. the hm2_eth manual page describes this"

man hm2_eth

 

Please Log in or Create an account to join the conversation.

More
05 Jan 2024 08:38 #289839 by rodw
I would be surprised if PCW's suggestion does not resolve the problem.

Please Log in or Create an account to join the conversation.

More
05 Jan 2024 15:47 #289876 by N3pu
Hard to find out the UEFI / Lagacy issue. In Bios both seems to be activiated, see attached. But I don´t know how the linuxcnc was installed. If computer choosed UEFI / Lagacy. Is there an information in Debian visible? For Windows I found a lot of informations in googel, but for Debian?!

ethtool is an unknown command. Before I wanted to install it (I am a bit afraid of chrashing my current installation because I don´t know what I am really doing in linux sometimes), I want to go on to modify the interface file hm2_eth. I read the docu, but there is no information how to it?

 
Attachments:

Please Log in or Create an account to join the conversation.

More
05 Jan 2024 16:01 #289877 by PCW
If ethtool is not installed, install with

sudo apt-get install ethtool

The instructions for editing the interfaces file is in the hm2_eth manual page:

man hm2_eth

(in a terminal)

 

Please Log in or Create an account to join the conversation.

More
05 Jan 2024 18:15 #289886 by N3pu
Thanks again. I found it and edited the network interface file.
After a restart my network connection disappeared in the graphics overview of debian (right upper corner on desktop). But it is connected.
I think I am using now a different network manager (in background)? Or what I have done?

Ok linuxCNC is still starting and mesa is connected. I will start to let it run tomorrow morning. For overnight my setup is still noch ready for...

Just by the way: Today was a good day without a mesa error finishing read. I don´t really know why. Only thing what was different is that I connected the LAN cable to my second networkcard (which is connected via USB - the USB-Networkcard itself is connected everytime but not the cable to my LAN). This is something I normally not have. It´s now disconnected again to see a different behavior tomorror.

I will report after test tomorrow.

Please Log in or Create an account to join the conversation.

More
05 Jan 2024 18:48 #289888 by PCW
Yes, that connection is no longer managed by the network manager GUI

You should verify that irq coalescing is disabled with

ethtool -c [your_ethernet_device_name]

(because if you have a syntax error in the interfaces file, it
is unreported and will silently fail)

Please Log in or Create an account to join the conversation.

More
06 Jan 2024 18:39 #289985 by N3pu
It´s running perfectly since 10 hours.
I also recheckt the setting with ethtool.

Just a small hint: After installation it wasn´t possible to run the command. You need to sudo ethtool -c [your_ethernet_device_name]

So for now the topic is solved I think.

Many thanks!
The following user(s) said Thank You: rodw

Please Log in or Create an account to join the conversation.

More
21 May 2024 19:00 - 21 May 2024 19:02 #301104 by N3pu
A long time ago the error was gone. Now on my final way to finish my mashine I am noticing this error again more and more. Because I am still setting setting up a lot of things I restarted linuxcnc and focusing on my work. 

Today it happens 2 times in about 15minutes. It happened when I was playing / changing the speed of the spindle drive. So I thought it could be an emi issue. I changed my wiring of the ethernet cable straight from pc to Mesa. Before it was interrupted on the wall of the cabinet with a feed trough connector (this connector was in contact with the cabinet potential earth). Inside was a unscreened 20cm ethernet cable to Mesa. This is now replaced by one screened ethernet cable from pc to Mesa, without using a feed trough connector. 

Few minutes later, after I did this. I got the next error "watchdog has bit" and a lot more faults, see attached. What is the initial fault? The packet loss on ethernet or has the watchdog trip something to do with the smartserial in this case? On smartserial is my handweel. Can I see somewhere on the pc how close I am for triggering one of these errors? If there is option, I will try to Analyse the emi noise and elite it. Spectrum Analyse and HF current probe is available, but I need an indicator when packages?! are lost or the delay increases. ​
Attachments:
Last edit: 21 May 2024 19:02 by N3pu.

Please Log in or Create an account to join the conversation.

More
21 May 2024 19:22 #301110 by PCW
The "Watchdog has bit" error typically indicates major latency issues on the host

What are your ping times if you ping the 7I76E?

ping -i .2 -c 4 10.10.10.10
sudo chrt 99 ping -i .001 -q 10.10.10.10

The first command populates the ARP cache so the ARP lookup doesn't add time to the first ping
The second command should be run for a few minutes (and the stopped with a control C)
 

Please Log in or Create an account to join the conversation.

Moderators: PCWjmelson
Time to create page: 0.098 seconds
Powered by Kunena Forum