Messages

RTL8126 is very solid to me in Windows. There is some CPU utilization but that could be come in part from SMB too. And I paid 30 bugs for two cards. Feels like a no-brainer.

5Gbit/s isn't 10 though...

I take it back about rtl8127. While it isn't giving me any issues, like you had with it per se. After I sent that message yesterday, I decided to compare CPU load on rtl8127. Copying single 50GB file from the unraid server (with AQC113 NIC) throughput was all over the place from 900 MB(yte)/s to 1.10GB/s with one core maxed out. I put the AQC113 NIC back in and got a stable smooth throughput at 1.09GB/s and none of the cores were maxed out. I did play with different settings including settings RSS queues from 8 to 4, but none helped. Is it a driver issue? maybe, but I think for now I'll stick to workarounds with AQC113 (i.e. disabling "Allow the computer to turn off this devices to save power" in Windows) instead of using rtl8127.

[meyergru]

I have x570 motherboard (AsRock x570 Phantom Gaming 4) with 5950x. I used to have AQC113 (non-C) installed into the bottom PCIe x1 slot and it was working fine at Gen4 speeds. The card itself was x4, but AsRock has those open ended slots where you can actually insert any card in them. I also made sure to get AQC113 and not AQC113C or other variant as that the one wiht PCIe Gen4 support. HWINFO64 confirmed it was Gen4 x1 speeds and iperf3 as well. The only problem was that every time I made my computer go to sleep it would come back with a non-working LAN. I had to reboot (or hibernate) for the card to start working again.

I bought rtl8127 card from amazon and it's working flawlessly in the same slot now. No issues at all.

My point, meyergru, it could be just your particular motherboard and not all x570 motherboards out there.

Thank you for trying to help. However 8125b (2.5Gb/s, pcie) is very different from 8153 (Gigabit,USB).

realtek-re-kmod package is for PCI-E only adapters.

Code Select Expand

Realtek PCIe FE / GBE / 2.5G / 5G Ethernet Family Controller
kernel driver.

Also I am not sure if it's exactly the same driver, but you can also install a package from within the GUI from under System -> Firmware -> Plugins -> os-realtek-re

I know 8153b is supported by "ure" driver in Freebsd, implementation is pretty good, it's just the driver doesn't recognise my hardware because there are a ton of different vendor/device ids for the same chip and the Surface adapter may not be on the list. I know in linux I can force a driver to start using a certain vendor/device id through configuration file/at boot. I am not that familiar with *bsd so I was hoping there is a similar mechanism

Hi, I need to add 2 extra ethernet ports to my firewall (a little Intel N100 fanless box with 4 ethernet ports) for low bandwidth-ish devices (PiKVM and a printer). I have 2 MS Surface USB3 gigabit adapters with realtek 8153b chips in them and got them to work with cdc generic driver by settings them to mode 1. The problem is that this driver is pretty bad in terms of maximum bandwidth (250Mbit/s) and latency. realtek 8153b is supposed to be decently supported with the "ure" driver which I am hoping will improve the latency at high load. I suspect that ure driver doesn't get loaded because the device id is just not in the driver code. Is there any way to make the driver load for my device passing device id at runtime (which is possible on linux).

I had these nics running under linux before and they were rock solid with months uptime.

Code Select Expand

ugen0.3: <RTL8153B GigE [Surface Ethernet Adapter] Microsoft Corp.> at usbus0, cfg=1 md=HOST spd=SUPER (5.0Gbps) pwr=ON (72mA)

  bLength = 0x0012
  bDescriptorType = 0x0001
  bcdUSB = 0x0300
  bDeviceClass = 0x0000  <Probed by interface class>
  bDeviceSubClass = 0x0000
  bDeviceProtocol = 0x0000
  bMaxPacketSize0 = 0x0009
  idVendor = 0x045e
  idProduct = 0x0927
  bcdDevice = 0x3100
  iManufacturer = 0x0001  <Microsoft>
  iProduct = 0x0002  <Ethernet Adapter>
  iSerialNumber = 0x0006  <001008A1A>
  bNumConfigurations = 0x0002

Thanks everyone. Running

Code Select Expand

pkg add -f /var/cache/pkg/pkg-1.19.2.pkg
fixed it for me

So this update pkg to 1.20.x and it now gives an error checking for new packages/update. Does anyone know how to rollback to the previous version of pkg. I have the package itself locally from a different system, just can't figure out how to properly downgrade.

