Adding OpenWrt support for Xiaomi AX3600 (Part 1)

castiel652 · 2021-07-26T14:15:44.997Z

first time building this repo?

Try delete the tmp folder and do ./script/feeds clean

perceival · 2021-07-26T16:31:49.515Z

That's not a problem with building system but with config on the router. Apparently I have left something that sysupgrade recognizes as config 1.0 while migrating from previous router. Wiping out AX3600 config solved the issue.

hgblob · 2021-07-26T16:49:23.374Z

Ok, I think I'm going crazy. There seems to be some packet corruption when only using ethernet with big(?) file downloads. I tried a few versions starting from robimarko's branch and all seem to have exactly the same problem:

$ wget https://cdimage.ubuntu.com/ubuntu-server/daily-live/20210628/impish-live-server-amd64.iso
--2021-07-26 19:32:15--  https://cdimage.ubuntu.com/ubuntu-server/daily-live/20210628/impish-live-server-amd64.iso
Resolving cdimage.ubuntu.com (cdimage.ubuntu.com)... 91.189.91.123, 91.189.91.124, 2001:67c:1562::25, ...
Connecting to cdimage.ubuntu.com (cdimage.ubuntu.com)|91.189.91.123|:443... connected.
HTTP request sent, awaiting response... 200 OK
Length: 1245161472 (1,2G) [application/x-iso9660-image]
Saving to: ‘impish-live-server-amd64.iso’

impish-live-server-amd64.iso             15%[===========>                                                                    ] 188,38M  23,3MB/s    in 9,2s    

2021-07-26 19:32:25 (20,4 MB/s) - Read error at byte 197525504/1245161472 (error:1408F119:SSL routines:ssl3_get_record:decryption failed or bad record mac). Retrying.

When trying to download some large file I get SSL errors. If I use my other router everything is OK, so it's not my ISP or something upstream. I can repro this 100% when using ethernet, Wifi is not affected.

I suppose the SSL error means there is some packet corruption when going WAN -> LAN. Sounds like a NSS error to me but I don't know enough about it.

emptyteeth · 2021-07-26T17:27:38.794Z

xabolcs:

I don't have any ath11k in the dmesg

Do you have these two packages in your device?

# opkg list-installed |grep kmod-ath11k
kmod-ath11k - 5.10.46+5.10.42-1-1
kmod-ath11k-ahb - 5.10.46+5.10.42-1-1

If yes, run this to see what happen

insmod ath11k
insmod ath11k_ahb

If not, make sure they are selected in menuconfig

xabolcs · 2021-07-26T19:05:26.045Z

Thanks for the hint!
I don't have those:

root@OpenWrt:~# opkg list-installed|grep ath11
ath11k-firmware-ipq8074 - 2021-07-20-d4003c19-2

But I can't find them in menuconfig also. (I searched in "Kernel modules / Wireless Drivers, where the kmod-ath10k & co are.)

Are they hidden behind some other switches?

emptyteeth · 2021-07-26T19:29:31.444Z

I think they should be there by default

make sure you're at the AX3600-5.10-restart branch

then try this to see if they show up

make clean
rm -rf tmp
./scripts/feeds update -a && ./scripts/feeds install -a
make menuconfig

xabolcs · 2021-07-26T19:40:58.936Z

emptyteeth:

rm -rf tmp

That helped! Thanks!
Now they showed up (and got selected automatically)!

  │ ┌──────────────────^(-)────────────────────────────────────────────────────────────────────────────────────┐ │  
  │ │                  -*- kmod-ath...................................... Atheros common driver part  --->     │ │  
  │ │                  < > kmod-ath10k...................... Atheros 802.11ac wireless cards support           │ │  
  │ │                  <*> kmod-ath10k-ct......... ath10k-ct driver optimized for CT ath10k firmware           │ │  
  │ │                  [*] Enable LED support                                                                  │ │  
  │ │                  < > kmod-ath10k-ct-smallbuffers                                                         │ │  
  │ │                  -*- kmod-ath11k..... Qualcomm 802.11ax wireless chipset support (common code)           │ │  
  │ │                  [*]   Enable thermal sensors and throttling support                                     │ │  
  │ │                  <*> kmod-ath11k-ahb........... Qualcomm 802.11ax AHB wireless chipset support           │ │  
  │ │                  < > kmod-ath11k-pci........... Qualcomm 802.11ax PCI wireless chipset support           │ │  
  │ │                  < > kmod-ath5k........................... Atheros 5xxx wireless cards support           │ │  
  │ └──────────────────v(+)────────────────────────────────────────────────────────────────────────────────────┘ │

Thank you very much for your help.

