What to do about Realtek RTL8125 RTL8126 RTL8127 drivers
-
@olivierlambert Realtek has released three related ethernet PCI chips: RTL8125 (2.5G), RTL8126 (5G), RTL8127 (10G).
Realtek started with a single driver that supported the 8125 and 8126 as they are closely related. The current XCP driver is based on that single set of code. Now they have split the code into three different drivers.
The problem is they have released new versions of the RTL8125 that require new driver code to work. This already caused problem in the mainline upstream Linux and required updates to the kernel driver (early 2025) and at least one XCP user has hit this problem on a desktop board. Since XCP is using the vendor driver, the question is what to do next:
- Do nothing and let people use other ethernets
- Make a new alt-driver for only the 8125 chip (no 8126 support), leave the old 8125 driver alone (maybe for 8.2?)
- Use the new Realtek 8125 code and add new drivers for the 8126 and 8127 (maybe for 8.3?)
- ???
The path forward could be different for XCP 8.3 vs. 8.2.1 ...
-
@Andrew I bought a RTL8126 5G card (because of the price) for my home lab that doesn't appear to be working with 8.3 like you said. It is trying to use the r8125 driver.
[08:53 tw-xcp-ng-02 ~]# ethtool -i eth2 driver: r8125 version: 9.012.04-NAPI-PTP-RSS firmware-version: expansion-rom-version: bus-info: 0000:02:00.0 supports-statistics: yes supports-test: no supports-eeprom-access: no supports-register-dump: yes supports-priv-flags: noI removed the r8125 driver and manually installed the r8126 driver which I downloaded from here. I can see the interface using the driver, and it shows the 5G capabilities of the driver, however I cannot get this to negotiate the speed (see below). I have tried this with a 2.5G switch and a 10G switch. The 10G switch can negotiate down to 5G so that shouldn't be an issue. On both switches it shows its connection as just fast ethernet. If I manually set the port on the switch to 5G I will not even get a link light.
Would you happen to have any suggestions or ideas to get the card working?
[09:49 tw-xcp-ng-02 ~]# ethtool -i eth2 driver: r8126 version: 10.015.00-NAPI firmware-version: expansion-rom-version: bus-info: 0000:02:00.0 supports-statistics: yes supports-test: no supports-eeprom-access: no supports-register-dump: yes supports-priv-flags: no [09:46 tw-xcp-ng-02 ~]# ethtool eth2 Settings for eth2: Supported ports: [ TP ] Supported link modes: 10baseT/Half 10baseT/Full 100baseT/Half 100baseT/Full 1000baseT/Full 2500baseX/Full 2500baseT/Full 5000baseT/Full Supported pause frame use: Symmetric Receive-only Supports auto-negotiation: Yes Supported FEC modes: Not reported Advertised link modes: 10baseT/Half 10baseT/Full 100baseT/Half 100baseT/Full 1000baseT/Half 1000baseT/Full 2500baseX/Full 2500baseT/Full 5000baseT/Full Advertised pause frame use: No Advertised auto-negotiation: Yes Advertised FEC modes: Not reported Speed: Unknown! Duplex: Unknown! (255) Port: Twisted Pair PHYAD: 0 Transceiver: internal Auto-negotiation: on MDI-X: on Supports Wake-on: pumbg Wake-on: d Current message level: 0x00000033 (51) drv probe ifdown ifup Link detected: no[09:52 tw-xcp-ng-02 ~]# lspci | grep Ethernet 00:1f.6 Ethernet controller: Intel Corporation Ethernet Connection (17) I219-LM (rev 11) 02:00.0 Ethernet controller: Realtek Semiconductor Co., Ltd. Device 8126 (rev 01) 03:00.0 Ethernet controller: Intel Corporation Ethernet Controller I225-V (rev 03) -
@snowstruk The stock XCP 8.3 r8125 driver works for my 8125 and 8126 devices. I have not tried it a 5G, so it's possible the 8126 is not working at 5G (I don't see it advertised) with the current driver.
I have run the current dedicated realtek 8125 driver and it works but I have not tried the 8126 driver. I don't know how many people are using the 8126 chipset (I guess not many).
If your device is not working at >1G then it sees to be a hardware incompatibility issue. You could try loading the new 8126 firmware to see if that helps.
