Re: RESOLVED! was: NetGear MA401, firmware 1.3.6, cannot use hostap mode, massive transmission errors, difficulty speaking to the rest of the network


From: Eric Johanson (ericj_at_cubesearch.com)
Date: 2002-09-13 07:12:42 UTC



It sounds like you are using fw 8.10, which is buggy as all getout. I stick with 6.06, and I've never had any problems like you described below, on quite a bit of hardware:

TI cardbus pcmcia chipsets
isa<>pcmcia vadem 395 bridges
pci<>pcmcia ricoh bridges
tcip (toshiba pcmcia)

If you are getting errors about Tx interrupts being dropped, then something is up with either your AP, or the card. That generally means the cards output buffers didn't get cleared, due to restranmissions, etc.

the wvlan_cs driver (other than obsolete) doesn't support wireless tools very well, based on the old HCF code from lucent, and has no methods for getting SNR/sig/quality numbers while in iBSS mode.

*shrug*, honestly, you are the first person I've heard of who still uses that driver.

I'm using all lucent silver cards, if that matters. I've never played with WEP or gold cards. Do you have wep enabled?

-Eric

On Fri, 13 Sep 2002, Brian Capouch wrote:

> Eric Johanson wrote:> Of all of the card/driver combos, I find the
> orinoco_cs to work the best
> > out of all of them.
> >
> > I've never had a single problem taking to any AP, iBSS or BSS mode.
> >
> > Using kernel 2.4.18, patched (monitor mode patches from snax/airsnort)
> > orinoco_cs .11b
> >
> > However, it sucks with prism cards.
> >
> > Could you provide more detail on what type of problems you've had?
> >
>
> From a userland perspective, dropouts in pings, long latencies, and
> stilted throughput.
>
> Many kernel errors logged about Tx interrupts being dropped.
>
> I would think perhaps I'm misconfiguring things wrt IRQ assigments, but
> it has happened on so many different machines, with and without so many
> other combinations of cards, and it is so consistent for me, that I
> really can't believe *all* of them are wrong.
>
> It is an erratic thing, though, and usually occurs only after the card
> has associated and been in use for a bit.
>
> It usually winds up getting me as "I rebuild the pcmcia tools, which
> overwrites my default /etc/pcmcia/config file, and the driver is then
> loaded on bootup. Then sometime later I notice my throughput shot to
> hell, and I remember that I'm using that driver."
>
> The wvlan_cs driver works PERFECTLY in each of these cases, with all
> other factors held constant, which further fuels my certainty that it's
> not a mis-configured interrupt issue.
>
> I've had this as well in desktop machines, both with ISA and PCI adapter
> cards. And the switch to the wvlan_cs driver always fixes it there, too.
>
> If they ever remove that "obsolete" driver I'm really going to be hurting.
>
> B.
>
>
>



This archive was generated by hypermail 2.1.4.