WCN7850 disconnects and reconnects in a loop

WCN7850 disconnects and reconnects in a loop

[Gregoire]

Hello,

I have a QCNCM865 card with a WCN7850 chipset. I'm running Arch Linux 
version 6.9.0-rc2-1-mainline.

I'm experiencing difficulty connecting to any Wi-Fi network; the device 
connects and reconnects almost immediately.

Here's a gist containing some logs from dmesg and journalctl related to iwd.

https://gist.github.com/GreyXor/5e6dc8e610a00c8aab37a1251d36dbc1

Let me know if you need anything else for debugging

I also tried 
https://git.kernel.org/pub/scm/linux/kernel/git/kvalo/ath.git last week 
on master, and same problem

Thanks

Re: WCN7850 disconnects and reconnects in a loop

[Baochen Qiang]

On 4/1/2024 5:56 PM, Gregoire wrote:
> Hello,
> 
> I have a QCNCM865 card with a WCN7850 chipset. I'm running Arch Linux 
> version 6.9.0-rc2-1-mainline.
> 
> I'm experiencing difficulty connecting to any Wi-Fi network; the device 
> connects and reconnects almost immediately.
> 
> Here's a gist containing some logs from dmesg and journalctl related to 
> iwd.
> 
> https://gist.github.com/GreyXor/5e6dc8e610a00c8aab37a1251d36dbc1

The log below seems strange. Actually we have a patch to fix this issue
https://git.kernel.org/pub/scm/linux/kernel/git/kvalo/ath.git/commit/drivers/net/wireless/ath/ath12k?id=a49732b1f1f8128608b1192c562880dd29037c06. 
So the question is are you still seeing this log with latest Kalle's 
ath.git?

[ 1132.255250] ath12k_pci 0000:01:00.0: Unknown htt tx status 8

Current logs don't provide enough info.  Could you reproduce it with 
full ath12k log and IWD log enabled and share them to us?
For ath12k, first unload module and then reload.
	modprobe -r ath12k
	modprobe ath12k dyndbg=+pflmt debug_mask=0xffffffff
For IWD, simply add '-d' option when starting it. You may directly add 
it to the systemd service entry if you like. See 
https://iwd.wiki.kernel.org/debugging

Better if you could also capture air sniffer as well.

> 
> Let me know if you need anything else for debugging
> 
> I also tried 
> https://git.kernel.org/pub/scm/linux/kernel/git/kvalo/ath.git last week 
> on master, and same problem
> 
> Thanks
> 
> 

Re: WCN7850 disconnects and reconnects in a loop

[Baochen Qiang]

On 4/3/2024 2:36 PM, Baochen Qiang wrote:
> 
> 
> On 4/1/2024 5:56 PM, Gregoire wrote:
>> Hello,
>>
>> I have a QCNCM865 card with a WCN7850 chipset. I'm running Arch Linux 
>> version 6.9.0-rc2-1-mainline.
>>
>> I'm experiencing difficulty connecting to any Wi-Fi network; the 
>> device connects and reconnects almost immediately.
>>
>> Here's a gist containing some logs from dmesg and journalctl related 
>> to iwd.
>>
>> https://gist.github.com/GreyXor/5e6dc8e610a00c8aab37a1251d36dbc1
> 
> The log below seems strange. Actually we have a patch to fix this issue
> https://git.kernel.org/pub/scm/linux/kernel/git/kvalo/ath.git/commit/drivers/net/wireless/ath/ath12k?id=a49732b1f1f8128608b1192c562880dd29037c06. So the question is are you still seeing this log with latest Kalle's ath.git?
> 
> [ 1132.255250] ath12k_pci 0000:01:00.0: Unknown htt tx status 8
> 
> Current logs don't provide enough info.  Could you reproduce it with 
> full ath12k log and IWD log enabled and share them to us?
> For ath12k, first unload module and then reload.
>      modprobe -r ath12k
>      modprobe ath12k dyndbg=+pflmt debug_mask=0xffffffff
> For IWD, simply add '-d' option when starting it. You may directly add 
> it to the systemd service entry if you like. See 
> https://iwd.wiki.kernel.org/debugging
> 
> Better if you could also capture air sniffer as well.
> 
And please share complete kernel/dmesg and IWD log, not need to do grep.

