Mesa hm2/hm2_7i96s.0: error finishing read

You're a lifesaver Tommy.

Ok, clean ISO from LCNC home page + Rod's instructions page.  Fingers crossed...

Still stumped, still getting read finishing errors.
[using MeLE N5105 w/realtek 8168 NIC.  I've given up on the HP 800 G2]

Checked 7i96s - yes, there's a red light when the error happens.
Currently I'm trying Deb13, pulled from Debian site.
Built 6.6rt kernel from kernel.org using JT's build instruction page.  Surprisingly no errors.

Installed LCNC using Andy's script. LCNC latency was fine. Still got read error

Switched to r8168 driver using Rod's instructions. Still got read error

Found a site describing similar NIC drops that was solved by changing "EEE" to disabled. I suspect this is an energy saving feature.
forums.debian.net/viewtopic.php?t=154817

Ping times terrible with r8168 (~1.5ms), then I stumbled across this post:
forum.linuxcnc.org/ethercat/50322-etherc...-with-minipcs#282701

Once I added those to grub, ping times got really good. Still got read error

Changed IRQ affinity to 3 with the sudo echo "8"... command. Still got read error

Changed servo thread to 2ms. Still got read error

At a loss at this point.  I've attached HAL and INI just in case there's something horribly stupid in there that's causing all this.

I just bought a Elite 800 D2 mini to try
Probably next week.

Should I try a different power supply for the Mesa?  I'm using a buck converter but it's not drooping on the built-in 5v volt display.

Are there any hardware configs or problems which would result in error finishing reads or the intermittent read-time spikes seen in halscope?

The bios has about a million settings, about three of which I understand.  But if it's something in there I don't get why both the realtek and intel NIC's appear to be acting the same.

If there was a 7I96s power issue, you would not get the 7I96s
watchdog bite indication (the red LED)

It did take me a while on the Zotac CI323 to get all the BIOS settings.
correct but then it worked with the stock 2.9.1 ISO and stock RTL8168
driver. I think it's still possible there is some BIOS setting issue involved.

Even if it's not a BIOS issue, I would like to chase down the problem with
the G2 and if not solve, at least clearly identify the actual problem.


AFAICT, its not related to anything on the LinuxCNC side


As an experiment, you might try setting the packet error limit to say 100
(setp hm2_7i96s.0.packet-error-limit 100) and watch hm2_7i96s.0.packet-error-total
to see when the errors occur.

The errors are happening so fast I cant get halscope open and set up before the error happens.

Tried setting error limit to 1000 and it's essentially immediate.

I found that I can get halscope to auto-start with 'loadusr' in hal file, but not sure how to call up a particular halscope file or get the pins set to capture everything immediately.
 

Managed to catch it.  This with error-level set to 1000.
Does not appear to be intermittent errors piling up, but a failure to commuicate from the start?
 
 

EDIT
Set error-level to 10000 and exactly the same thing happened, just took longer to trigger.

This feels like some sort of setting I've got wrong and not some intermittent communications error.

Yes, that looks like it simple times out every packet
(so about 1/2 second to get the 500 time out errors needed to trip the read error)

Maybe is is worth looking into the configuration a bit,

Can you try a minimal configuration? (nothing but axis stepgens for example)

Slight change with minimal config - errors stop briefly, then accumulate again.

Entire HAL file is in upper left of screenshot.