mac80211/ath11k regression in next-20220720

mac80211/ath11k regression in next-20220720

[Johan Hovold]

Hi,

After moving from next-20220714 to next-20220720, ath11k stopped
working here and instead spits out a bunch of warnings (see log below).

I noticed that this series from Johannes was merged in that period:

	https://lore.kernel.org/all/20220713094502.163926-1-johannes@sipsolutions.net/

but can't say for sure that it's related. I also tried adding the
follow-up fixes from the mld branch:

	https://git.kernel.org/pub/scm/linux/kernel/git/wireless/wireless-next.git/log/?h=mld

but it didn't seem to make any difference.

Any ideas about what might be going on here?

Johan


[  256.674853] wlan0: authenticate with 70:3a:cb:f7:3a:e4
[  256.687231] wlan0: bad VHT capabilities, disabling VHT
[  256.687324] wlan0: Invalid HE elem, Disable HE
[  256.687389] wlan0: 80 MHz not supported, disabling VHT
[  256.779398] wlan0: send auth to 70:3a:cb:f7:3a:e4 (try 1/3)
[  256.788290] wlan0: authenticated
[  256.807167] wlan0: 80 MHz not supported, disabling VHT
[  256.823138] wlan0: associate with 70:3a:cb:f7:3a:e4 (try 1/3)
[  256.932099] wlan0: associate with 70:3a:cb:f7:3a:e4 (try 2/3)
[  257.047894] wlan0: associate with 70:3a:cb:f7:3a:e4 (try 3/3)
[  257.156109] wlan0: association with 70:3a:cb:f7:3a:e4 timed out
[  257.171179] ath11k_pci 0006:01:00.0: failed to lookup peer 70:3a:cb:f7:3a:e4 on vdev 0
[  257.171340] ------------[ cut here ]------------
[  257.171360] WARNING: CPU: 4 PID: 71 at drivers/net/wireless/ath/ath11k/mac.c:7181 ath11k_mac_op_unassign_vif_chanctx
+0x1e8/0x2c0 [ath11k]

[  257.172255] Call trace:
[  257.172267]  ath11k_mac_op_unassign_vif_chanctx+0x1e8/0x2c0 [ath11k]
[  257.172313]  ieee80211_assign_link_chanctx+0x74/0x320 [mac80211]
[  257.172380]  __ieee80211_link_release_channel+0x58/0x150 [mac80211]
[  257.172443]  ieee80211_link_release_channel+0x3c/0x60 [mac80211]
[  257.172503]  ieee80211_destroy_assoc_data+0xcc/0x170 [mac80211]
[  257.172563]  ieee80211_sta_work+0x220/0xa20 [mac80211]
[  257.172621]  ieee80211_iface_work+0x388/0x3f0 [mac80211]
[  257.172681]  process_one_work+0x1cc/0x320
[  257.172708]  worker_thread+0x14c/0x450
[  257.172729]  kthread+0x10c/0x110
[  257.172747]  ret_from_fork+0x10/0x20
[  257.172768] ---[ end trace 0000000000000000 ]---
[  257.173752] ------------[ cut here ]------------
[  257.173771] WARNING: CPU: 4 PID: 71 at drivers/net/wireless/ath/ath11k/mac.c:6813 ath11k_mac_vdev_stop+0x134/0x1a0 [
ath11k]

[  257.175919] Call trace:
[  257.177237]  ath11k_mac_vdev_stop+0x134/0x1a0 [ath11k]
[  257.178582]  ath11k_mac_op_unassign_vif_chanctx+0x74/0x2c0 [ath11k]
[  257.179925]  ieee80211_assign_link_chanctx+0x74/0x320 [mac80211]
[  257.181281]  __ieee80211_link_release_channel+0x58/0x150 [mac80211]
[  257.182633]  ieee80211_link_release_channel+0x3c/0x60 [mac80211]
[  257.183990]  ieee80211_destroy_assoc_data+0xcc/0x170 [mac80211]
[  257.185350]  ieee80211_sta_work+0x220/0xa20 [mac80211]
[  257.186779]  ieee80211_iface_work+0x388/0x3f0 [mac80211]
[  257.188137]  process_one_work+0x1cc/0x320
[  257.189455]  worker_thread+0x14c/0x450
[  257.190771]  kthread+0x10c/0x110
[  257.192082]  ret_from_fork+0x10/0x20
[  257.193382] ---[ end trace 0000000000000000 ]---
[  257.194721] ath11k_pci 0006:01:00.0: failed to find peer vdev_id 0 addr 70:3a:cb:f7:3a:e4 in delete
[  257.196084] ath11k_pci 0006:01:00.0: failed to delete peer 70:3a:cb:f7:3a:e4 for vdev 0: -22