>>
>> Let me know if you need anything else for debugging
>>
>> I also tried 
>> https://git.kernel.org/pub/scm/linux/kernel/git/kvalo/ath.git last 
>> week on master, and same problem
>>
>> Thanks
>>
>>
> 

Re: WCN7850 disconnects and reconnects in a loop

[Gregoire Stein]

Hello,
now i'm on the last commit of master-pending 
(6.9.0-rc2-wt-ath-1-mainline-g853450ce5ca0)
here are the various logs you asked for: 
https://gist.github.com/GreyXor/927e7871e70630beabb522aeddc5891e
There are iwd, networkmanager and dmesg
iwd is launched with -d and the ath12k module with the debug flag
thanks

Re: WCN7850 disconnects and reconnects in a loop

[Baochen Qiang]

On 4/4/2024 11:41 PM, Gregoire Stein wrote:
> Hello,
> now i'm on the last commit of master-pending 
> (6.9.0-rc2-wt-ath-1-mainline-g853450ce5ca0)
> here are the various logs you asked for: 
> https://gist.github.com/GreyXor/927e7871e70630beabb522aeddc5891e

Thanks for the logs.

 From dmesg and NetworkManager logs, connection and EAPOL frames are 
good, but DHCP transaction fails. IWD logs show 'Del Station' event 
received after EAPOL is done. So guessing 3rd party triggers 'Del 
Station' due to some unkown reasons (DHCP session failed?).

In order to know who triggers 'Del Station', could you help reproduce 
this issue and collect mac80211/cfg80211 trace?

Below steps for your consideration:

1. enable tracing on mac80211/cfg80211
	# echo 1 > /sys/kernel/debug/tracing/events/cfg80211/enable
	# echo 1 > /sys/kernel/debug/tracing/events/mac80211/enable

2. recording trace to local file
	# cat /sys/kernel/debug/tracing/trace_pipe > trace.log

3. reproduce this issue

4. share full/verbose kernel/ath12k/IWD/NetworkManager logs, just like 
before, and trace log to us. Air sniffer logs are useful as well if 
possible.

> There are iwd, networkmanager and dmesg
> iwd is launched with -d and the ath12k module with the debug flag
> thanks
> 
And one more question, is this issue specific to AP? could you try other 
APs or could you try other client with this same AP? We can do 
comparison to narrow down.

Re: WCN7850 disconnects and reconnects in a loop

[Kalle Valo]

Baochen Qiang <quic_bqiang@quicinc.com> writes:

> On 4/4/2024 11:41 PM, Gregoire Stein wrote:
>> Hello,
>> now i'm on the last commit of master-pending
>> (6.9.0-rc2-wt-ath-1-mainline-g853450ce5ca0)
>> here are the various logs you asked for:
>> https://gist.github.com/GreyXor/927e7871e70630beabb522aeddc5891e
>
> Thanks for the logs.
>
> From dmesg and NetworkManager logs, connection and EAPOL frames are
> good, but DHCP transaction fails. IWD logs show 'Del Station' event
> received after EAPOL is done. So guessing 3rd party triggers 'Del
> Station' due to some unkown reasons (DHCP session failed?).
>
> In order to know who triggers 'Del Station', could you help reproduce
> this issue and collect mac80211/cfg80211 trace?
>
> Below steps for your consideration:
>
> 1. enable tracing on mac80211/cfg80211
> 	# echo 1 > /sys/kernel/debug/tracing/events/cfg80211/enable
> 	# echo 1 > /sys/kernel/debug/tracing/events/mac80211/enable

This is just a suggestion but it might be helpful to include also driver
and wpasupplicant traces to the same file. For example, here's what we
documented for ath10k:

https://wireless.wiki.kernel.org/en/users/drivers/ath10k/debug#tracing

>> There are iwd, networkmanager and dmesg
>> iwd is launched with -d and the ath12k module with the debug flag
>> thanks
>> 
> And one more question, is this issue specific to AP? could you try
> other APs or could you try other client with this same AP? We can do
> comparison to narrow down.

