Only one USB dongle RTL8822BU is working if we connect two same dongles to the same ARM processor

[madhug89]

We are using RTL8822BU usb dongles in our wireless router products. One dongle will be used as STA + AP (i.e 2.4GHz) mode and another dongle will be used for 5GHz band AP.

By using this driver, we can use only one dongle operational at a time. Is there a way to operate both dongles simultaneously?

[morrownr]

@madhug89

By using this driver, we can use only one dongle operational at a time. Is there a way to operate both dongles simultaneously?

It is unlikely to work well. I certainly would not do it in a commercial product. My experience over many years is that it works well to have many adapters connected to a system... as long as all adapters use different drivers. This is particularly true of Realtek chipset based products. There may be an exception here and there but I am not aware of any exceptions.

With that said, I regularly run multiple usb wifi adapters with a single computer. Up to 4 at a time in fact. It can be very very stable. I just make sure that I do not have more than one adapter with the same chipset. I think I remember doing a test with 2 mt7612u adapters at some point and it worked but I would have to retest as my memory is not as good as it used to be. Let me know you want me to test this.

Regards,

@morrownr

[madhug89]

Mediatek Chipset adapters works well. But we are using RTL8822BU dongles connected to the same system.

Only one dongle is working at a time.

[morrownr]

Mediatek Chipset adapters works well.

That statement indicates that you have tested Mediatek based adapters. Now I am wondering why we are having this conversation.

But we are using RTL8822BU dongles connected to the same system.

Yes, you made that clear in the original post. Let me rephrase my reply. That is a REALLY BAD IDEA.

I run a dual band router that is based on an ARM processor. I made it. It uses two usb wifi adapters. It is rock solid.Why is it rock solid? Because I know what works. I'll be glad to make some suggestions if you share some details of your plan. My suggestions will not include using two rtl8822bu adapters in the same setup because that is a REALLY BAD IDEA.

@morrownr

[madhug89]

Thanks for your quick responses !!
We are using ARM based system, which runs the Linux Kernel of version-3.4.113. Initially we are using WIFI4 based dongle RT5572 in the SoftAP mode and RTL8822BU in the Mesh/STA mode. Since the Ralink modules are not available in the market, we want to use the both dongles as RTL8822BU for SoftAP, Mesh and STA connections for the single system at a time.

Even though its a bad idea, is there any way to make it work for the single system with two RTL8822BU?

[morrownr]

which runs the Linux Kernel of version-3.4.113

Oh, my. Basing a WiFi product on a kernel this old? Planning on support WPA3? Basing a WiFi product on a kernel that old is a security nightmare. I would not use a kernel older than 5.4 but newer would be better.

is there any way to make it work for the single system with two RTL8822BU?

If you want to spend a lot of money on a developer to rewrite a lot of code, possibly. There is no practical way for this to happen that I am aware of.

[bjlockie]

Do 2 Mediatek adapters with the same chipset work together?

[morrownr]

@bjlockie

I just added a second adapter with a mt7921au chipset to my homemade wifi router:

$ iw dev
phy#1
	Interface wlan1
		ifindex 5
		wdev 0x100000001
		addr 00:c0:ca:b3:7a:bb
		type managed
		txpower 3.00 dBm
		multicast TXQ:
			qsz-byt	qsz-pkt	flows	drops	marks	overlmt	hashcol	tx-bytes	tx-packets
			0	0	0	0	0	0	0	0		0
phy#0
	Interface wlan0
		ifindex 4
		wdev 0x1
		addr e0:e1:a9:35:c5:1f
		ssid myPI-WiFi6
		type AP
		channel 36 (5180 MHz), width: 80 MHz, center1: 5210 MHz
		txpower 3.00 dBm
		multicast TXQ:
			qsz-byt	qsz-pkt	flows	drops	marks	overlmt	hashcol	tx-bytes	tx-packets
			0	0	0	0	0	0	0	0		0

A second interface came up. I'll need to do some reconfiguring and activate it as an AP to give it a good test. I think I will set the new adapter to the 2.4 GHz band since that kind of setup is what the OP is talking about.

[madhug89]

Since the MTK drivers are part of the kernel source tree, so if we connect both dongles to the same system, then both dongles works flawlessly.

