tx_timeout problem

[Joe Parks]

On Wed, 18 May 2005 00:25:06 +0300
"vlad (bis)" <vlad at bis.net.ua> wrote:
> If I use full channel speed (450 kb/s) one of the router very often looks like
> this: "tx_timeout, hostap is restarting" and no more.. The server is
> down. 

I've noticed the same problem in Master mode with both WDS connections and regular STA connections.   In addition I've noticed the problem occasionally occurring with various ethernet drivers as well.   In all cases the problem was only occurring under high-speed, heavy throughput conditions.   

But because the problem seems to happen to all sorts of drivers (not just hostap) under those conditions I'm inclined to suspect that the problem lives in more generic queue handling code in the kernel.

I've also noticed that disabling and re-enabling the interface seems to allow recovery from the problem and is the current workaround that I'm using (scripted) when the problem is encountered (I have a daemon watching the error logs for the timeout error).

I'm using an older 2.4 kernel and have tested only briefly with 2.6 and didn't encounter the problem there so I had suspected the problem no longer existed.   But since you mention FC3 having it, I guess it must still exist.

I expect this is a problem best discussed on some other list, but please let me know if you learn anything more about it.