Another things to try narrow down the problem is manually running
wpasupplicant. And also using manual IPs (no DHCP at all) or just use
'iw connect' without encryption.

-- 
https://patchwork.kernel.org/project/linux-wireless/list/

https://wireless.wiki.kernel.org/en/developers/documentation/submittingpatches

Re: WCN7850 disconnects and reconnects in a loop

[Gregoire]

I've found a workaround:
I had to add
[General]
  ControlPortOverNL80211=false

in my /etc/iwd/main.conf

And now I can connect and I don't have this issue anymore.

are there any plans to avoid having to add this option in iwd?

Re: WCN7850 disconnects and reconnects in a loop

[Kalle Valo]

Gregoire <gregoire.s93@live.fr> writes:

> I've found a workaround:
> I had to add
> [General]
>  ControlPortOverNL80211=false
>
> in my /etc/iwd/main.conf
>
> And now I can connect and I don't have this issue anymore.
>
> are there any plans to avoid having to add this option in iwd?

To make it easier to track this I recommend filing a bug to bugzilla:

https://wireless.wiki.kernel.org/en/users/drivers/ath12k/bugreport

-- 
https://patchwork.kernel.org/project/linux-wireless/list/

https://wireless.wiki.kernel.org/en/developers/documentation/submittingpatches

Re: WCN7850 disconnects and reconnects in a loop

[Baochen Qiang]

On 4/9/2024 6:49 PM, Gregoire wrote:
> I've found a workaround:
> I had to add
> [General]
>   ControlPortOverNL80211=false
Why are you adding this? Any logs or behavior get your attention to it?

 From naming of this option, seems it affects the delivery of some kinds 
of frames from cfg80211/mac80211 to IWD. What I know now one example of 
them is the EAPOL frames, but from IWD logs EAPOL is good. Any other 
kind of frames?

> 
> in my /etc/iwd/main.conf
> 
> And now I can connect and I don't have this issue anymore.
> 
> are there any plans to avoid having to add this option in iwd?
> 
> 

Re: WCN7850 disconnects and reconnects in a loop

[Gregoire]

I'm adding this because I saw it here: 
https://bbs.archlinux.org/viewtopic.php?pid=2144901#p2144901

I've been lucky, it's the only one I've changed and now it works 
perfectly well. I have another problem but it's not related, I think. 
that when I upload very fast (at 100mb/s, for 30 seconds) then the whole 
ath12k driver crashes and I have to reboot my system.

I will make another thread for this, or is there already an existing thread?

Thanks

Re: WCN7850 disconnects and reconnects in a loop

[Johan Hovold]

On Mon, Apr 01, 2024 at 11:56:39AM +0200, Gregoire wrote:

> I have a QCNCM865 card with a WCN7850 chipset. I'm running Arch Linux 
> version 6.9.0-rc2-1-mainline.
> 
> I'm experiencing difficulty connecting to any Wi-Fi network; the device 
> connects and reconnects almost immediately.
> 
> Here's a gist containing some logs from dmesg and journalctl related to iwd.
> 
> https://gist.github.com/GreyXor/5e6dc8e610a00c8aab37a1251d36dbc1

I'm seeing iwd repeatedly failing to connect with reason
PREV_AUTH_NOT_VALID with WCN7850 on the Lenovo ThinkPad T14s (and the
Snapdragon X1E CRD reference design).

I first noticed this with 6.10 when starting to work on the Qualcomm X
Elite platform, but never got around to investigating it. I've confirmed
that the issue is still there with 6.13-rc1 and that adding
ControlPortOverNL80211=false allows iwd to connect.

Looking at Gregoire's logs now, it looks like the deauthentication
reason is different:

	(Reason: 7=CLASS3_FRAME_FROM_NONASSOC_STA)

instead of

	(Reason: 2=PREV_AUTH_NOT_VALID)

but the same control port workaround allows iwd to connect.