4. (Optional) If you want to verify that the updates are working, you can install the package x86info. This is not contained in OpnSense, therefore you have to edit /usr/local/etc/pkg/repos/FreeBSD.conf to enable FreeBSD repositories temporarily. You can use these commands:

Code Select Expand

echo "FreeBSD: { enabled: yes }" > /usr/local/etc/pkg/repos/FreeBSD.conf
echo y | pkg install x86info"
echo "FreeBSD: { enabled: no }" > /usr/local/etc/pkg/repos/FreeBSD.conf
rehash
kldload -q cpuctl
x86info -a | fgrep -i microcode


This is the log

Code Select Expand

root@OPNsense:~ # echo y | pkg install x86info
Updating FreeBSD repository catalogue...
Fetching meta.conf: 100%    163 B   0.2kB/s    00:01
Fetching packagesite.pkg: 100%    7 MiB   6.9MB/s    00:01
Processing entries: 100%
FreeBSD repository update completed. 34062 packages processed.
Updating OPNsense repository catalogue...
OPNsense repository is up to date.
All repositories are up to date.
New version of pkg detected; it needs to be installed first.
The following 1 package(s) will be affected (of 0 checked):

Installed packages to be UPGRADED:
        pkg: 1.19.2 -> 1.20.7 [FreeBSD]

Number of packages to be upgraded: 1

The process will require 21 MiB more space.
9 MiB to be downloaded.

[1/1] Fetching pkg-1.20.7.pkg: 100%    9 MiB   9.0MB/s    00:01    %
Checking integrity... done (0 conflicting)
[1/1] Upgrading pkg from 1.19.2 to 1.20.7...
[1/1] Extracting pkg-1.20.7: 100%
Updating FreeBSD repository catalogue...
FreeBSD repository is up to date.
Updating OPNsense repository catalogue...
pkg: No SRV record found for the repo 'OPNsense'
pkg: packagesite URL error for pkg+http://mirror.sfo12.us.leaseweb.net/opnsense/FreeBSD:13:amd64/23.7/latest/packagesite.pkg -- pkg+:// implies SRV mirror type
pkg: packagesite URL error for pkg+http://mirror.sfo12.us.leaseweb.net/opnsense/FreeBSD:13:amd64/23.7/latest/packagesite.txz -- pkg+:// implies SRV mirror type
Unable to update repository OPNsense
Error updating repositories!
root@OPNsense:~ # echo "FreeBSD: { enabled: no }" > /usr/local/etc/pkg/repos/FreeBSD.conf
root@OPNsense:~ # rehash
root@OPNsense:~ # kldload -q cpuctl
root@OPNsense:~ # pkg update
Updating OPNsense repository catalogue...
pkg: No SRV record found for the repo 'OPNsense'
Fetching meta.conf: 100%    163 B   0.2kB/s    00:01
pkg: packagesite URL error for pkg+http://mirror.sfo12.us.leaseweb.net/opnsense/FreeBSD:13:amd64/23.7/latest/packagesite.pkg -- pkg+:// implies SRV mirror type
pkg: packagesite URL error for pkg+http://mirror.sfo12.us.leaseweb.net/opnsense/FreeBSD:13:amd64/23.7/latest/packagesite.txz -- pkg+:// implies SRV mirror type
Unable to update repository OPNsense
Error updating repositories!

@meyergru I would delete that part before someone else stumbles upon it.

Just ran into the same issues. I had specifically checked all the rules and logging is disabled for every one of them, yet the space was still getting filled up. Thank you for the solution. I limited the logs to 7 days.

I agree, it's a little confusing, but to add phase 2 entry there is a + button on the same line with your phase 1 entry.

Removing "Description" from all Dynamic DNS entries seems to fix this problem.

Here you go.

I have no problem with my upload shaping by the way, but it is only 15Mbit/s pipe.

Where is the link in the Wiki with wrong explanation?

https://docs.opnsense.org/manual/how-tos/shaper.html
I suppose after reading those two topics on the forum + github and none of them were marked solved and I was seeing similar behavior I just assumed it wasn't fixed. After reading the help text for masks I realized mask behaves differently for "Pipe" and for "Queue".

Did you read the Help Text of masks?

I did, as I stated in the second post.

Whenever possible avoid generalisation of "it's broken" assessments. It can keep help away from your thread.

Apologies, duly noted. Topic subject edited.

I'm still having issue with only half of bandwidth available to me. I've now tested with 3 clients at the same time, and when the pipe bandwidth is set to 300, I only can get ~150 (+/- 10) Mbit/s cumulatively on all clients

So after reading the comments, one does not need to set the mask. Somebody needs to edit the wiki to get rid of that. Once mask is unset, it limits the bandwidth for the whole pipe.

The other problem is still there though.

Here's the list of corresponding values that I got from testing.

It is value set under Shaper -> Pipe (downpipe) vs actual value I get from speedtest.
Something is really off, and the results are consistent after many runs.

Pipe  - actual
150 -  80
300 -  150
320 -  200
330 -  250
340 -  ~310 My max connection speed is 300 with initial burst of up to 320

There are 2 issues with it:
- It halves the total bandwidth for every user
- It limits the bandwidth per user, instead of per pipe. Given enough users, the cumulative bandwidth will exceed that of the maximum bandwidth of the pipe.

There are at least 2 topics on this on this forum. One from 2016 and one from 2018. Github issue was closed in 2019 with no resolution (and 1 more problem added while trying to fix the initial one).

Is anyone working on this at all?

https://forum.opnsense.org/index.php?topic=7235.0
https://forum.opnsense.org/index.php?topic=3966.0
https://github.com/opnsense/core/issues/2191

There is an issue report on github: https://github.com/opnsense/plugins/issues/1564