X520 errors in 2.6 and 22.01?
-
So I upgraded from 2.5.2 first to 2.6.0 and then to 22.01. I noticed that immediately after installing 2.6.0 my error in counter for the ix1 interface was something like 150 which I thought was odd but didn't pay any attention to it as I decided to try out 22.01. After rebooting I saw the error counter on ix1 was in the mid 30's which was weird but I thought it was due to some timing thing or other.
This evening I logged in and I see the counter has gone up to 83 on ix1 and I'm seeing 3 on ix0 which is a secondary WAN interface going to a LTE modem.
I'm running pfSense on an HP T730 with 16GB of RAM and an Intel X520 card. My WAN is connected to the onboard Realtek NIC, my LAN is connected to a copper SFP which is read as ix1 and my backup LTE WAN is connected via copper SFP to ix0.
I've been running the same hardware since 2.4.5 and have not had any issues (outside of needing to use the Realtek kernel module). Are these errors a sign of failing card or SFPs? Or is this, more likely, either an issue with the new kernel drivers or some variable that needs to be set?
I'm a newbie when it comes to FreeBSD so not even sure where to start looking.
-
Those seem like pretty small numbers especially if they happened at boot when the link comes up. How many good packets have passed in the time you're seeing 83 bad packets?
If you have an actual bad cable/fibre/module the error rates are usually far higher. Like 10s or 100s of thousands after some GBs have been passed.
If you're not actually seeing connectivity issues I would just keep an eye on it.
Steve
-
@stephenw10 I guess I'm just a little sensitive about any errors after the growing pains of 2.4.5 on the old Realtek driver. Before the system became completely unreachable it would report interface errors like this.
This is what I have as of this morning. A small number of packets with errors but higher than the 0 count over the last year and change.
-
Mmm, it's nice to see all zeros but those numbers wouldn't really worry me especially if they are not climbing steadily. 83 pkts could be a momentary glitch in something.
Here's my edge box:
Note the errors on the WIFI interface. That's an actual WIFI NIC and you expect to see errors.
All the others are actually VLANs over a LAGG of igb NICs. Those errors likely happened when the lagg was coming up or the switch was rebooted etc.Steve
-
Thanks for the info.
Quick followup; is there a difference between Errors In and Errors Out when it comes to worrying about a potential issue?
-
Well errors in could be a cable issue or dirt on the optics etc. That would not normally cause errors out which imply it's unable to send for some reason. SFP over-heated etc. Though in my case it's probably just that the parent interfaces was down.
Steve
-
i also have the same problem after upgrading to 2.6. with the sysctl command I found that the errors are associated with this:
mac_stats.checksum_errs: 41177
it's annoying and I'd like to fix it.
-
So you definitely didn't see those errors in 2.5.2?
41177 is very different to 3170, what makes you think they are related?
Do you have checksum hardware off-loading enabled?
Steve
-
The image is after a reboot. Never an error with the 2.5.2 . The "Hardware Checksum Offloading" is enable by default.
-
Ok, so before that the checksum error count matched?
Try disabling checksum off-loading.
Steve
-
Yes, the value is the same. I tried to disable "Hardware Checksum Offloading" but it didn't work.
thank you for the support
-
Like it didn't disable it or just didn't change the error count?
The errors could be genuine of course. Perhaps the updated driver is now reflecting the correct values.
Steve
-
With a VM i compiled the latest driver ("if_ix.ko" ver3.3.29) for freebsd 12.3 and I put it in the pfsense, but I got nothing!
Any idea?
-
I'm not sure how helpful this will be, but it looks like there were several updates made to the Intel drivers in FreeBSD 12.3 (look under Drives):
https://www.freebsd.org/releases/12.3R/relnotes/
The issue might be upstream from pfSense.
-
@stephenw10 said in X520 errors in 2.6 and 22.01?:
Like it didn't disable it or just didn't change the error count?
^?
Does it show your new driver was correctly loaded at boot? If not does it show an error?
-
The driver is loaded correctly. I think the driver is not the real problem. If it can help, my switch is a hp/aruba JL261A. In the next few days I will try to install some fiber optic transceivers. -
@cescofs said in X520 errors in 2.6 and 22.01?:
I tried to disable "Hardware Checksum Offloading" but it didn't work.
It didn't get disabled or it didn't affect the error count?
-
@stephenw10 Can't speak for @cescofs but all that stuff is disabled on my end:
My error count is relatively small at 1105 with 5 days of uptime and passing through 464GiB of data but it's still higher than the 0 on 2.5.2
-
Mmm, that's a very small number really though. That would not concern me.
Is that still rising? That could all have been when the interface came up for example.Steve
-
@stephenw10
After disabling it, the errors were still there.
Since there are no errors in the output, the drivers should be excluded as a cause.