But the problem with the RTL drivers only as they are not part of the kernel source tree. We are using the external driver code for RTL8822BU dongles in our system. For us also both physical interfaces are up.

root@madhug-linux ~ # iw dev
phy#1
Interface wlan1
ifindex 4
type managed
phy#0
Interface wlan0
ifindex 3
type AP
root@madhug-linux ~ # ethtool -i wlan0
driver: rtl88x2bu
version: 3.4.113
firmware-version: N/A
expansion-rom-version:
bus-info: 1-1:1.2
supports-statistics: no
supports-test: no
supports-eeprom-access: no
supports-register-dump: no
supports-priv-flags: no
[email protected] ~ # ethtool -i wlan1
driver: rtl88x2bu
version: 3.4.113
firmware-version: N/A
expansion-rom-version:
bus-info: 3-1.1:1.2
supports-statistics: no
supports-test: no
supports-eeprom-access: no
supports-register-dump: no
supports-priv-flags: no
root@madhug-linux ~ # uname -a
Linux madhug-linux 3.4.113 #1 SMP Sat Apr 13 20:28:21 IST 2024 armv7l GNU/Linux

root@1madhug-linux ~ # ping www.google.com
PING www.google.com (142.250.179.196): 56 data bytes
64 bytes from 142.250.179.196: seq=0 ttl=54 time=8.415 ms
64 bytes from 142.250.179.196: seq=1 ttl=54 time=8.053 ms
64 bytes from 142.250.179.196: seq=2 ttl=54 time=8.820 ms
^C
--- www.google.com ping statistics ---
3 packets transmitted, 3 packets received, 0% packet loss
round-trip min/avg/max = 8.053/8.429/8.820 ms

root@madhug-linux ~ # cat /proc/net/rtl88x2bu/wlan0/tx
tx_aclt_confs tx_ampdu_density tx_max_agg_num tx_power_idx tx_quick_addba_req txbf_cap
tx_aclt_flags tx_bw_mode tx_power_by_rate tx_power_limit tx_rate_bmp txpwr_total_dbm
tx_aclt_force_val tx_info_msg tx_power_ext_info tx_power_offset tx_stat
root@madhug-linux ~ # cat /proc/net/rtl88x2bu/wlan0/tx_stat

[madhug89]

@morrownr

Is it good idea to use one RTL8822BU (i.e WIFI-5 based dongle) and one RTL8852CU (i.e WIFI-6 based) dongle connected to the single ARM based system?

[morrownr]

Since the MTK drivers are part of the kernel source tree, so if we connect both dongles to the same system, then both dongles works flawlessly.

I am testing. I burned a new sd with RasPiOS 2024-03-15, kernel 6.6. The hardware is a RasPi4B with 2 usb wifi adapters based on the mt7921au chipset. The driver is mt7921u which is in-kernel and has been since kernel 5.18. The chipset/driver are tri-band WiFi 6 but I will use the 2.4 and 5 Ghz bands.

My testing so far shows the 2 interfaces below. I am connected and have used APs on both interfaces. No problems noted so far but this is a quick and dirty test. A longer term test is needed. I will leave this setup going and will continue to use it over the next few days to see if I can find any problems. You can ping me here when you want an update.

$ iw dev
phy#1
	Interface wlxe0e1a935c51f
		ifindex 5
		wdev 0x100000001
		addr e0:e1:a9:35:c5:1f
		ssid myPI-WiFi4
		type AP
		channel 6 (2437 MHz), width: 20 MHz, center1: 2437 MHz
		txpower 3.00 dBm
		multicast TXQ:
			qsz-byt	qsz-pkt	flows	drops	marks	overlmt	hashcol	tx-bytes	tx-packets
			0	0	0	0	0	0	0	0		0
phy#0
	Interface wlx00c0cab37abb
		ifindex 4
		wdev 0x1
		addr 00:c0:ca:b3:7a:bb
		ssid myPI-WiFi6
		type AP
		channel 36 (5180 MHz), width: 80 MHz, center1: 5210 MHz
		txpower 3.00 dBm
		multicast TXQ:
			qsz-byt	qsz-pkt	flows	drops	marks	overlmt	hashcol	tx-bytes	tx-packets
			0	0	5042	0	0	0	0	1780310		5050

[morrownr]

Is it good idea to use one RTL8822BU (i.e WIFI-5 based dongle) and one RTL8852CU (i.e WIFI-6 based) dongle connected to the single ARM based system?

