Are you in good testing condition?
You are in a fully functional version and updating your mwlwifi package.
Whatever your openwrt version, I can compile one for you at the appropriate version ;).
@jbsky I tested the mwlwifi packages I built for 23.05.0-rc1 and couldn't connect to wifi. I replaced the both the kmod-mwlwifi and mwlwifi-firmware files. I'm not sure if I did something wrong with compiling it because I had to update 005-mac80211_update.patch to get it to compile. I had to change hif/pcie/tx.c:
- if (sta && (sta->ht_cap.ht_supported || sta->vht_cap.vht_supported)
+ if (sta && (sta->deflink.ht_cap.ht_supported || sta->deflink.vht_cap.vht_supported)
I was trying to follow what the original patch did, but I don't know what any of that means.
My router was functioning before replacing the files and the wifi went back to functioning after reinstalling the original files.
1 Like
@trinidude4 , thank you for your test, can you please test this sha1? 5404712
When I reread the code, I think the 1st version is really the right one.
Are you in CCMP only force CCMP(AES), for TKIP, we'll see later.
Best regards,
jbs
1 Like
I ended up creating a new wireless config and was still not able to connect to wifi.
If you could possibly create packages for 23.05.0-rc1, that would be best for testing purposes at the moment. Thank you.
1 Like
From what I can see, openwrt 23.05 is a 6.1 linux kernel.
It's compiled and available here:
I get the following compile error now:
2023-06-17T11:36:53.4813028Z /home/runner/work/openwrt/openwrt/build_dir/target-arm_cortex-a9+vfpv3-d16_musl_eabi/linux-mvebu_cortexa9/mwlwifi-2023-06-17-54047122/hif/pcie/tx_ndp.c: In function 'pcie_tx_xmit_ndp':
2023-06-17T11:36:53.4815030Z /home/runner/work/openwrt/openwrt/build_dir/target-arm_cortex-a9+vfpv3-d16_musl_eabi/linux-mvebu_cortexa9/mwlwifi-2023-06-17-54047122/hif/pcie/tx_ndp.c:508:13: error: unused variable 'tailpad' [-Werror=unused-variable]
2023-06-17T11:36:53.4815793Z 508 | int tailpad = 0;
2023-06-17T11:36:53.4816875Z | ^~~~~~~
2023-06-17T11:36:53.4820880Z /home/runner/work/openwrt/openwrt/build_dir/target-arm_cortex-a9+vfpv3-d16_musl_eabi/linux-mvebu_cortexa9/mwlwifi-2023-06-17-54047122/hif/pcie/tx_ndp.c:499:36: error: unused variable 'k_conf' [-Werror=unused-variable]
2023-06-17T11:36:53.4821837Z 499 | struct ieee80211_key_conf *k_conf;
2023-06-17T11:36:53.4822375Z | ^~~~~~
2023-06-17T11:36:53.8375160Z cc1: all warnings being treated as errors
2023-06-17T11:36:53.8459796Z make[4]: *** [scripts/Makefile.build:289: /home/runner/work/openwrt/openwrt/build_dir/target-arm_cortex-a9+vfpv3-d16_musl_eabi/linux-mvebu_cortexa9/mwlwifi-2023-06-17-54047122/hif/pcie/tx_ndp.o] Error 1
Also, I'm using WPA2 so no TKIP.
1 Like
openwrt 23.0.5-0.rc1 is based on linux 6.1.
This commit compiles well https://github.com/jbsky/mwlwifi/commit/16d9f7c
23.05 is based on Linux kernel 5.15. 23.05.0-rc1 is running 5.15.114.
Master snapshot is using 6.1 afaik.
2 Likes
You're probably right, the starting point is a 5.15, but today it looks more like a 6.1 than a 5.15 in terms of compilation.
In any case, the kmod-mwlwifi packages I'm compiling are KO. @trinidude4 , I'm counting on you 
I'll find out what's wrong with these 5.15 modules 
23.05 should stay on kernel 5.15 for it’s lifetime from what I read. 23.05 can be expected to get 5.15 kernel updates. 5.15 has a projected upstream EOL in October 2026.
The bump to kernel 6.1 was made on master recently. The next release 24.xx or 25.xx should then get Linux kernel 6.1.
