Current status of TL-WR1043ND

[Adrian Chadd]

On 6 June 2012 12:47, Harm Weites <harm at weites.com> wrote:
> Adrian,
>
> thanks for the additional information. Looking through the scripts
> (again) I've now added some other tools to the image.
> Are the ath* tools needed for wifi to 'just work', or are they installed
> for debug purposes? Since leaving them out frees some valuable Kbytes to
> store bsnmpd :)

The ath* tools are for debugging, yes. But believe me when I say
they're invaluable. :)

It would be better to look at how to slim down the image by adding
options to the generated 'make.conf' - eg by stripping out
kerberos/pam from the build. i haven't really put a lot of effort into
that just yet. I bet the userland can be quite significantly slimmed.

> Could you explain a bit about the issues that you are still seeing
> (regarding aggregate traffic)?

Right now I'm just debugging out of order frames and discarded frames
by the receiver. I think I've jus fixed that. Now I just need to
implement some better buffer management so higher priority traffic (eg
management) has a higher chance of having buffers available. If all
buffers are exhausted when a BAR TX needs to occur, the TX will just
fail and all traffic stops for a while. This is not a good behaviour.
:)

> Furthermore, the OpenWrt wiki mentions roughly 250Mbit throughput for
> routed traffic and you mention 50Mbit on your blog (although you still
> have debug options enabled). Did you/someone else do a test without
> debug options (and where do I disable debug options in your scripts)?

You can just disable INVARIANTS/WITNESS in the kernel. But I've not
seen 250MBit over the TL-1043ND Wifi, sorry.



Adrian