My current recommendation to all Linux users is to avoid usb wifi adapters based on Realtek's WiFi 6 chipsets. This includes the following chipsets:

rtl8852/32au
rtl8852/32bu
rtl8852/32cu

I have a driver for the rtl8852/32bu up here. I had previously tried to get a Realtek out-of-kernel driver up for the rtl8852/32au chipset. That driver terrible and I have no intention of ever going to back to work on it. The driver that I have up is not a good driver. I tell people in the README that basically I am only supporting managed mode and I try to convince them to avoid adapters with the chipset. Do I expect the rtl8852/32cu to be any better? I doubt it. Realtek has used the WEXT era driver model way beyond what was a good idea so these WiFi 6 driver are really really bad.

The good news is that the Linux kernel management is putting an end to Realtek and their way out dated WEXT era drivers. There is code in the Linux kernel these days that will detect WEXT use for WiFi 7 drivers and it will stop the driver in its tracks so Realtek is at the end of the line. They are being forced to do something thay should have done many years ago. On the other hand, Mediatek is so far ahead of Realtek with their usb wifi support that it is hard to believe. Mediatek's support the mt7925 chipset, which WiFi 7, is already in the Linux kernel and is standards compliant... modern standards compliant. We are simply waiting for products to ship.

[madhug89]

Thanks @morrownr for your suggestion.

Right now our system supports only usb based wlan modules and is running based on very old kernel-3.4.113.

Our product is a dual band based wireless router with Mesh support. We are in the hunting & evaluation stage of usb based modules.

I evaluated MTK7921au dongles in the last yr with 6.1.xx kernel. The dual band is working fine with some performance issues.

Do you have any usb based wlan module suggestions otherthan MTK7921?

[morrownr]

Do you have any usb based wlan module suggestions other than MTK7921?

mt7612u. Solid driver. Solid AC1200 chipset.

I evaluated MTK7921au dongles in the last yr with 6.1.xx kernel. The dual band is working fine with some performance issues.

Do you mind sharing the issues? The tremendous complexity of the WiFi 6 and 7 drivers is incredible. Patches continue to roll in for the mt7921u driver. Test it on kernel 6.6 and see if you see a difference... and make sure you use up-to-date firmware for your testing. I have a menu item for firmware upgrades.

running based on very old kernel-3.4.113.

There have been massive updates to the wifi stack in Linux over the last 5-6 years. You really need to think about going with a newer kernel or you will have to deal with little incompatibilities all over the place.

[madhug89]

Do you mind sharing the issues?

I tried with the kernel-6.3.5, I am facing the below stability issues -

1. Device connection with the internet is unstable in the managed mode.
2. Mesh connection between the two devices is not happening always

[madhug89]

mt7612u. Solid driver. Solid AC1200 chipset.

This wlan module is also EOL from the suppliers. Do you know the suppliers of this component?

Note: We are not looking the end product dongles. We are looking for the components to assemble them as a part of our production.

[morrownr]

Do you know the suppliers of this component?

My opinion is that the mt7612u chipset may be EOL by a specific maker but I think it should be available in usb module format for several years as it has the single best Linux driver in the WiFi 5, AC1200 category. Here are the links for a site that I check at times:

https://www.aliexpress.us/item/3256805689944637.html?spm=a2g0o.productlist.main.51.69ef46fdrIphyD&algo_pvid=2ba61ffd-9bf0-41ee-8626-7fe74550947a&algo_exp_id=2ba61ffd-9bf0-41ee-8626-7fe74550947a-25&pdp_npi=4%40dis%21USD%2120.00%2114.60%21%21%2120.00%2114.60%21%402101ef8717131535731136307edde0%2112000034666194263%21sea%21US%210%21AB&curPageLogUid=bVrslbCX0lew&utparam-url=scene%3Asearch%7Cquery_from%3A

https://www.aliexpress.us/item/3256802419694070.html?spm=a2g0o.productlist.main.65.69ef46fdrIphyD&algo_pvid=2ba61ffd-9bf0-41ee-8626-7fe74550947a&algo_exp_id=2ba61ffd-9bf0-41ee-8626-7fe74550947a-32&pdp_npi=4%40dis%21USD%2140.00%2118.00%21%21%2140.00%2118.00%21%402101ef8717131535731136307edde0%2112000021366024999%21sea%21US%210%21AB&curPageLogUid=i6CMNTrdJ29l&utparam-url=scene%3Asearch%7Cquery_from%3A

https://www.aliexpress.us/item/3256806443807364.html?spm=a2g0o.productlist.main.71.69ef46fdrIphyD&algo_pvid=2ba61ffd-9bf0-41ee-8626-7fe74550947a&algo_exp_id=2ba61ffd-9bf0-41ee-8626-7fe74550947a-35&pdp_npi=4%40dis%21USD%2117.45%2116.58%21%21%21126.00%21119.70%21%402101ef8717131535731136307edde0%2112000037870861938%21sea%21US%210%21AB&curPageLogUid=4tZoOHY45dhZ&utparam-url=scene%3Asearch%7Cquery_from%3A