And sorry for the half-offtopic comments. I'd like to test the AX3600 as a mesh node.

vit0r · 2021-07-26T22:17:55.602Z

I have seen that issue before in my units, but only when running without the NSS kmod and usually only after a fews days of uptime.

hgblob · 2021-07-27T12:56:58.821Z

Looks like it was either connected to NSS or ECM. Either way I swiped the kernel 5.10.51, mac80211(with patches) and nss(they call it nss-64??) from the lede above and everything works nicely now.

I should mentioned that I also ran the lede build by itself and that was also not causing any package corruption.

Two things I did notice:

when it works correctly both the rx and tx lan paths are accelerated, I see the packet count in nss debugfs
for me ecm did always load correctly and as far as I understand ecm has something to do with the bridge accel(?)

Also for what it's worth, I got ath11k to create a 160Mhz network, but this seems kinda useless cause it doesn't go above a 866Mbit MCS rate.

Edit: This also fixed my Wireguard problem, now from wired LAN I also get line speed.

castiel652 · 2021-07-27T14:10:27.611Z

hgblob:

160Mhz

they just sent the 160MHz patch upstream so you are gonna need to wait
ath11k: add support for 80P80 and 160 MHz bandwidth

hgblob:

nss-64

They call it NSS-64 so they can have different packages for ipq806x and ipq807x

hgblob · 2021-07-27T18:50:37.121Z

castiel652:

they just sent the 160MHz patch upstream so you are gonna need to wait
ath11k: add support for 80P80 and 160 MHz bandwidth

This patch is already applied so I think more is needed to actually get 160Mhz rates.

joba-1 · 2021-07-29T22:30:05.150Z

About to reboot for a new firmware, soo

ax1

image734×325 33.4 KB

ax2

image734×325 36.3 KB

6 days uptime, "normal" usage as dumb APs. Doesn't look like there is a memory leak in this scenario.

perceival · 2021-07-30T04:48:47.010Z

Same here, even earlier - after just a few MB. Also observed freshly after reboot, with and without packet steering and regardless of software offloading setting:

Log

:: ~ » wget https://ftp.icm.edu.pl/pub/Linux/fedora/linux/releases/34/Workstation/x86_64/iso/Fedora-Workstation-Live-x86_64-34-1.2.iso                                                                                           130 ↵
--2021-07-30 06:42:15--  https://ftp.icm.edu.pl/pub/Linux/fedora/linux/releases/34/Workstation/x86_64/iso/Fedora-Workstation-Live-x86_64-34-1.2.iso
Resolving ftp.icm.edu.pl (ftp.icm.edu.pl)... 193.219.28.2
Connecting to ftp.icm.edu.pl (ftp.icm.edu.pl)|193.219.28.2|:443... connected.
HTTP request sent, awaiting response... 200 OK
Length: 2007367680 (1,9G) [application/octet-stream]
Saving to: ‘Fedora-Workstation-Live-x86_64-34-1.2.iso.3’

Fedora-Workstation-Live-x86_64-34-1.2.iso.3                   0%[                                                                                                                                          ] 160,00K   603KB/s    in 0,3s    

2021-07-30 06:42:16 (603 KB/s) - Read error at byte 163840/2007367680 (Decryption has failed.). Retrying.

--2021-07-30 06:42:17--  (try: 2)  https://ftp.icm.edu.pl/pub/Linux/fedora/linux/releases/34/Workstation/x86_64/iso/Fedora-Workstation-Live-x86_64-34-1.2.iso
Connecting to ftp.icm.edu.pl (ftp.icm.edu.pl)|193.219.28.2|:443... connected.
HTTP request sent, awaiting response... 206 Partial Content
Length: 2007367680 (1,9G), 2007203840 (1,9G) remaining [application/octet-stream]
Saving to: ‘Fedora-Workstation-Live-x86_64-34-1.2.iso.3’

Fedora-Workstation-Live-x86_64-34-1.2.iso.3                   0%[                                                                                                                                          ] 416,00K   892KB/s    in 0,3s    

2021-07-30 06:42:17 (892 KB/s) - Read error at byte 425984/2007367680 (Decryption has failed.). Retrying.

