Looking forward to build and test later. Anything to look at? Fan control? Trying some SFP(+) modules?
Testing some SFPs would be great, it's still a bit spooky that most SFP stuff works fine on my RTL9313 devices while others don't seem to work. I only tested with 10GBase-R, 1000Base-X and SGMII modules, if you verify this and have some more that would help.
Regarding fan control, I didn't do anything yet. The fan is annoying but it wasn't obvious to me how fan control might be done on that device, if it's even possible. Feel free to contribute something if you know more already 
Just installed 25.12-SNAPSHOT r32602 on my D-Link DGS-1210-16 and switching speed is back to normal. I get again 920-930 MBit/s in a bidirectional iperf test.
Thanks again to all involved parties for this quick analysis and fix! 
2 Likes
Would the USW-Flex-2.5G-8-PoE be able to use OpenWrt?
With RTL8239 on the PoE board, RTL8372N appears to be the main 2.5G switching chip and RTL8261BE1 is the 10gb SFP+ controller.
I don't have this switch myself, but saw a tear down here https://www.reddit.com/r/Ubiquiti/comments/1iga7tn/uswflex25g8poe_teardown/
AFAIK that SoC can't run OpenWrt.
Ubiquiti has a 10G switch out though that very probably runs on an RTL93xx SoC, if you like Ubiquiti stuff. Pricey though.
You are right. It only has a 8051 8-bit microcontroller and can barely run the most minimal rtos. If you are lucky ubiquiti used the option to pair it with an external cpu somewhere under the poe board, because the switch appears to be (somewhat) managed and has more ports than the RTL8372N can provide on its own.
What is the current state of the XGS-1250-12? Is that functional now? I tried it before and it was pretty broken. So I went back to stock. I currently have a BPi-R4 (used as router) connected via a SFP+ DAC 10gb cable (H!Fiber.com 10Gb/s SFP+ DAC Cable 3m - 10GBase-CU Passive Direct Attach Copper Twinax) to the XGS-1250-12. Then via the RJ-45 ports at 10GB, 2.5 GB, 2x 1 Gbit , and 2x 100Mbit to different devices/APs.
Thank you
I have an XGS1250-12 A1 (with the aquantia PHYs) and everything just works(tm) on 25.12 RC4.
4 Likes
Sorry it took me so long but I have uploaded pictures of the RAM and Flash, as well as the rear of the board, for the ZyXEL GS1900-8HP Rev B2.
Hopefully everything is legible: the original image I took was too high resolution to upload (I did my best to reduce size without losing resolution using cropping) so the flash may be a little hard to read. I added the text to the image description but am not certain of exact model numbers. There is also a slight scratch (marking) on the flash chip so I could not tell if the final two characters were G0, 60, or C0.
1 Like
It's not my daily driver here anymore (that's an XGS1210-12 B1 with 2x SFP+ now) but I use it as a testbed and still seems to work fine.
Realtek target has come a long way since and it's seeing a lot of development.
3 Likes
Looking at the last year, the Realtek target is the most active target by far from a development perspective.
| target | commits |
|---|---|
| mxs | 7 |
| tegra | 7 |
| ixp4xx | 8 |
| malta | 8 |
| uml | 8 |
| zynq | 8 |
| bcm4908 | 10 |
| omap | 10 |
| sifiveu | 10 |
| microchipsw | 11 |
| qoriq | 11 |
| loongarch64 | 12 |
| pistachio | 12 |
| octeon | 14 |
| bcm47xx | 15 |
| at91 | 16 |
| kirkwood | 16 |
| bcm53xx | 17 |
| d1 | 17 |
| armsr | 18 |
| econet | 19 |
| stm32 | 19 |
| imx | 20 |
| layerscape | 22 |
| apm821xx | 26 |
| sunxi | 26 |
| gemini | 29 |
| mpc85xx | 30 |
| starfive | 31 |
| siflower | 36 |
| x86 | 39 |
| bmips | 44 |
| ipq806x | 45 |
| mvebu | 53 |
| lantiq | 60 |
| qualcommbe | 61 |
| bcm27xx | 75 |
| rockchip | 85 |
| ipq40xx | 106 |
| airoha | 110 |
| ath79 | 136 |
| ramips | 174 |
| qualcommax | 186 |
| mediatek | 319 |
| generic | 366 |
| realtek | 708 |
In case someone wants to replicate this, here is the script:
for a in target/linux/*; do git log --format=%cd --after=$(date -d "now -1 year" +%F) $a|wc -l|xargs echo $a; done|sort -k 2 -g
5 Likes
FWIW I'm currently using the XGS 1250-12 with with an ipolex SFP+ (both Base-T and Base-S tested). Have also tested with an SFP+ equivalent ipolex DAC. Both work at 10Gbps on 25.12 RC1+. Did not work on 24.10.5.
1 Like
How do I see if I have a rev A1 or B1?
For B1, you should see "Rev. B1" written on the label on the bottom of the device, probably in the lower right corner. If there's nothing, it should be A1.
1 Like
ok thx, i have an A1 (I expected that because at the time i 1st tried installing OpenWrt there was no distinction, but just wanted to make sure). Anyway I upgraded in the meanwhile and everything seems to work fine.
Thx all, including devs that did all the hard work!
Maybe I spoke slightly too soon, im getting these:
[ 5417.496754] rtl838x-eth 1b00a300.ethernet eth0: Ring contention: r: 0, last a280007c, cur a2800078
Any idea how to fix?
I don't know what causes this and if it's harmful. Though, I'm seeing this on all my switches after some time and it doesn't cause any issues to me.
seems to be a ring buffer check for traffic going to the CPU, so should not affect the switching. But maybe this causes packet loss for CPU traffic? I guess that not the post important thing for a switch. Im going to leave it as is to see if it give any issues. Up till now I did not see any negative consequence. I did notice that Luci is a bit on the slow slide, but that could be the CPU itself being slower than e.g. the BPi-R4 CPU.
As we are going to receive one more RC for 25.12.
Which commits need or should be picked to 25.12 from the last weeks?
I would prep a PR but need your input?
The obvious crasher bug fixes are already merged AFAICS. Other than that, we could probably transplant the whole of git main for realtek to get better support. But I'm not sure if that is acceptable or even a good idea. (Not saying it's a bad idea, just totally unsure.) It would help people wanting more switches supported in the release, though.
2 Likes