USB Thunderbolt NIC Coldboot Problem

[eracerxrs]

I have a Asus NUC 15 PRO and a QNAP QNA-UC10G1SF USB/TB4 10G SFP NIC.

I can get it up and running after a hot-plug and subsequent warm reboot, but it won't come up when already plugged in on a coldboot.

I have tried on the 6.8 kernel from Prox8 and the 6.14.8-2 kernel from Prox9.

My buddy CGPT and I have been digging for days, and have concluded that force_power is not exposed for the Thunderbolt NHI's, and that is really what would be needed to manually kick these awake.

The thing is that I installed Win11 to see if the problem persisted, but it had no problem bringing up the interface on a cold boot. So how is this achieved?

I have read about others' previous issues that were solved by setting the bios thunderbolt security to "legacy" (which this bios does not have).
But, it looks like it's not even getting to the security check level because "boltctl list" shows the device as authorized, but disconnected.

Barring ASUS modifying the Bios for more granular TB4 control, anyone have any ideas?

 

[codemangel]

Hey,


just made an account for your thread. Since we also bought theses adapter and had same problem. We solved it with allow-hotplug:

iface enp193s0 inet manual
allow-hotplug enp99s0
iface enp99s0 inet manual
        ovs_type OVSPort
        ovs_bridge vmbr2
        ovs_mtu 9000

auto vmbr2
iface vmbr2 inet manual
        ovs_type OVSBridge
        ovs_ports enp99s0
        ovs_mtu 9000

EDIT: Be careful with edditing in the gui, allow-hotplug seems to be not supported?

EDIT2: Actually I only tested warm reboot at the moment, where I had also this problem, will test cold boot tomorrow.

Do you reach max speed with these, at the moment the best I got (with crossover) was 7gbit, to switch connected I get less at the moment.

 

[beisser]

speed on those things highly depends on the USB interface they are connected to.
if its connected to a 40gbit USB-C port or a thunderbolt-port it should reach closer to 9.7-9.8 gbit (according to official reviews)
maybe its connected to something other than usb4/thunderbolt and slower because of that?

 

[codemangel]

for me, its connected to a thunderbolt port on an AMD system.

boltctl list

● QNAP System, Inc QNA-UC10G1SF #2
   ├─ type:          peripheral
   ├─ name:          QNA-UC10G1SF
   ├─ vendor:        QNAP System, Inc
   ├─ uuid:          03034c17-0003-0303-ffff-ffffffffffff
   ├─ generation:    USB4
   ├─ status:        authorized
   │  ├─ domain:     39a23804-e113-2178-ffff-ffffffffffff
   │  ├─ rx speed:   40 Gb/s = 2 lanes * 20 Gb/s
   │  ├─ tx speed:   40 Gb/s = 2 lanes * 20 Gb/s
   │  └─ authflags:  none
   ├─ authorized:    Mi 14 Jan 2026 08:55:30 UTC
   ├─ connected:     Mi 14 Jan 2026 08:55:26 UTC
   └─ stored:        Mo 12 Jan 2026 11:59:15 UTC
      ├─ policy:     iommu
      └─ key:        no

I also had to switch to OVSBridge to get highspeed, testing this with LXC container at the moment. Is it true, that a VM would perform better here? I read thie somewhere...

 

[DerekG]

Barring ASUS modifying the Bios for more granular TB4 control, anyone have any ideas?

Have you checked for the latest USB4 update from ASUS.? (it appears to be separate from the BIOS firmware)

I'm having a problem with a USB4 egpu dock, which I've worked on for weeks, and my system (Aoostar WTR MAX), uses the same USB4 controller, the ASM4242. ASUS recently released a firmware update or both the PD system and the USB4 interface, so it might be worth investigating that.

Unfortunately I can't risk that update as it might be ASUS specific, but looking online, it's seems that a number of users have already applied it.

On completely separate topic, Realtek have released a USB3.2 to 10Gbe NIC, at a great cost. I've only tested it on Windows so far and the performance seems OK via USB4 (not extensively tested, just iperf3 tests).