Fwd: Re: performance question

[Marc Cousin]

> What kernel version does each one have?  What version of wireless tools?

one is 2.4.18 (the one that works) the other is 2.4.19

> When in doubt, install the latest version of wireless tools.  Also, very

wireless tools version 24

> important, what are the vendor and model number of your wireless cards, and
> what firmware type and version do they use (e.g. Lucent/Agere 6.16;
> Intersil 1.0.3).  The orinoco driver reports this on syslog when starting
> up.

i've put all of it below
hermes.c: 5 Apr 2002 David Gibson <hermes at gibson.dropbear.id.au>
orinoco.c 0.11b (David Gibson <hermes at gibson.dropbear.id.au> and others)
orinoco_plx.c 0.11b (Daniel Barlow <dan at telent.net>)
PCI: Found IRQ 12 for device 00:08.0
PCI: Sharing IRQ 12 with 00:07.2
PCI: Sharing IRQ 12 with 00:07.3
orinoco_plx: CIS:
01:03:6900:7400:61FF:7217:7404:6F67:705A:6F08:6DFF:1D:05:03:67:5A:
orinoco_plx: Local Interrupt already enabled
Detected Orinoco/Prism2 PLX device at 00:08.0 irq:12, io addr:0xa000
eth0: Station identity 001f:0003:0000:0008
eth0: Looks like an Intersil firmware version 0.08
eth0: Ad-hoc demo mode supported
eth0: IEEE standard IBSS ad-hoc mode supported
eth0: WEP supported, 104-bit key
eth0: MAC address 00:04:75:88:BD:E4
eth0: Station name "Prism  I"
eth0: ready

The card is a 3Com airconnect PCI. I had to add
 {0x10b7, 0x7770, PCI_ANY_ID, PCI_ANY_ID,}, in the pci_device_id struct in
orinoco_plx from 0.11b to have it work (pasted it from 0.12c)

> > i think the debian machine works perfectly, in all modes, with all
> > drivers, and that the gentoo machine is the real problem :
> > if i put the debian as an AP (using hostap), the gentoo can connect using
> > either driver, except from the very high cpu load i mentionned in the
> > previous mail
>
> It looks like you're doing it correctly, except for the broken orinoco
> 0.12* driver.  Revert to 0.11b.  The Gentoo machine is in Managed mode,
> right, when the Debian one is being a host-AP?

the gentoo is managed, the debian is host-ap using hostap_plx

> > if i put the debian in ad-hoc mode, i cannot connect with either orinoco
> > driver, but i can connect with the hostap driver.
>
> In some cases the machine in Ad-Hoc mode will be able to talk with an
> access point, but it isn't the right way to do it.  You should be able to
> connect if both machines are in Ad-Hoc mode, and the channels are
> explicitly set to the same value, and the WEP keys are equal or are turned
> off.  Take the default for speed; the cards will negotiate the highest
> feasible value, which is 11 Mb for normal home or office installations.
> Use iwconfig to check if you set everything up the same on both ends.
> Unfortunately there are bugs in some vendors' Ad-Hoc code; for example, I
> had endless trouble getting a Dell (Agere) card to work with a Linksys
> (Intersil) card in Ad-Hoc mode.  (I finally gave up and bought an
> inexpensive AP.)

both cards are the same, and i use them in AP/Managed or both in ad-hoc. The
weird part is that i cannot set the channel on the gentoo

> > what suprises me most is
> > that i don't have the same behaviour on both machines. for instance, i
> > can change the channel on the debian, but i don't on the gentoo (it
> > accepts the command, but iwconfig still tells me the frequency is 2.462,
> > should i put channel 10, 11, 12 or 13.)
>
> The access point controls which channel. The client (in Managed mode)
> should ignore any channel set into its card, and should automatically
> follow the AP.

I meant i cannot change the channel in adhoc mode

> > another thing i noticed : when i use the orinoco 11b driver, if i put my
> > card in managed mode, iwconfig tells there is a signal, if i put it in
> > ad-hoc mode, it says there is no signal anymore.
>
> That's not a bug, that's a feature :-)  In managed mode, there's a
> particular station (the access point) whose signal should be reported.  In
> Ad-Hoc mode there's no fixed partner in the general case, so link quality
> reports would be meaningless.  You can use iwspy to designate which partner
> should be reported.  There's a plan to include an "autospy" feature which
> will be helpful in the common case of an Ad-Hoc net with exactly two
> stations, but other areas in the driver have a lot higher priority for
> development, like the locking scheme.

Here too, I have quality reports on the debian when in ad-hoc (i'll double
check this one, but i'm quite sure of it)

> James F. Carter          Voice 310 825 2897    FAX 310 206 6673
> UCLA-Mathnet;  6115 MSA; 405 Hilgard Ave.; Los Angeles, CA, USA  90095-1555
> Email: jimc at math.ucla.edu    http://www.math.ucla.edu/~jimc (q.v. for PGP
> key)

-------------------------------------------------------