wifi: mt76: mt7915: BW160 and BW80+80 add support back!

[CornelCotoara]

Please revert this removals regarding MT7915, MT7915 works just fine with HE160, been using this for over a month when I got this powerful client. Please see my tests in the comment below. https://github.com/openwrt/mt76/commit/98aa346042bd21fe8c0b1594736edca16be436db#commitcomment-99775400 @nbd168 @MeiChia.Chiu

[simontretter]

Same issue here. Have been using HE160 on mt7915e (AW7915-NP1 card) with 2gbit link speeds until support was taken out (now iw only reports HE80 capabilities).

[DragonBluep]

HE160 has serious compatibility issues with Apple devices. Unless someone fix it, the 160 MHz won't come back.

[CornelCotoara]

Wow, compatibility issues with Apple devices, that explains !all (mind the hairpin). Btw, who's forcing you to even use Aple or HE160 if you have major issues with? I have done tests on my side, HE160 has a huge gain against HE80 on my particular setup, so yes, for me it's a must have. The hardware supports HE160 so every owner should be able to use that if needed. Here's a small quote from OpenWrt repo itself,: `` For developers, OpenWrt is the framework to build an application without having to build a complete firmware around it; for users this means the ability for full customization, to use the device in ways never envisioned.

Sunshine! ``

[ThiloteE]

fix "what". Could you link to the issues?

[CornelCotoara]

I can link the tests performed, they are placed as comments to the HE160 cutoff commit. https://github.com/openwrt/mt76/commit/98aa346042bd21fe8c0b1594736edca16be436db#commitcomment-99775400 https://github.com/openwrt/mt76/commit/98aa346042bd21fe8c0b1594736edca16be436db#commitcomment-100460501

[DragonBluep]

fix "what". Could you link to the issues?

https://forum.openwrt.org/t/802-11ax-worse-than-802-11ac-with-mt76-driver/126466

[ThiloteE]

fix "what". Could you link to the issues?

https://forum.openwrt.org/t/802-11ax-worse-than-802-11ac-with-mt76-driver/126466

Thank you for your answer.

Just trying to summarize a little what we know here.

1. 👍 The first comment mentioning that the new snapshot fixed "802-11ax-worse-than-802-11ac-with-mt76-driver" for rt3200 was https://github.com/openwrt/mt76/issues/709#issuecomment-1414438364 and since then multiple people have voiced that this update fixed rt3200 for them

   This was right after mt76: update to the latest version, import WED related mtk_eth_soc patches was merged. This update of the mt76 driver indeed contained the commit "98aa346042bd wifi: mt76: mt7915: remove BW160 and BW80+80 support".

2. 👎 Unfortunately, it is not quite as clear, since this update contained A LOT of other commits as well...

3. 👍 👎 Also, there are comments like these, which show that 160MHz can improve performance in certain situations, but decrease performance in others:

   Is the slow down observed happening to just Apple devices? @dsouza

   I recently, took a spare RT3200 and used it as a client only just like a phone or a laptop. When I placed it behind the couch several walls away from the Redmi AX6000 AP, I could not reproduce the single digit iperf3 values that my iPhone 13 was reporting.

   I am wondering if the AX slow down is only affecting Apple hardware.

   Example 1. Redmi AX6000 on channel 149/80MHz width.

   iphone 13: 6 Mbps RT3200 client: 604 Mbps

   Example 2. Redmi AX6000 on channel 100/160MHz width.

   iphone 13: 515 Mbps RT3200 client: 153 Mbps

   I switched channels/width a total of 3 times. The results are consistent.

4. 👍 👎 This report points to other cause why performance might deteroriate with 160 MHz in certain circumstances, but reports better performance with 160 MHz overall.

   I have great performance on 160mhz on my rt3200 as well. Would be a shame if that is removed. It is true that it drops down to 2+2 instead of 4+4 spatial streams. But that's not such a big downside since not many clients are 4+4, nor do I experience high throughout from multiple 2+2 devices at the same time.

   160mhz isn't strictly better, but in plenty of use cases it was better so I am sad to see it go

To sum it all up: it probably should be investigated if it was really the removal of the 160 MHz that fixed things for some people, or if it was one of the other commits...

[CornelCotoara]

It should definitely be investigated, as you mentioned there were many code changes in this regard included in the build of the openwrt snapshot image. I'm no programming expert, but I find it hard to believe that butchering the ability to use HE160 for MT7915 (and related) miraculously solved the "AX" speed problems. Personally, I don't have and haven't encountered these problems using the AX in HE160 mode (I've been using it this way for about 2 months already). Correct me if I'm wrong, but having these capabilities available does not mean that they are used in default mode (does /etc/config/wireless have a default, right?), the choice is definitely up to the user.

