[Madwifi-users] Performance problems and errors

[GolemMTM <golem@mt...>]

Hello madwifi-users!

I'm using Linux with kernel 2.4.23, running Atheros card 5ghz
My current driver is CVS:

wlan: 0.7.2.0 BETA
ath_hal: 0.9.6.3                                                                                   
ath_pci: 0.8.4.0 BETA                                                                              
PCI: Found IRQ 12 for device 02:00.0                                                               
PCI: Sharing IRQ 12 with 00:1f.3                                                                   
ath0: 11a rates: 6Mbps 9Mbps 12Mbps 18Mbps 24Mbps 36Mbps 48Mbps 54Mbps                             
ath0: turbo rates: 6Mbps 9Mbps 12Mbps 18Mbps 24Mbps 36Mbps 48Mbps 54Mbps                           
ath0: 802.11 address: 00:05:5d:89:63:2f                                                            
ath0: Atheros 5211: mem=0xe8000000, irq=12

Why I have ath_hardstart: discard, no xmit buf
and while transferring large files transfers is chopping
and receiving hardware error resetting... I'm using turbo mode
because normal don't seems work (too many packets lost in normal mode)
Link Latency is ~1ms on 1km link, Real transfer speed is ~25mbit.
Link Quality:24/0  Signal level:-71 dBm Noise level:-95 dBm
Rate set to 54M
Why big transfers cause hardware error ?

Also my other question is about bridge ARP patch, Lastest CVS is bug
fixed ?




-- 
Pozdrowienia,
 GolemMTM

Re: [Madwifi-users] Probable source of noise in laptop affecting madwifi

[Sam Leffler <sam@er...>]

On Wednesday 07 January 2004 05:10 pm, Greg Chesson wrote:
> Goncalo Valverde wrote:
> > On Tuesday 06 January 2004 22:49, Greg Chesson wrote:
> >>Gon=E7alo Valverde wrote:
> >>>power from the battery. Btw, this only happens when using the madwifi
> >>>drivers in the desktop machine. When using either driverloader or the
> >>>computer in Windows (both in Ad-Hoc
> >>
> >>The only optimization in the windows drivers on XP is to turn off
> >>power-save mode
> >>on the wlan driver when AC power is available.    The madwifi driver
> >>doesn't implement
> >>this power-saving mode, so this would not be an issue.
> >
> >   Well,actually this could be an issue because on the laptop i'm running
> > windows and not Linux.  Linux+madwifi is only running in the desktop (in
> > master mode). I'll try instaling linux on the laptop to try the madwifi
> > driver
>
> aha.
> I think what might be happening is that the windows laptop in battery mode
> operates the driver in power-saving mode.   Your madwifi desktop is
> acting as an AP,
> but it does not (a) generate DTIM signals to the sleeping laptop, or
> (b) implement a power save queue and respond to PS-Poll messages from
> the laptop.
> If the laptop is running on AC power it won't do the power save dance
> and should work ok.    This reasoning only applies to the case where
> madwifi is the AP and is talking to a station that is using the PS
> protocol.
>

A station should never expect PS-Poll support from an AP unless the AP=20
specifies it in the capabilities it broadcasts in each beacon.  The current=
=20
AP support for madwifi does not support PS-Poll and so does not advertise i=
t=20
in beacons.

> When the roles are reversed and you see radically different behavior
> on a madwifi laptop depending on battery mode or AC power, then
> it's much more of a mystery.

As I said previously, if the OS's time maintenance is off when operating on=
=20
battery power then the ANI support in the HAL could easily get confused and=
=20
do something weird.  This seems unlikely (that time is not managed correctl=
y)=20
but would be something to check.

	Sam

Re: [Madwifi-users] Probable source of noise in laptop affecting madwifi

[Goncalo Valverde <goncalo.valverde@ne...>]

Greetings

On Thursday 08 January 2004 01:10, Greg Chesson wrote:
> Goncalo Valverde wrote:
> > On Tuesday 06 January 2004 22:49, Greg Chesson wrote:
> >>Gon=E7alo Valverde wrote:
> >   Well,actually this could be an issue because on the laptop i'm running
> > windows and not Linux.  Linux+madwifi is only running in the desktop (in
> > master mode). I'll try instaling linux on the laptop to try the madwifi
> > driver
>
> aha.
> I think what might be happening is that the windows laptop in battery mode
> operates the driver in power-saving mode.   Your madwifi desktop is
> acting as an AP,
> but it does not (a) generate DTIM signals to the sleeping laptop, or

 I've installed linux on the laptop, and the problems i was experiecing wit=
h=20
Windows seem to disapear. So this is probably an issue of windows putting t=
he=20
card in power-saver mode, and the PCI card acting as an AP not generating t=
he=20
DTIM signals you mentioned. Any workaround for this (because i want to use=
=20
the laptop both with windows and linux)

Best regards