Open jcard0na opened 12 years ago
The above commit has nothing to do with this issue. It was just picked up by github because #1 appeared in the commit log.
<4>[ 106.190000] Call Trace:
<4>[ 106.190000] [<83b29074>] ath_txchainmask_reduction+0x154/0x1170
[ath9k]
<4>[ 106.190000] [<83b28fc8>] ath_txchainmask_reduction+0xa8/0x1170 [ath9k]
<4>[ 106.190000]
<4>[ 106.190000]
<4>[ 106.190000] Code: 00831821 70522002 8c630000 <8c630004> 00839021
8e2308ec 92420007 30630020 10600004
<4>[ 106.400000] ---[ end trace 574d10138db4623f ]---
Looks ath9k related.
This may have fixed it... ath9k would have crashed with no rates set up while ath5k warned.
commit 9cbbffe2ded494429b0d005a51a88242bd9b3095 Author: Bob Copeland me@bobcopeland.com Date: Wed Jan 9 12:34:55 2013 -0500
mac80211: set NEED_TXPROCESSING for PERR frames
I do not know if that this issue has already been fixed, but what is meant by "the radios will reboot"? It's just an AR93XX reset or a real AirOS/Openwrt reboot? I'm asking this because in a real-life 10 MAPs mesh (with a single Mesh Gate) and 40-50 people associated with, I see that internet browsing starts quickly then stop dead and continues a less bit quickly. I do not know if it depends on a path-switch (since its metric degrades, see hwmp ping-pong effect) or just on a transmit buffer overrun (AR93XX resets ?) due to the heavy traffic. Anyway, no devices reboot as been observed.
Submitted by David Fulgham via the mailing list:
I've been experiencing an issue with our mesh network and mesh nodes repeating constantly. After some troubleshooting it looks like I may have come across a transmit buffer overrun issue or the like.
I've been able to reproduce it quite readily by simply pushing as many packets through the mesh as possible between two nodes that have a low signal level connection (i.e. < 75dbm) and thus have a situation where there is a need to re-transmit a larger number of packets.
PC -> MeshNode1 ~~ <75dbm link ~~ Meshnode2
I then ping mesh node 2 from the PC with a command like
ping -i .01 <ip of mesh node 2>
and after a few minutes (and I assume the transmit buffer fills up and overflows) one or both of the radios will reboot. I can make the reboot happen pretty much anytime I want by moving the radios to almost out of range of each other. I'm using the openwrt trunk snapshot from Feb26th on UBNT Rocket 5M devices, and thus don't have any debugging turned on.