[riKirsti]

Based on my research those codes came from wifi: mt76: mt7915: remove BW80+80 support and [v2] wifi: mt76: mt7915: remove BW160 and BW80+80 support by MeiChia Chiu (邱美嘉) MeiChia.Chiu@mediatek.com and Money Wang (王信安) Money.Wang@mediatek.com. They seem to be the employees of MediaTek.

On the first patch, MeiChia Chiu answered why.

On 12/28/22 6:11 PM, MeiChia Chiu wrote: mt7915 doesn't support bw80+80. You are removing a bunch of 160Mhz support flags as well, was that on purpose? From what I can tell, 7915 does support 160Mhz in 2x2 mode. Thanks, Ben

While on the second patch, MeiChia Chiu made some changes.

MeiChia Chiu MeiChia.Chiu@mediatek.com writes: Remove BW160 and BW80+80 capability in mt7915. This patch is already applied, so too late to change it, but I comment anyway as this is important: The commit log should always answer to the question "why?" and explain the reason behind the patch. For example, I have no clue why you are removing these features from mt7915.

Since this is an open-source driver for OpenWRT. We don't need to obey the rules from MediaTek.

[nbd168]

I will look into adding back support for 160 MHz, possibly guarded by a module parameter

[nbd168]

Please test this patch: https://nbd.name/p/7945d62a

[joaochainho]

Hi @nbd168 , many thanks for providing the patch. I've tested it on my RT3200 and M1 Mac, and I didn't notice any significant decrease in performance with HE80 (base channel = 36). However, if I select HE160, there's a noticeable decrease in performance, namely upload, but not as bad as before. Note: the M1 Mac only supports HE80.

Hope it helps.

[nbd168]

It's normal that performance decreases in HE160 mode if your client doesn't support it, since that limits the available number of tx/rx chains to 2. I'd like to get some test feedback from people using HE160 with clients that support it.

[CornelCotoara]

I need a hint about how the apply the patch , I've successfully setup a build container using alpine linux, but the image i build is missing the patch.

wget https://nbd.name/p/7945d62a -O patches/01_restore_HE160.patch
quilt import patches/01_restore_HE160.patch
quilt push
Applying patch 01_restore_HE160.patch
can't find file to patch at input line 5
Perhaps you used the wrong -p or --strip option?
The text leading up to this was:
--------------------------
|diff --git a/mt7915/init.c b/mt7915/init.c
|index ec297e047fe8..4f9188d825e1 100644
|--- a/mt7915/init.c
|+++ b/mt7915/init.c
--------------------------
No file to patch.  Skipping patch.
4 out of 4 hunks ignored

[nbd168]

You could place the patch as package/kernel/mt76/patches/100-restore_HE160.patch and rebuild.

[CornelCotoara]

I applied the patch, I can select HE160 but my client get only 300/300Mbps! HE80 acts the same way, only 300Mbps both ways.

 -----------------------------------------------------
 OpenWrt SNAPSHOT, r22124-bfa5e4e4eb
 -----------------------------------------------------
root@OpenWrt:~# iwinfo wl1-ap0 htmodelist
HT20 HT40 VHT20 VHT40 VHT80 HE20 HE40 HE80 HE160

netsh wlan show all
There is 1 interface on the system:

    Name                   : Wi-Fi
    Description            : Intel(R) Wi-Fi 6E AX211 160MHz
    GUID                   : f2066b41-7c39-4064-80b7-fff1bxxxxxxx
    Physical address       : f4:ce:23:xx:xx:xx
    State                  : connected
    SSID                   : OpenWrt_5G
    BSSID                  : 5c:02:14:xx:xx:xx
    Network type           : Infrastructure
    Radio type             : 802.11n
    Authentication         : WPA3-Personal  (H2E)
    Cipher                 : CCMP
    Connection mode        : Profile
    Channel                : 100
    Receive rate (Mbps)    : 300
    Transmit rate (Mbps)   : 300
    Signal                 : 93%
    Profile                : OpenWrt_5G

[ThiloteE]

Radio type : 802.11n

@CornelCotoara I assume this is your client. For some reason, it is configured to run with 802.11n mode, instead of 802.11ax...

[CornelCotoara]

It's definitely configured as AX client. I will try and rebuild from scratch, maybe I broke something trying to patch with quilt.