Here's the output about my test 8126 card (attached to a 2.5G switch):
# lspci -v -s 04:00.0 04:00.0 Ethernet controller: Realtek Semiconductor Co., Ltd. RTL8126 5GbE Controller (rev 01) Subsystem: Realtek Semiconductor Co., Ltd. Device 0123 Flags: bus master, fast devsel, latency 0, IRQ 18 I/O ports at 3000 [size=256] Memory at e0d00000 (64-bit, non-prefetchable) [size=64K] Memory at e0d10000 (64-bit, non-prefetchable) [size=16K] Expansion ROM at e0d20000 [disabled] [size=64K] Capabilities: [40] Power Management version 3 Capabilities: [50] MSI: Enable- Count=1/1 Maskable+ 64bit+ Capabilities: [70] Express Endpoint, MSI 01 Capabilities: [b0] MSI-X: Enable+ Count=32 Masked- Capabilities: [d0] Vital Product Data Capabilities: [100] Advanced Error Reporting Capabilities: [148] Virtual Channel Capabilities: [170] Device Serial Number 01-00-00-00-b4-b3-c9-88 Capabilities: [180] #19 Capabilities: [190] Transaction Processing Hints Capabilities: [21c] Latency Tolerance Reporting Capabilities: [224] L1 PM Substates Capabilities: [234] Vendor Specific Information: ID=0002 Rev=4 Len=100 <?> Kernel driver in use: r8125 Kernel modules: r8125# ethtool -i eth1 driver: r8125 version: 9.012.04-NAPI-PTP-RSS firmware-version: rtl8126a-2_0.0.2 02/01/24 expansion-rom-version: bus-info: 0000:04:00.0 supports-statistics: yes supports-test: no supports-eeprom-access: no supports-register-dump: yes supports-priv-flags: no# ethtool eth1 Settings for eth1: Supported ports: [ TP ] Supported link modes: 10baseT/Half 10baseT/Full 100baseT/Half 100baseT/Full 1000baseT/Full 2500baseX/Full Supported pause frame use: Symmetric Receive-only Supports auto-negotiation: Yes Supported FEC modes: Not reported Advertised link modes: 10baseT/Half 10baseT/Full 100baseT/Half 100baseT/Full 1000baseT/Full 2500baseX/Full Advertised pause frame use: Symmetric Receive-only Advertised auto-negotiation: Yes Advertised FEC modes: Not reported Link partner advertised link modes: 10baseT/Half 10baseT/Full 100baseT/Half 100baseT/Full 1000baseT/Full 2500baseX/Full Link partner advertised pause frame use: Symmetric Receive-only Link partner advertised auto-negotiation: Yes Link partner advertised FEC modes: Not reported Speed: 2500Mb/s Duplex: Full Port: Twisted Pair PHYAD: 0 Transceiver: internal Auto-negotiation: on MDI-X: Unknown Supports Wake-on: pumbg Wake-on: g Current message level: 0x00000033 (51) drv probe ifdown ifup Link detected: yes -
I've had this thread in my TODO list for long.
I'm not sure what the different options we have mean for users.
As
r8125-moduleis installed by default, we'll want to avoid breaking it for existing users. Would using the new code risk causing regressions? Maybe that's what you meant but I'm not sure. Does our currentr8125driver support both 8125 and 8126, and not 8127? Or just 8125?If we can't update the driver without causing regressions, then can we offer an alternate driver based on the new code for each chip (8125, 8126, 8127), and what amount of work might this represent?
The target would be XCP-ng 8.3. In XCP-ng 9.0 we'll have a newer kernel and newer drivers anyway.
CCing @Team-Hypervisor-Kernel for their opinion.
-
@stormi Unless new Dom0 is Kernel 6.15 or newer (with updates), this will be an ongoing problem to support the newest hardware as they seem to keep updating the chips (you can't stop progress).
The current
r8125driver in 8.2/8.3 supports the 8125 and some 8126 chips, but not 8127. Other than supporting the new hardware versions of the chips, I have not seen any real complaints posted about the currentr8125driver (may be there are some Vates tickets?)While the risk on the new driver is low, it is different (not just an update). I agree that it would be best to have
r8125-module-altfor the new code and thenr8126-moduleandr8127-module.Note, that the new
r8126-moduledriver should not be installed without ther8125-module-altas the existingr8125-modulewould conflict with the newr8126-moduledriver.The work is already done (compiled and running)... I'm just waiting for Realtek to fix some issues with the new code before it's worth fully testing the drivers (issues already acknowledged by Realtek).
