Problem seen with wireless mesh

2017-07-27 Thread George H
Hi

I have two wireless products forming a wireless mesh. One (let's call
it A) runs 2.6.37 kernel while the other (let's call it B) runs 3.18
kernel. The wireless mesh based on IEEE80211s runs successfully
(mac80211 module).

I see a problem where the communication between A to B freezes
intermittently, each time the duration is between a minute and two
minutes. The product A is at fault here. During this period, no ping
from a PC to A succeeds. Using a serial terminal to A, I investigated
and found:

1) Top doesn't show anything suspicious when the issue is hit. The
same processes are seen always.

2) The watchdog on A is not triggered when the issue is hit.

3) It's not a firewall issue, had tried clearing all iptables entries.

4) I checked netstat, nothing that gave further clues.

5) Added debug printk's through the mesh code in mac80211, but I'm not
seeing anything with strong evidence so far.

Any tips to debug the issue? Thanks.


RE: Problem seen with wireless mesh

2017-07-28 Thread Stam, Michel
Hello George,

Do you happen to be running AuthSAE for authenticated mesh? I've seen this go 
wrong with ath9k cards using hardware encryption; in that particular case, you 
may want to load the module with nohwcrypt=1, it should work then.

Kind regards,

Michel Stam
-Original Message-
From: linux-wireless-ow...@vger.kernel.org 
[mailto:linux-wireless-ow...@vger.kernel.org] On Behalf Of George H
Sent: Friday, July 28, 2017 3:43 AM
To: linux-wireless@vger.kernel.org
Subject: Problem seen with wireless mesh

Hi

I have two wireless products forming a wireless mesh. One (let's call
it A) runs 2.6.37 kernel while the other (let's call it B) runs 3.18
kernel. The wireless mesh based on IEEE80211s runs successfully
(mac80211 module).

I see a problem where the communication between A to B freezes
intermittently, each time the duration is between a minute and two
minutes. The product A is at fault here. During this period, no ping
from a PC to A succeeds. Using a serial terminal to A, I investigated
and found:

1) Top doesn't show anything suspicious when the issue is hit. The
same processes are seen always.

2) The watchdog on A is not triggered when the issue is hit.

3) It's not a firewall issue, had tried clearing all iptables entries.

4) I checked netstat, nothing that gave further clues.

5) Added debug printk's through the mesh code in mac80211, but I'm not
seeing anything with strong evidence so far.

Any tips to debug the issue? Thanks.



Re: Problem seen with wireless mesh

2017-07-31 Thread George H
Hi Michel,

I didn't have SAE authentication enabled, it was open mode. The latest
logs give some clues. The mesh appears to have gone down and this is
possibly the reason for the ping failure. After a while, the mesh
discovery starts again and restores the mesh. On product B, I see the
dropped_frames_no_route statistic incrementing during the period of
ping failure. I'm investigating why the mesh became inactive.

Thanks for your response.