[CornelCotoara]

No luck, rebuilding de image from scratch did not help, the wifi is still capped at 300Mbps both ways like it shows in the above post. PS. Just flashed SNAPSHOT (r22124-bfa5e4e4eb) downloaded from openwrt and HE80 works as expected the client caps at 1201/1201 (Mbps), no HE160 obviously.

 OpenWrt SNAPSHOT, r22124-bfa5e4e4eb
 -----------------------------------------------------
root@OpenWrt:~# iwinfo wl1-ap0 htmodelist
HT20 HT40 VHT20 VHT40 VHT80 HE20 HE40 HE80

[nbd168]

Please try this patch: https://nbd.name/p/d9c34f29

[CornelCotoara]

The patch it's working!

OpenWrt SNAPSHOT, r22171-fdd1af9a44
 -----------------------------------------------------
root@OpenWrt:~# iwinfo wl1-ap0 htmodelist
HT20 HT40 VHT20 VHT40 VHT80 VHT160 HE20 HE40 HE80 HE160

 Name                   : Wi-Fi
    Description            : Intel(R) Wi-Fi 6E AX211 160MHz
    GUID                   : f2066b41-7c39-4064-80b7-ffffffffffff
    Physical address       : f4:ce:23:xx:xx:xx
    State                  : connected
    SSID                   : OpenWrt_5G
    BSSID                  : 5c:02:14:xx:xx:xx
    Network type           : Infrastructure
    Radio type             : 802.11ax
    Authentication         : WPA3-Personal  (H2E)
    Cipher                 : CCMP
    Connection mode        : Profile
    Channel                : 100
    Receive rate (Mbps)    : 2162
    Transmit rate (Mbps)   : 1729
    Signal                 : 92%
    Profile                : OpenWrt_5G

Iperf summary:

 .\iperf3.exe -c docker -u -b 2G -P 4 -V
iperf 3.1.3
CYGWIN_NT-10.0 ccornel-2ZB81T3 2.5.1(0.297/5/3) 2016-04-21 22:14 x86_64
Time: Wed, 01 Mar 2023 15:17:22 GMT
Connecting to host docker, port 5201
      Cookie: ccornel-2ZB81T3.1677683842.711089.0a
[truncated blah blah content]
[SUM]   0.00-10.00  sec  1.24 GBytes  1.06 Gbits/sec  0.000 ms  0/0 (0%)
CPU Utilization: local/sender 79.3% (14.7%u/64.6%s), remote/receiver 0.0% (0.0%u/0.0%s)

PS. I only need to add bridger, but i cannot find the package in make menuconfig, there's no "services" under network. Any clues on this? I can share the build if someone wanna test but i warn you that's a very basic build, no GUI included.

[nbd168]

To enable bridger, you need to enable LLVM toolchain support. You can download the precompiled llvm toolchain if you want to and extract it to your openwrt source directory.

[wired-dev]

ndb168; ty for fixing this with https://github.com/openwrt/mt76/commit/f9ca70d6367a16fecd80b93fec60fb0d4e3b9fd0

[nbd168]

Added in f9ca70d6367a

[Anteus]

This specific commit broke something on the rt3200 80MHz with some obstacles. Iperf went from around 400Mbit/s to 40Mbit/s. Edit: It was just L419 https://github.com/openwrt/mt76/commit/f9ca70d6367a16fecd80b93fec60fb0d4e3b9fd0#commitcomment-104764055

[csharper2005]

HE160 has serious compatibility issues with Apple devices. Unless someone fix it, the 160 MHz won't come back.

AX 160MHz works perfectly with kmod-7915e on Mediatek filogic devices (mt7986).

Can't apple users disable 160 MHz on their APs?

[darmach]

Wow, compatibility issues with Apple devices, that explains !all (mind the hairpin). Btw, who's forcing you to even use Aple or HE160 if you have major issues with? I have done tests on my side, HE160 has a huge gain against HE80 on my particular setup, so yes, for me it's a must have. The hardware supports HE160 so every owner should be able to use that if needed. Here's a small quote from OpenWrt repo itself,: `` For developers, OpenWrt is the framework to build an application without having to build a complete firmware around it; for users this means the ability for full customization, to use the device in ways never envisioned.

Sunshine! ``

Right? If some specific group has problems, it's not like they can't disable it. Unlike enabling it when it's no longer supported - and for VR Meta Airlink 160MHz is the prerequisite, tried at 80MHz and it was unusable.