Johan

Re: WCN7850 disconnects and reconnects in a loop

[Kalle Valo]

Johan Hovold <johan@kernel.org> writes:

> On Mon, Apr 01, 2024 at 11:56:39AM +0200, Gregoire wrote:
>
>> I have a QCNCM865 card with a WCN7850 chipset. I'm running Arch Linux 
>> version 6.9.0-rc2-1-mainline.
>> 
>> I'm experiencing difficulty connecting to any Wi-Fi network; the device 
>> connects and reconnects almost immediately.
>> 
>> Here's a gist containing some logs from dmesg and journalctl related to iwd.
>> 
>> https://gist.github.com/GreyXor/5e6dc8e610a00c8aab37a1251d36dbc1
>
> I'm seeing iwd repeatedly failing to connect with reason
> PREV_AUTH_NOT_VALID with WCN7850 on the Lenovo ThinkPad T14s (and the
> Snapdragon X1E CRD reference design).
>
> I first noticed this with 6.10 when starting to work on the Qualcomm X
> Elite platform, but never got around to investigating it. I've confirmed
> that the issue is still there with 6.13-rc1 and that adding
> ControlPortOverNL80211=false allows iwd to connect.
>
> Looking at Gregoire's logs now, it looks like the deauthentication
> reason is different:
>
> 	(Reason: 7=CLASS3_FRAME_FROM_NONASSOC_STA)
>
> instead of
>
> 	(Reason: 2=PREV_AUTH_NOT_VALID)
>
> but the same control port workaround allows iwd to connect.

Is there a bugzilla report about this? Unfortunately I have not been
able to follow bugzilla lately due to MLO work but it would be still
good to file this (if that's not done already).

-- 
https://patchwork.kernel.org/project/linux-wireless/list/

https://wireless.wiki.kernel.org/en/developers/documentation/submittingpatches

Re: WCN7850 disconnects and reconnects in a loop

[Johan Hovold]

On Wed, Dec 11, 2024 at 09:35:37AM +0200, Kalle Valo wrote:
> Johan Hovold <johan@kernel.org> writes:

> > I'm seeing iwd repeatedly failing to connect with reason
> > PREV_AUTH_NOT_VALID with WCN7850 on the Lenovo ThinkPad T14s (and the
> > Snapdragon X1E CRD reference design).
> >
> > I first noticed this with 6.10 when starting to work on the Qualcomm X
> > Elite platform, but never got around to investigating it. I've confirmed
> > that the issue is still there with 6.13-rc1 and that adding
> > ControlPortOverNL80211=false allows iwd to connect.
> >
> > Looking at Gregoire's logs now, it looks like the deauthentication
> > reason is different:
> >
> > 	(Reason: 7=CLASS3_FRAME_FROM_NONASSOC_STA)
> >
> > instead of
> >
> > 	(Reason: 2=PREV_AUTH_NOT_VALID)
> >
> > but the same control port workaround allows iwd to connect.
> 
> Is there a bugzilla report about this? Unfortunately I have not been
> able to follow bugzilla lately due to MLO work but it would be still
> good to file this (if that's not done already).

Yes, sorry, I should have included the link (it was posted later in the
original thread):

	https://bugzilla.kernel.org/show_bug.cgi?id=218733

I only noticed that Gregoire's symptoms were slightly different when
replying to this thread. Let me know if you prefer that I file a
separate bugzilla report for the PREV_AUTH_NOT_VALID case.

The exact same setup (kernel, distro and router) works fine with ath11k
and WCN6855 (e.g. as found in the Lenovo ThinkPad X13s).

Johan

Re: WCN7850 disconnects and reconnects in a loop

[Kalle Valo]

Johan Hovold <johan@kernel.org> writes:

