Disconnected due to excessive missing ACKs on Ath10k (OpenWRT Chaos Chalmer RC2)

Discussion:

Fernando S

2015-07-14 17:29:03 UTC

I'm having dropouts on my TP-Link Archer C7, 2.4Ghz Wlan runs perfect,
but 5Ghz is unstable (my laptop and my Ipad 4 loss connection after a
couple of minutes, specially on heavy traffic).
Looking on router's log, I found:

Thu Jul 9 20:51:32 2015 daemon.info hostapd: wlan0: STA
01:02:03:04:05:06 IEEE 802.11: disconnected due to excessive missing
ACKs Thu Jul 9 20:52:02 2015 daemon.info hostapd: wlan0: STA
01:02:03:04:05:06 IEEE 802.11: deauthenticated due to inactivity
(timer DEAUTH/REMOVE)

And on client side:

Jul 9 20:51:05 laptop wpa_supplicant[1011]: wlan0:
CTRL-EVENT-DISCONNECTED bssid=30:30:30:30:30:30 reason=7 Jul 9
20:51:05 laptop NetworkManager[861]: <warn> Connection disconnected
(reason 7)
A little search on Google show me some old bugs, apparently resolved
and only on ath9k (2.4Ghz): maybe the same issue is present on ath10k
(5Ghz)?

Please let me know if I can help to debug this issue.

Thanks and regards.

Michal Kazior

2015-07-15 07:09:38 UTC

Permalink

Post by Fernando S
I'm having dropouts on my TP-Link Archer C7, 2.4Ghz Wlan runs perfect,
but 5Ghz is unstable (my laptop and my Ipad 4 loss connection after a
couple of minutes, specially on heavy traffic).
Thu Jul 9 20:51:32 2015 daemon.info hostapd: wlan0: STA
01:02:03:04:05:06 IEEE 802.11: disconnected due to excessive missing
ACKs Thu Jul 9 20:52:02 2015 daemon.info hostapd: wlan0: STA
01:02:03:04:05:06 IEEE 802.11: deauthenticated due to inactivity
(timer DEAUTH/REMOVE)

Hmm.. This means there had to be a peer kickout event generated in
firmware and your hostapd is configured to handle low-ack events as
disassoc triggers.

Ben recently raised a problem concerning ACK problems [1]. Perhaps
that's one of the manifestations.

You could just try disabling the `disassoc_low_ack` option in hostapd.
Stations will still be disassociated when they go out of range and
don't respond to NullFunc frames pings after a few minutes of idling
(from AP point of view).

[1]: http://lists.infradead.org/pipermail/ath10k/2015-July/005618.html

Post by Fernando S
CTRL-EVENT-DISCONNECTED bssid=30:30:30:30:30:30 reason=7 Jul 9
20:51:05 laptop NetworkManager[861]: <warn> Connection disconnected
(reason 7)
A little search on Google show me some old bugs, apparently resolved
and only on ath9k (2.4Ghz): maybe the same issue is present on ath10k
(5Ghz)?
Please let me know if I can help to debug this issue.

Can you point to the ath9k thread/patches, please?

Michał

Fernando S

2015-07-15 19:31:37 UTC

Permalink

Same suggestion was posted on Openwrt forum (disable
disassoc_low_ack), I'll try it, do some stress test and come back with
results.

Here are two links to Openwrt forum threads related to this problem:

https://forum.openwrt.org/viewtopic.php?id=43188
https://forum.openwrt.org/viewtopic.php?pid=228738

And two tickets with some more information:

https://dev.openwrt.org/ticket/12372
https://dev.openwrt.org/ticket/15464

Post by Michal Kazior

Hmm.. This means there had to be a peer kickout event generated in
firmware and your hostapd is configured to handle low-ack events as
disassoc triggers.
Ben recently raised a problem concerning ACK problems [1]. Perhaps
that's one of the manifestations.
You could just try disabling the `disassoc_low_ack` option in hostapd.
Stations will still be disassociated when they go out of range and
don't respond to NullFunc frames pings after a few minutes of idling
(from AP point of view).
[1]: http://lists.infradead.org/pipermail/ath10k/2015-July/005618.html

Can you point to the ath9k thread/patches, please?
Michał

Michal Kazior

2015-07-16 05:42:00 UTC

Permalink

Post by Fernando S
Same suggestion was posted on Openwrt forum (disable
disassoc_low_ack), I'll try it, do some stress test and come back with
results.
https://forum.openwrt.org/viewtopic.php?id=43188
https://forum.openwrt.org/viewtopic.php?pid=228738
https://dev.openwrt.org/ticket/12372

This seems to be related to ANI. I guess you could try start the
ath10k AP and issue:

echo 0 > /sys/kernel/debug/ieee80211/phy*/ath10k/ani_enable

You can check the current state of the setting by reading the file obviously.

Post by Fernando S
https://dev.openwrt.org/ticket/15464

This one is related to software A-MPDU support in ath9k. ath10k
offloads Tx aggregation to firmware so this bug doesn't really apply.

Michał

f***@gmail.com

2015-07-19 03:04:39 UTC

Permalink

Michal, do you know how to enable debug? Because I haven't that file, the directory with some other files exists, but there is not one called "ani_enable".
I suspect some debugging option not enabled on firmware/driver.

Post by Michal Kazior

This seems to be related to ANI. I guess you could try start the
echo 0 > /sys/kernel/debug/ieee80211/phy*/ath10k/ani_enable
You can check the current state of the setting by reading the file obviously.

Post by Fernando S
https://dev.openwrt.org/ticket/15464

This one is related to software A-MPDU support in ath9k. ath10k
offloads Tx aggregation to firmware so this bug doesn't really apply.
Michał

Michal Kazior

2015-07-20 05:02:21 UTC

Permalink

Post by f***@gmail.com
Michal, do you know how to enable debug? Because I haven't that file, the directory with some other files exists, but there is not one called "ani_enable".
I suspect some debugging option not enabled on firmware/driver.

You need to enable mac80211 debugfs and ath debug options in OpenWRTs
menuconfig.

Michał

Kevin Darbyshire-Bryant

2015-07-20 08:26:33 UTC

Permalink

Post by Michal Kazior

You need to enable mac80211 debugfs and ath debug options in OpenWRTs
menuconfig.

If it helps, menuconfig:

Kernel modules -> Wireless drivers -> kmod-ath -> Atheros wireless debugging

Kevin