2 Likes
Remember, mac80211 is backported in openwrt from more recent kernels.
jbsky is correct, as far as mac80211 is concerned it is 6.1. The underlying kernel for the OS is 5.15.
3 Likes
Indeed, we are not in a full 6.1, but a 5.15 with mac80211 backported, thank you @lantis1008 for this clarification.
Thank you @trinidude4 for your pipeline, very useful to understand how to compile the mwlwifi module, I managed to compile working modules (without crashing )
@WildByDesign, can you test these compiled kmods and tell me if there's at least one that works please.
https://github.com/jbsky/mwlwifi/raw/testing/23.05.rc1/kmod/kmod-mwlwifi_5.15.114%2B2023-04-29-6682ce5-1_arm_cortex-a9_vfpv3-d16.ipk
https://github.com/jbsky/mwlwifi/raw/testing/23.05.rc1/kmod/kmod-mwlwifi_5.15.114%2B2023-04-29-955fc21-1_arm_cortex-a9_vfpv3-d16.ipk
https://github.com/jbsky/mwlwifi/raw/testing/23.05.rc1/kmod/kmod-mwlwifi_5.15.114%2B2023-04-29-a59896d-1_arm_cortex-a9_vfpv3-d16.ipk
Thank you in advance.
jbs
1 Like
Yes, for sure. I am ready to do testing in a few hours. As soon as the family is off the network for the night.
I am more prepared now to recover from possible failures compared to last night.
I am very thankful for your work and time on this and I am happy to help whenever possible.
1 Like
Here is the results from my testing.
955fc21 - fail - can see both 2g/5g but can’t connect
6682ce5 - success- everything good and slightly faster wifi than usual
a59896d - fail - can see both 2g/5g but can’t connect
So the two commits were unsuccessful, but 6682ce5 was successful and my wireless speed tests were actually faster than normal. EDIT: Although that speed difference is likely coincidence because that commit is only for 88W8864 firmware which does not affect WRT3200/32X devices.
Please let me know if/when you have more to test. I have a good testing environment setup now.
1 Like
Have you tried creating a new guest wifi on 2ghz on WRT3200ACM once I add a new WiFi and create a new interface for it and save it, radio 1 immediately becomes permanently disabled. Let me know if it not just me.
1 Like
Hello,
@WildByDesign, Thanks for your tests, it's important to have a good test environment.
I need you to test these next 3 kmod.
https://github.com/jbsky/mwlwifi/raw/testing/mwl8964/20230618/kmod-mwlwifi_5.15.114%2B2023-04-29-1edfacb-1_arm_cortex-a9_vfpv3-d16.ipk
https://github.com/jbsky/mwlwifi/raw/testing/mwl8964/20230618/kmod-mwlwifi_5.15.114%2B2023-04-29-bdf2704-1_arm_cortex-a9_vfpv3-d16.ipk
https://github.com/jbsky/mwlwifi/raw/testing/mwl8964/20230618/kmod-mwlwifi_5.15.114%2B2023-04-29-bf9e847-1_arm_cortex-a9_vfpv3-d16.ipk
@openwrtforever, your problem is a regression for version 23.05.0-rc1.
It's great that you've noticed, can you send a message here if you haven't already?
Thank you in advance,
Best regards,
Jbs
3 Likes
Hello @WildByDesign ,
No worries, everyone's time is precious.
Debugging this driver is a good challenge. It's a good entry point for getting to know the Linux kernel.
Thank you for your time.
I'd especially like to thank Marvel for developing a driver of dubious quality.
Best regards,
Jbs
3 Likes
You're welcome. Absolutely. I'm excited to test more. I wont be able to test them until about the same time as last night, so about 12 hours from now. I will have testing results for you by morning again.
1 Like
Same issue. Hopefully that guest wifi issue gets resolved.
If the issue resides in LuCI, I wonder if it would be possible to create the new interface manually by editing the config files. I haven't tried that though.
Actually i believe the issue is outside of Luci. My initial attempt (which failed) was loading back in the settings via the config files. Committing the settings caused the radio to drop. My next plan was to do it purely via the GUI but i suspect between us, we've covered both bases.
Interestingly, all the other settings for the Guest network (DHCP, firewall...) were accepted fine. Enabling the radio once the ssid was configured caused the drop.
1 Like