> On Wed, Dec 11, 2024 at 09:35:37AM +0200, Kalle Valo wrote:
>> Johan Hovold <johan@kernel.org> writes:
>
>> > I'm seeing iwd repeatedly failing to connect with reason
>> > PREV_AUTH_NOT_VALID with WCN7850 on the Lenovo ThinkPad T14s (and the
>> > Snapdragon X1E CRD reference design).
>> >
>> > I first noticed this with 6.10 when starting to work on the Qualcomm X
>> > Elite platform, but never got around to investigating it. I've confirmed
>> > that the issue is still there with 6.13-rc1 and that adding
>> > ControlPortOverNL80211=false allows iwd to connect.
>> >
>> > Looking at Gregoire's logs now, it looks like the deauthentication
>> > reason is different:
>> >
>> > 	(Reason: 7=CLASS3_FRAME_FROM_NONASSOC_STA)
>> >
>> > instead of
>> >
>> > 	(Reason: 2=PREV_AUTH_NOT_VALID)
>> >
>> > but the same control port workaround allows iwd to connect.
>> 
>> Is there a bugzilla report about this? Unfortunately I have not been
>> able to follow bugzilla lately due to MLO work but it would be still
>> good to file this (if that's not done already).
>
> Yes, sorry, I should have included the link (it was posted later in the
> original thread):
>
> 	https://bugzilla.kernel.org/show_bug.cgi?id=218733

No, I should have checked the bugzilla myself but have been too busy.
Thanks for the link.

> I only noticed that Gregoire's symptoms were slightly different when
> replying to this thread. Let me know if you prefer that I file a
> separate bugzilla report for the PREV_AUTH_NOT_VALID case.
>
> The exact same setup (kernel, distro and router) works fine with ath11k
> and WCN6855 (e.g. as found in the Lenovo ThinkPad X13s).

Good to know. We need to get this fixed ASAP, I marked it as P1 in bugzilla.

-- 
https://patchwork.kernel.org/project/linux-wireless/list/

https://wireless.wiki.kernel.org/en/developers/documentation/submittingpatches

Re: WCN7850 disconnects and reconnects in a loop

[Johan Hovold]

On Thu, Dec 12, 2024 at 10:04:18AM +0200, Kalle Valo wrote:
> Johan Hovold <johan@kernel.org> writes:
> 
> > On Wed, Dec 11, 2024 at 09:35:37AM +0200, Kalle Valo wrote:
> >> Johan Hovold <johan@kernel.org> writes:
> >
> >> > I'm seeing iwd repeatedly failing to connect with reason
> >> > PREV_AUTH_NOT_VALID with WCN7850 on the Lenovo ThinkPad T14s (and the
> >> > Snapdragon X1E CRD reference design).
> >> >
> >> > I first noticed this with 6.10 when starting to work on the Qualcomm X
> >> > Elite platform, but never got around to investigating it. I've confirmed
> >> > that the issue is still there with 6.13-rc1 and that adding
> >> > ControlPortOverNL80211=false allows iwd to connect.
> >> >
> >> > Looking at Gregoire's logs now, it looks like the deauthentication
> >> > reason is different:
> >> >
> >> > 	(Reason: 7=CLASS3_FRAME_FROM_NONASSOC_STA)
> >> >
> >> > instead of
> >> >
> >> > 	(Reason: 2=PREV_AUTH_NOT_VALID)
> >> >
> >> > but the same control port workaround allows iwd to connect.
> >> 
> >> Is there a bugzilla report about this? Unfortunately I have not been
> >> able to follow bugzilla lately due to MLO work but it would be still
> >> good to file this (if that's not done already).
> >
> > Yes, sorry, I should have included the link (it was posted later in the
> > original thread):
> >
> > 	https://bugzilla.kernel.org/show_bug.cgi?id=218733
> 
> No, I should have checked the bugzilla myself but have been too busy.
> Thanks for the link.
> 
> > I only noticed that Gregoire's symptoms were slightly different when
> > replying to this thread. Let me know if you prefer that I file a
> > separate bugzilla report for the PREV_AUTH_NOT_VALID case.
> >
> > The exact same setup (kernel, distro and router) works fine with ath11k
> > and WCN6855 (e.g. as found in the Lenovo ThinkPad X13s).
> 
> Good to know. We need to get this fixed ASAP, I marked it as P1 in bugzilla.

For the record, this appears to have been fix in linux-next now by the
following series:

	https://lore.kernel.org/all/20250523-ath12k-unicast-key-first-v1-0-f53c3880e6d8@quicinc.com/

Johan