--2021-07-30 06:42:19--  (try: 3)  https://ftp.icm.edu.pl/pub/Linux/fedora/linux/releases/34/Workstation/x86_64/iso/Fedora-Workstation-Live-x86_64-34-1.2.iso
Connecting to ftp.icm.edu.pl (ftp.icm.edu.pl)|193.219.28.2|:443... connected.
HTTP request sent, awaiting response... 206 Partial Content
Length: 2007367680 (1,9G), 2006941696 (1,9G) remaining [application/octet-stream]
Saving to: ‘Fedora-Workstation-Live-x86_64-34-1.2.iso.3’

Fedora-Workstation-Live-x86_64-34-1.2.iso.3                   0%[                                                                                                                                          ] 528,00K   584KB/s    in 0,2s    

2021-07-30 06:42:21 (584 KB/s) - Read error at byte 540672/2007367680 (Decryption has failed.). Retrying.

--2021-07-30 06:42:24--  (try: 4)  https://ftp.icm.edu.pl/pub/Linux/fedora/linux/releases/34/Workstation/x86_64/iso/Fedora-Workstation-Live-x86_64-34-1.2.iso
Connecting to ftp.icm.edu.pl (ftp.icm.edu.pl)|193.219.28.2|:443... connected.
HTTP request sent, awaiting response... 206 Partial Content
Length: 2007367680 (1,9G), 2006827008 (1,9G) remaining [application/octet-stream]
Saving to: ‘Fedora-Workstation-Live-x86_64-34-1.2.iso.3’

Fedora-Workstation-Live-x86_64-34-1.2.iso.3                   0%[                                                                                                                                          ]   2,97M  1,72MB/s    in 1,4s    

2021-07-30 06:42:25 (1,72 MB/s) - Read error at byte 3112960/2007367680 (Decryption has failed.). Retrying.

--2021-07-30 06:42:29--  (try: 5)  https://ftp.icm.edu.pl/pub/Linux/fedora/linux/releases/34/Workstation/x86_64/iso/Fedora-Workstation-Live-x86_64-34-1.2.iso
Connecting to ftp.icm.edu.pl (ftp.icm.edu.pl)|193.219.28.2|:443... connected.
HTTP request sent, awaiting response... 206 Partial Content
Length: 2007367680 (1,9G), 2004254720 (1,9G) remaining [application/octet-stream]
Saving to: ‘Fedora-Workstation-Live-x86_64-34-1.2.iso.3’

Fedora-Workstation-Live-x86_64-34-1.2.iso.3                   0%[                                                                                                                                          ]   3,19M   977KB/s    in 0,2s    

2021-07-30 06:42:30 (977 KB/s) - Read error at byte 3342336/2007367680 (Decryption has failed.). Retrying.

--2021-07-30 06:42:35--  (try: 6)  https://ftp.icm.edu.pl/pub/Linux/fedora/linux/releases/34/Workstation/x86_64/iso/Fedora-Workstation-Live-x86_64-34-1.2.iso
Connecting to ftp.icm.edu.pl (ftp.icm.edu.pl)|193.219.28.2|:443... connected.
HTTP request sent, awaiting response... 206 Partial Content
Length: 2007367680 (1,9G), 2004025344 (1,9G) remaining [application/octet-stream]
Saving to: ‘Fedora-Workstation-Live-x86_64-34-1.2.iso.3’

Fedora-Workstation-Live-x86_64-34-1.2.iso.3                   0%[                                                                                                                                          ]   5,19M  2,97MB/s    in 0,7s    

2021-07-30 06:42:36 (2,97 MB/s) - Read error at byte 5439488/2007367680 (Decryption has failed.). Retrying.

--2021-07-30 06:42:42--  (try: 7)  https://ftp.icm.edu.pl/pub/Linux/fedora/linux/releases/34/Workstation/x86_64/iso/Fedora-Workstation-Live-x86_64-34-1.2.iso
Connecting to ftp.icm.edu.pl (ftp.icm.edu.pl)|193.219.28.2|:443... connected.
HTTP request sent, awaiting response... 206 Partial Content
Length: 2007367680 (1,9G), 2001928192 (1,9G) remaining [application/octet-stream]
Saving to: ‘Fedora-Workstation-Live-x86_64-34-1.2.iso.3’

Fedora-Workstation-Live-x86_64-34-1.2.iso.3                   0%[                                                                                                                                          ]   5,28M   433KB/s    in 0,2s    

2021-07-30 06:42:43 (433 KB/s) - Read error at byte 5537792/2007367680 (Decryption has failed.). Retrying.

