Oh, that's great, that will create a customized image with the drivers integrated? Thanks !
Too bad I didn't know that before though.
yep... seems to be very popular for this board...
Popular Profiles
Counts of each profile, by firmware release...
SNAPSHOT:
linksys_e8450-ubi: 1304
rpi-4: 131
linksys_wrt1900acs: 62
tplink_archer-c60-v1: 60
21.02:
rpi-4: 137
xiaomi_mi-router-4a-gigabit: 125
Can we get the stats for drivers?
In particular and if that's possible, I'd like to know the most used drivers for USB to Ethernet adapters?
interesting question... afaik what you are looking at probably covers 90% of usb3-eth in that order...
- asix would typically represent maybe 70% 'outside' of OpenWrt
- for arm in particular UE300/rtl8152(3) is popular here so it probably represents 65-ish percent in OpenWrt-usb3 terms
( other adapters are rare or 1G+ )
290, "kmod-usb-net"
169, "kmod-usb-net-cdc-ether"
140, "kmod-usb-net-rtl8152" ***
109, "kmod-usb-net-rndis"
82, "kmod-usb-net-qmi-wwan"
74, "kmod-usb-net-cdc-ncm"
71, "kmod-usb-net-huawei-cdc-ncm"
62, "kmod-usb-net-ipheth"
60, "kmod-usb-net-cdc-mbim"
56, "kmod-usb-net-sierrawireless"
53, "kmod-usb-net-asix-ax88179" ***
30, "kmod-usb-net-cdc-subset"
27, "kmod-usb-net-cdc-eem"
24, "kmod-usb-net-smsc95xx"
24, "kmod-usb-net-rtl8150" **
24, "kmod-usb-net-hso"
24, "kmod-usb-net-asix" *
23, "kmod-usb-net-sr9700" *
23, "kmod-usb-net-pl"
23, "kmod-usb-net-pegasus" *
23, "kmod-usb-net-mcs7830"
23, "kmod-usb-net-kaweth"
23, "kmod-usb-net-kalmia"
23, "kmod-usb-net-dm9601-ether" *
11, "kmod-usb-net-aqc111" **
( cc: @richb-hanover great suggestion )
Thank you !
May I ask where you got this information from?
For ARM in particular ... means (I didn't check) that most routers are based on the RISC architecture, right? No royalties to pay after all ...
Off topic, but I can't see the devices that were previously connected to a WiFi AP in my LAN, I'll have to investigate to see what's wrong ...
Anyway, I really appreciate all of you for the help you provided me, looks like a friendly community here .
Really, actually i have the same problem, omg !!!
Seems like the ax88179 is really power hungry, so better give it some juice !
hahaha okey i will give
Hello,
I finally bought this Y cable . But something is wrong:
Looks like the internet connection is capped at USB 2.0 speed, which is obviously not what I wanted.
The male USB 2.0 end is connected to a 5V 2A charger, like when I make the test previously with the USB 3.0 hub.
lsusb seems to confirm USB 2.0 speed, right?
Now, what puzzles me, is that on another linux computer (Linux Mint), I connected a USB 3.0 SSD (USB key form factor) using the same Y cable, and compared that to the USB SSD attached directly to a USB 3.0 port on the PC.
I obtained virtually the same results !
That is, about 300 MB/s using the following command:
dd if=/dev/zero of=/media/root/1877-6625/output bs=1M count=10k; rm -f /media/root/1877-6625/output
So, the Y cable apparently allows transfers at USB 3.0 Gen 1 speeds (5 Gbps), I don't get it why on the RPI 4 that doesn't work. Of course, I double checked that I used one of the USB 3.0 port on the PI with the USB 3.0 male end of the Y cable.
Any idea?
curl -sSL https://asu.aparcar.org/api/v1/stats/packages/SNAPSHOT | jq | grep -B1 kmod-usb-net
Thank you.
