Re: [Madwifi-devel] no xmit buf

[bruno r. <br...@th...>]

On Thursday 28 February 2008 15:28:03 bruno randolf wrote:
> On Wednesday 27 February 2008 19:47:46 Emanuele Colombo wrote:
> > Hi!
> >
> > While deeply investigating the problem of "no xmit buf", I've found that
> > some time before the error message appears, the "av_mcastq" queue starts
> > filling.
> > Frames in this queue are never sent; this fact empties out the
> > "sc_txbuf" queue, and then the "no xmit buf" appears.
> >
> > I've tried to understand why it happens, but without succeding. Since I
> > needed a quite stable solution for this problem as soon as possible,
> > I've made a patch that resets the interface if the "av_mcastq" queue
> > starts filling too much (attached).
> >
> > I know this is everything but a good solution, and I don't know if this
> > could bring to a worse situation, but it actually appears to mitigate
> > the problem.
> >
> > If someone of you has a better idea or can correct the problem at the
> > source (why are that frames not sent?) I could test the solution.
> >
> > Currenlty I'm working with madwifi-0.9.4-rev3317 with some additional
> > patches to make it work on my IXP425 boards.
>
> hi emanuele!
>
> i think this was really critical information - thanks for figuring that
> out! :)
>
> do you have any way to reliably reproduce the problem and could you tell me
> how, so i could test solutions here? are you testing in IBSS or AP mode?
>
> and could you please test if the attached patch also solves the problem? it
> should avoid putting frames into this queue in the first place. that breaks
> power manangement, but it's just for testing if this would be the right
> track to solve the problem. also, elektra, i would be curious if any of the
> two patches (no-mit-buf-mcastq.diff or emanueles max_axq_depth.diff) help
> in your case.

sorry i didn't compile test my patch before. here is a version which actually 
compiles... please check and report here, thanks.

bruno