Re: mac80211/ath11k regression in next-20220720

[Johannes Berg]

On Fri, 2022-07-22 at 09:52 +0200, Johan Hovold wrote:
> Hi,
> 
> After moving from next-20220714 to next-20220720, ath11k stopped
> working here and instead spits out a bunch of warnings (see log below).
> 
> I noticed that this series from Johannes was merged in that period:
> 
> 	https://lore.kernel.org/all/20220713094502.163926-1-johannes@sipsolutions.net/
> 
> but can't say for sure that it's related. I also tried adding the
> follow-up fixes from the mld branch:
> 
> 	https://git.kernel.org/pub/scm/linux/kernel/git/wireless/wireless-next.git/log/?h=mld
> 
> but it didn't seem to make any difference.
> 
> Any ideas about what might be going on here?
> 

We think the "fix" is this:

https://git.kernel.org/pub/scm/linux/kernel/git/wireless/wireless-next.git/commit/?h=mld&id=dd5a559d8e90fdb9424e0580b91702c5838928dc

Do you want to try it?

Note that if that fixes it, it's still a bug in the driver, but one that
you'd otherwise not hit.

Anyway I'll do some tree shuffling today and get that in.

johannes

Re: mac80211/ath11k regression in next-20220720

[Johan Hovold]

On Fri, Jul 22, 2022 at 09:54:57AM +0200, Johannes Berg wrote:
> On Fri, 2022-07-22 at 09:52 +0200, Johan Hovold wrote:
> > Hi,
> > 
> > After moving from next-20220714 to next-20220720, ath11k stopped
> > working here and instead spits out a bunch of warnings (see log below).
> > 
> > I noticed that this series from Johannes was merged in that period:
> > 
> > 	https://lore.kernel.org/all/20220713094502.163926-1-johannes@sipsolutions.net/
> > 
> > but can't say for sure that it's related. I also tried adding the
> > follow-up fixes from the mld branch:
> > 
> > 	https://git.kernel.org/pub/scm/linux/kernel/git/wireless/wireless-next.git/log/?h=mld
> > 
> > but it didn't seem to make any difference.
> > 
> > Any ideas about what might be going on here?
> > 
> 
> We think the "fix" is this:
> 
> https://git.kernel.org/pub/scm/linux/kernel/git/wireless/wireless-next.git/commit/?h=mld&id=dd5a559d8e90fdb9424e0580b91702c5838928dc
> 
> Do you want to try it?

Thanks for the quick reply. And yes, that fixes the problem.

I apparently failed to apply all commits from that mld branch, but this
one alone fixes it.

> Note that if that fixes it, it's still a bug in the driver, but one that
> you'd otherwise not hit.

Yeah, those warnings looked like secondary issues if that's what you're
referring to?
 
> Anyway I'll do some tree shuffling today and get that in.

Thanks!

Johan

Re: mac80211/ath11k regression in next-20220720

[Johannes Berg]

On Fri, 2022-07-22 at 10:05 +0200, Johan Hovold wrote:
> > We think the "fix" is this:
> > 
> > https://git.kernel.org/pub/scm/linux/kernel/git/wireless/wireless-next.git/commit/?h=mld&id=dd5a559d8e90fdb9424e0580b91702c5838928dc
> > 
> > Do you want to try it?
> 
> Thanks for the quick reply. And yes, that fixes the problem.

OK great, thanks for checking!

> I apparently failed to apply all commits from that mld branch, but this
> one alone fixes it.

OK, I'll check later.

> > Note that if that fixes it, it's still a bug in the driver, but one that
> > you'd otherwise not hit.
> 
> Yeah, those warnings looked like secondary issues if that's what you're
> referring to?

No, I mean there are legitimate cases where override==true, and then the
driver breaks. It's just uncommon (I think perhaps only specific
userspace [debug] configurations), but it could happen.

johannes