I had to modify the command a bit on a RPI to get a result identical to the output you have previously quoted, but that was a god starting point anyway:
root@raspberrypi:~# curl -sSL https://asu.aparcar.org/api/v1/stats/packages/SNAPSHOT | jq '.' | grep -B1 kmod-usb-net | sed '/--/d' | sed 's/\s//g' | sed -r '/^[0-9]+,/N;s/\n/ /'
301, "kmod-usb-net"
179, "kmod-usb-net-cdc-ether"
146, "kmod-usb-net-rtl8152"
111, "kmod-usb-net-rndis"
90, "kmod-usb-net-qmi-wwan"
81, "kmod-usb-net-cdc-ncm"
78, "kmod-usb-net-huawei-cdc-ncm"
66, "kmod-usb-net-cdc-mbim"
64, "kmod-usb-net-ipheth"
62, "kmod-usb-net-sierrawireless"
55, "kmod-usb-net-asix-ax88179"
32, "kmod-usb-net-cdc-subset"
29, "kmod-usb-net-cdc-eem"
26, "kmod-usb-net-hso"
26, "kmod-usb-net-asix"
25, "kmod-usb-net-smsc95xx"
25, "kmod-usb-net-rtl8150"
24, "kmod-usb-net-sr9700"
24, "kmod-usb-net-pl"
24, "kmod-usb-net-pegasus"
24, "kmod-usb-net-mcs7830"
24, "kmod-usb-net-kaweth"
24, "kmod-usb-net-kalmia"
24, "kmod-usb-net-dm9601-ether"
11, "kmod-usb-net-aqc111"
root@raspberrypi:~#
Hello,
Since the Y cable didn't do the trick, I have ordered an externally powered USB 3.0 hub, I will report back my results here.
Just to be safe, I also got a TP-Link UE300 USB 3.0 to Gigabit Ethernet adapter.
Do you want me to run some specific tests to compare both options?
Is it possible to switch from one to the other without rebooting?
And independently of any reboot, what is the fattest way to modify the WAN interface to reflect the current adapter used?
My ISP requires the WAN interface to have a given MAC address.
Perhaps the easier method is to use the CLI and comment out the part that is irrelevant to the actual configuration benchmarked?
if usb = wan = eth1
- make sure you've setup the mac override
- make sure the
kmod-usb-net-rtl8152
kmod installed - power off switch! adapters (only boot with one connected ever)
when you power up the router, the new nic will claim eth1 providing the kmod/driver is installed, the mac override will still be applied based on that...
no additional config editing is needed, just make sure the kmod is installed...
(90% confident the above would work without a reboot... but it's not something to mess around with if you dont have console/kbd-mon)
at the moment... no special tests/bench's, the primary tasks is to ensure absence of crashing
a) with powered usb hub (+ asix)
b) with a different nic (will be ok)
however... feel free to post additional results/tests as i'm sure there will be more to learn here...
OK, thank you.
usb = wan = eth1
kmod-usb-net-rtl8152 kmod installed
root@OpenWrt:~# opkg list-installed | grep -Ei 'ax|rtl'
kmod-usb-net-asix-ax88179 - 5.4.154-1
kmod-usb-net-rtl8152 - 5.4.154-1
root@OpenWrt:~#
power off switch! adapters (only boot with one connected ever) => OK, I will do that
But just to make sure I understand this part:
For example, since lsusb
would list a different USB device (Ethernet adapter in our case) depending on which one is connected, why would OpenWRT consider it should apply the same network configuration regardless of the plugged in device?
Don't get me wrong, if that works just like that, that's even more convenient than what I imagined .
because it claims eth1... and all (wan) config is tied to this already
yes.. if you only ever have 1 additional usb nic + have the driver it is good...
OK, so that will be automatic?
No need to configure anything the first time I connect the UE300?
OK, so the first test, was performed with the powered USB hub (5V 2A in), and the asix ax88179 chipset based adapter.
I ran a batch download, that simultaneously fetched up to 1000 files, with a total of 970 files. The 3 first files are 1 GB each in size, the rest consists of various ipk files (I copied all the links of the .ipk files from this page, so that excludes the first few ones that are other types files' extensions).
The Packet Steering option was enabled.
The 4 cores of the poor RPI 4 were maxed out during the beginning of the torture test (probably for the time it took to get all ipk files, then there were only the 3 biggest files left to complete so the RPI could breathe a little).
