Comment 32 for bug 941695

So you have tested the patch from the fedora bug, "brcm80211: smac: fix endless retry of A-MPDU transmissions", and found that it didn't help, correct? Really I think that's only intended to be a fix for the warning that's triggered in brcms_c_wait_for_tx_completion(), which seems like it was a symptom of the problem and not related to the cause. These attempts to fix the warning don't seem to be getting at the root of your hangs and disconnects.

I'd say that some of the AMPDU messages are normal, especially in noisy RF environments, but an extreme number of them may be indicative of a problem.

The illegal channel messages are more interesting. The driver is trying to transmit some packets, but the hardware comes back with an error that the comments in the code note as "channel mismatch." The driver won't retry when it sees this error, so the BA timeout messages are a result of the code iterating through all the fragments that were part of this failed transfer. I'm going to have to go through the code more to try and better understand what it all means.

Once you get into this state, do you continually get a stream of these messages when trying to use the network until you use network manager to reconnect? Can you explain exactly what you're doing when you use network manager to reconnect? I.e., do you disable and re-enable wirless, or do you just disconnect and reconnect to the network?

Can you also trigger this error and take note of whether the channel number in the "illegal channel" messages matches the channel the AP you're connected to is using?

Thanks!