If I understand correctly NSS is on:

 1070 root         0 IW<  [nss_data_plane_]
 1073 root         0 IW<  [nss_freq_queue]

Directly from AX3600 works fine:

Log

root@router:~# wget https://ftp.icm.edu.pl/pub/Linux/fedora/linux/releases/34/Wo
rkstation/x86_64/iso/Fedora-Workstation-Live-x86_64-34-1.2.iso -O /dev/null 
--2021-07-30 06:59:16--  https://ftp.icm.edu.pl/pub/Linux/fedora/linux/releases/34/Workstation/x86_64/iso/Fedora-Workstation-Live-x86_64-34-1.2.iso
Resolving ftp.icm.edu.pl... 193.219.28.2
Connecting to ftp.icm.edu.pl|193.219.28.2|:443... connected.
HTTP request sent, awaiting response... 200 OK
Length: 2007367680 (1.9G) [application/octet-stream]
Saving to: '/dev/null'

/dev/null           100%[===================>]   1.87G  34.5MB/s    in 55s     

2021-07-30 07:00:12 (34.6 MB/s) - '/dev/null' saved [2007367680/2007367680]

Fine also while connected over WiFi. MTU mismatch?

hgblob · 2021-07-30T07:36:13.916Z

My system has been rock solid for some days now so I'm trying to back-track.

I slowly began reverting the changes I swiped from the lede project above. First on the list was nss(&friends) which seemed identical to robi's nss feeds. The system still works so I guess the problem was somewhere else.

To be honest the patch list is pretty small and nothing really stands out(besides the switch to kernel 5.10.51) but I'll keep going.

dspalu32 · 2021-07-31T10:04:13.957Z

@robimarko ath11k WDS PR submitted, using the recent upstream Kernel versions rather than the older ones from CodeAurora.

Compiled from scratch and running OK for me here. Build log shows all three patches being applied cleanly despite line offsets.

The OpenWRT mac80211 HE80 WDS patch was already in 5.10-restart.

plevine · 2021-07-31T17:46:06.345Z

I've been running LEDE for 24 hours It's the first time a self-compiled build didn't crash overnight. I noticed LEDE committed a number of ath11k: memory leak patches 4 days ago.

Ansuel · 2021-07-31T17:47:42.409Z

CAN WE PLS STOP CALLING THAT LEDE ?

It's a custom repo based on the old openwrt project name...
Someone coming here for the first time would think that the old lede is back again producing official firmware...

kostadinov · 2021-07-31T18:03:24.085Z

github.com/coolsnowwolf/lede

AX3600仍然内存泄漏

opened 12:34AM - 25 Jul 21 UTC

cath-palug

反馈bug/问题模板，提建议请删除 ## 1.关于你要提交的问题 Q：是否搜索了issue (使用 "x" 选择) * [x] 没有类似的issu…e ## 2. 详细叙述在长时间运行后，仍然会内存泄漏，内存可用非常低，打游戏时最为明显内存一旦低下就会把Clash进程杀掉，Clash守护进程会把Clash再次复活，内存越来越少现在已经接入N1为旁路由维持科学上网还是漏，俩小时内存吃完现在换成QSDK了，不漏了，内存稳定剩余70M ![QQ截图20210725083809](https://user-images.githubusercontent.com/44462661/126884208-421ea778-4665-4033-a101-dba448d2a1a1.png) ![P`0DAVT)93WZL3(E @C2{(8](https://user-images.githubusercontent.com/44462661/126884210-18c89a20-1331-456e-94aa-4e71d2f4372b.png) ![QQ截图20210727004652](https://user-images.githubusercontent.com/44462661/127027314-46bebaa9-3709-446b-83c8-80e18df3b0ff.png)

that fork have the same problem

hgblob · 2021-07-31T18:12:54.904Z

Well, the chinese fork then has for sure some patch that solves the leak, I've been running it as my main router for a while now and no OOM.

My suspicion is that it's this patch, because I only swiped their mac80211 patches, which aren't many.

Bisection is kinda difficult because of the other packet corruption bug but I have a rather limited number of patches applied to robi's tree and the system is still stable.

Stat_headcrabed · 2021-07-31T20:32:00.402Z

Thx guys. I think it's time to have a try for this. Btw which branch should I use? And how to switch between two firmwares? I remember having seen how to switch but I couldn't find that.