I tried with the kernel-6.3.5, I am facing the below stability issues -

Device connection with the internet is unstable in the managed mode.

This is about the mt7921au (I have to remind myself). This is a surprise. I'm trying to remember any managed mode stability issues that have been reported and I am coming up blank. Were you keeping the firmware updated? I'd be interested in the details as I would like to see if I can find more info or duplicate it.

Mesh connection between the two devices is not happening always

II do nothing with mesh so I would have to spin up to speed to see what is going on there. How would I duplicate this problem?

I just happened to think that while this conversation is interesting, if you must stick with kernel 3.4, most of what we are talking about does not matter. Backporting any of these modern drivers to that kernel would be a challenge. Well, I understated that. It would be unpractical.

[dubhater]

I think the problem with using two identical Realtek dongles is that the out of tree drivers have some global variables. The in kernel drivers should not have this problem because, I was reminded recently, global variables are a big no no in the kernel. On the other hand, rtw88_8822bu doesn't work well in AP mode yet.

Maybe you can identify the global variables and rename them, make two copies of the driver.

The good news is that the Linux kernel management is putting an end to Realtek and their way out dated WEXT era drivers. There is code in the Linux kernel these days that will detect WEXT use for WiFi 7 drivers and it will stop the driver in its tracks so Realtek is at the end of the line. They are being forced to do something thay should have done many years ago.

I'm pretty sure the wifi 4 and 5 drivers with version 5.x.y.z don't use wext anymore. The wifi 6 drivers seem to have a newer codebase, with the version number starting from 1. They do use cfg80211 instead of mac80211, which is not ideal, but they are not affected by the wext deprecation.

Only the older drivers abandoned earlier still use wext, e.g. rtl8723au, rtl8192cu, rtl8192du.

[morrownr]

@dubhater

I think the problem with using two identical Realtek dongles is that the out of tree drivers have some global variables.

I investigated this about 2 years ago. It did not pan out as there appears to be more to the problem.

rtw88_8822bu doesn't work well in AP mode yet.

Agree. AP mode has not cooked enough in the in-kernel driver yet. Then there is the problem with getting the in-kernel driver working on kernel 3.4.

I'm pretty sure the wifi 4 and 5 drivers with version 5.x.y.z don't use wext anymore.

Noted. I think this may be a per driver thing. Last time I looked I saw the "wext detected, be advised wext support will not be allowed starting with WiFi 7."

The wifi 6 drivers seem to have a newer codebase, with the version number starting from 1. They do use cfg80211 instead of mac80211, which is not ideal, but they are not affected by the wext deprecation.

Realtek's out-of-kernel usb drivers for WiFi 6 cause me to start saying many bad things. Those drivers are bad and in no way could be of use to the needs of the OP. The difference between Realtek's WiFi 6 drivers and Mediatek's is frankly astounding. If it were practical to backport mt7921u to kernel 3.4, then the OPs problem would likely be solved.

[bjlockie]

AP mode has not cooked enough in the in-kernel driver yet.

I wonder if there will be much demand for AP mode.
I don't think there are many community members working on the Realtek in-kernel driver. :-)

[morrownr]

@bjlockie

I wonder if there will be much demand for AP mode.

Yes, there already is high demand for AP mode. I get direct emails from companies working on commercial products. They often use usb wifi modules instead of adapters but it is the same driver. The most heavily used modes are AP and STA. There are a lot of cool products out there... robotics, automation and more and they need to communicate.

I don't think there are many community members working on the Realtek in-kernel driver. :-)

I can't argue with this observation. I'll add that it appears there are far fewer community members working on the out-of-kernel Realtek drivers than there were maybe 5-6 years ago. My best guess is that Mediatek has disrupted this market and many users of usb wifi adapters have moved over to adapters and modules based on Mediatek chips.