The RPI 4 was running with all cores clocked at 1.5 GHz (default speed, I tried OCing them to 2 GHz, but that made virtually no difference, so the bottleneck was probably not the SOC). force_turbo
was enabled to keep them constantly at their frequency of 1500 MHz, I hoped this would help reduce variability between differents tests.
root@OpenWrt:~# wc -l download.txt
970 download.txt
root@OpenWrt:~#
root@OpenWrt:~# head download.txt
http://ipv4.appliwave.testdebit.info/1G.iso
http://ipv4.scaleway.testdebit.info/1G.iso
http://ipv4.k-net.testdebit.info/1G.iso
https://downloads.openwrt.org/releases/21.02.1/targets/bcm27xx/bcm2711/packages/base-files_1434-r16325-88151b8303_aarch64_cortex-a72.ipk
https://downloads.openwrt.org/releases/21.02.1/targets/bcm27xx/bcm2711/packages/bcm27xx-gpu-fw_2021-02-16-ba6259246c702b04ea56ff1034325e476d460ae8_aarch64_cortex-a72.ipk
https://downloads.openwrt.org/releases/21.02.1/targets/bcm27xx/bcm2711/packages/bcm27xx-userland-dev_97bc8180ad682b004ea224d1db7b8e108eda4397-1_aarch64_cortex-a72.ipk
https://downloads.openwrt.org/releases/21.02.1/targets/bcm27xx/bcm2711/packages/bcm27xx-userland_97bc8180ad682b004ea224d1db7b8e108eda4397-1_aarch64_cortex-a72.ipk
https://downloads.openwrt.org/releases/21.02.1/targets/bcm27xx/bcm2711/packages/block-mount_2021-01-04-c53b1882-1_aarch64_cortex-a72.ipk
https://downloads.openwrt.org/releases/21.02.1/targets/bcm27xx/bcm2711/packages/blockd_2021-01-04-c53b1882-1_aarch64_cortex-a72.ipk
https://downloads.openwrt.org/releases/21.02.1/targets/bcm27xx/bcm2711/packages/comgt-directip_0.32-33_aarch64_cortex-a72.ipk
root@OpenWrt:~#
root@OpenWrt:~# time xargs -P 1000 -n 1 curl -4 -o /dev/null < download.txt
...
real **0m26.592s**
user 0m44.372s
sys 0m37.487s
root@OpenWrt:~#
So, it took 26.592s to download a little bit over 3 GB of data.
root@raspberrypi:~# wget https://downloads.openwrt.org/releases/21.02.1/targets/bcm27xx/bcm2711/packages/
--2021-12-24 20:57:58-- https://downloads.openwrt.org/releases/21.02.1/targets/bcm27xx/bcm2711/packages/
Resolving downloads.openwrt.org (downloads.openwrt.org)... 168.119.138.211, 2a01:4f8:251:321::2
Connecting to downloads.openwrt.org (downloads.openwrt.org)|168.119.138.211|:443... connected.
HTTP request sent, awaiting response... 200 OK
Length: unspecified [text/html]
Saving to: ‘index.html.1’
index.html.1 [ <=> ] 207.80K --.-KB/s in 0.08s
2021-12-24 20:57:59 (2.50 MB/s) - ‘index.html.1’ saved [212790]
root@raspberrypi:~#
root@raspberrypi:~# echo "$(grep ipk index.html.1 | grep -Po '[0-9]+\.?[0-9]* KB' | sed 's/ KB//g' | tr '\n' '+' | sed 's/\+/ \+ /g' | sed -rn 's/^(.*) \+ $/\1/p')" | bc
25417.7
root@raspberrypi:~#
The ipk files contribute to add precisely 25417.7 KB of data, i.e. a little bit over 25 MB (that's less than 1% overall).
So, the average download speed, when considering all the files, is almost 114 MB/s.
Now, the same test, but this time the TP-Link UE300 adapter was used instead (directly attached to the RPI 4, no USB 3.0 hub in between):
real 0m26.487s
user 0m45.030s
sys 0m31.585s
root@OpenWrt:~#
So, that looks quite similar, even equivalent in practice.
By the way, that indeed worked out of the box, as you told me, thanks ! (I just rebooted in the meantime, I'm not sure that was really necessary, but I could not hurt either way).
The main difference between the 2 adapters, is that while using the Ugreen adapter (Asix), it seems that generated a lot more networking errors or packets dropped than the TP-Link.
This topic was automatically closed 10 days after the last reply. New replies are no longer allowed.