Bad Ethernet Performance on Newer 465 CPU revision

[sdefisher]

I have been trying to troubleshoot an odd ethernet issue. It's been narrowed down to the CPU board proper. I have 3 CPU cards, only one has good ethernet performance. This board has a revision number 604045-106. The boards that are not working well are 604045-107. We use GPASCII and SENDGETSENDS with the ethernet communction. And they are not pushing data though at the same rates with the new boards. Is there any configuration on these that would affect them?

 

I have all 3 boards running the same firmware (2.3.2.5), and same project downloaded by the same version of the IDE (in-case the complier is differnt. As far as I can tell all boards are also running the same backend linux kernel version as well (3.2.21-powerpmac-smp). All of the CPUs were acquired as part of a full UMAC system at various dates/times in the last few years.

 

Update: The boards that don't work were setup with IDE version 4.5.2.9 and the working one was done with 3.1.1.0.  Re-downloaded my project via 4.5.2.9 into the good board and now that CPU no longer performing well. The symptom seems to stem from GPASCII not transmitting the ACK character. Which is was previously doing without fail. Would 4.5.2.9 have anything to do with this? By default it's should be on, I don't see how the PowerPMAC Ide would change this. Is there a GPASCII switch to force the ACK to be active? I know there is one to turn it off.

Edited September 26, 2022 by sdefisher

[steve.milici]

There is no setting in PMAC to suppress the <ACK> character. My educated guess is that you have some other issue causing communications problems in the “bad” PMACs.

The IDE would not be the source as it only “downloads” the project code. There may have been an issue when the project was “upgraded” with the later IDE version though. You may want to compare it with the original.

This would be a technical support issue. Please contact ODT's direct technical support with your full issue details at: ODT-Support@Omron.com

 

[sdefisher]

I found my issue. It was rather taxing. Related to the following post. I think the ACK was a false problem. The main issue I later determined was my GPASCII instances were being killed by the PowerPMAC on the linux side.

 

mount -o remount,rw /
nano /.readonly/etc/ssh/sshd_config
[Move to end of file with arrow keys]

[Remove All instances of MaxSessions, ClientAliveInterval, and ClientAliveCountMax]

[Ctrl-X, Y, ENTER]
mount -o remount,ro /

 

Below is what my sshd_config looked like after downloading my project with the latest version of the PowerPMAC IDE. It doesn't happen with the older revision. I noticed you can attempt to set  MaxSessions, ClientAliveInterval, and ClientAliveCountMax in the newest version, but it seems buddy and doesn't actually save (from what I have found)

UsePAM yes
UseDNS no
MaxSessions 256
ClientAliveCountMax 3
ClientAliveInterval 15
MaxSessions 256
MaxSessions 256
MaxSessions 256
MaxSessions 256
MaxSessions 256
MaxSessions 256
MaxSessions 512
MaxSessions 256
MaxSessions 512
MaxSessions 256
MaxSessions 256
MaxSessions 256
MaxSessions 256
MaxSessions 256

 

I confirmed this by using WinSCP to download the readonly folder from a working machine then compared the configs.

 

Edited September 27, 2022 by sdefisher

[Eric Hotchkiss]

I am attaching an application note on how these values are used and how to set them in PMAC. It may be that MaxSessions was added repeatedly because a step was missed. I hope this is helpful.

Configuring an SSH Timeout In Power PMAC.pdf