MADWIFI

Andrew Gaffney wrote:
> I've got a D-link DWL-AB650 card. Using the driver pulled from CVS on 
> 02-12-2004, it seems to be working fine, except I lose about 1/3 of my 
> packets. I did a 'ping -c 100 192.168.0.2' and lost 37% of the packets 
> (that IP is the AP/router's IP). I just pulled the driver from CVS today 
> and it's still dropping packets. It's also now giving me the error 
> 'ping: sendmsg: No buffer space available' when I try to run the same 
> command as above. I'm monitoring my signal with wavemon and have a 
> steady 58/94. Is anyone else experiencing this?

More info: modules are compiled against a 2.6.4 kernel with Gentoo patches. Card is in a 
Thinkpad 770. No problems with card in Windows.

-- 
Andrew Gaffney
Network Administrator
Skyline Aeronautics, LLC.
636-357-1548

Andrew Gaffney wrote:
> Andrew Gaffney wrote:
> 
>> I've got a D-link DWL-AB650 card. Using the driver pulled from CVS on 
>> 02-12-2004, it seems to be working fine, except I lose about 1/3 of my 
>> packets. I did a 'ping -c 100 192.168.0.2' and lost 37% of the packets 
>> (that IP is the AP/router's IP). I just pulled the driver from CVS 
>> today and it's still dropping packets. It's also now giving me the 
>> error 'ping: sendmsg: No buffer space available' when I try to run the 
>> same command as above. I'm monitoring my signal with wavemon and have 
>> a steady 58/94. Is anyone else experiencing this?
> 
> 
> More info: modules are compiled against a 2.6.4 kernel with Gentoo 
> patches. Card is in a Thinkpad 770. No problems with card in Windows.

Does anyone else have this problem with the D-link DWL-AB650 and the madwifi driver?

-- 
Andrew Gaffney
Network Administrator
Skyline Aeronautics, LLC.
636-357-1548

Hi Michael,

I have been successful in getting my 5211 (802.11g Proxim card) and 
5212 (802.11a/b and 802.11a/b/g proxim cards) working with with my 
WRT54G v2 (Interestingly enough, the packaging for my AP says it's a 
v1.1, but the hardware itself is a v2!)

I updated to version 2.02.7 of the firmware (the AP shipped with 2.02). 
and I reset the configuration to defaults just to be on the safe side.

There are some weird issues with you throw the cards into a specific 
mode... sometimes it works, sometimes it doesn't, but when the card is 
left on auto things seem fine.

A few things to try:

Make sure your interface is up (ifconfig ath0 up) or you won't 
associate.  The driver leaves the card in a "down" state by default, so 
unless you have a script that does this for you, or you do it by hand, 
it won't come up.

Also, you might want to fiddle with the b/g/mixedmode options on the 
AP.   You can find the options under the Wireless->Basic Wireless 
Settings tab on the web interface.

Good luck.

If you provide more information about your specific setup (commands 
issued, etc...), I can probably help do some more testing.

- Terry

On Mar 25, 2004, at 9:46 PM, Michael Halcrow wrote:

> Hi.  Me again.
>
> I picked up a Linksys WRT54G AP/router/switch wonder box during my
> lunch break (the Fry's deli in Austin makes some fairly decent buffalo
> wings), and I was unable to get the madwifi Atheros driver to talk to
> it.  My Linksys 802.11b PC-card had no problems communicating with it
> though.  I then decided, just for kicks, to try out the ndiswrapper.
> I grabbed the binary drivers from D-Link's site
> (DWL-G520B_driver_317.zip) and has ndiswrapper do its thing with
> them.  That gave me this message (in case anyone here cares):
>
> ndiswrapper: Driver is serialized. This will not work!
>
> As predicted, it did not work (although I did get a network interface
> that I could parameterize).  That I thought I would try out the
> Linuxant thing.  Lo and behold, it works!  54Mbps and everything.  Of
> course, Linuxant will set me back $20.
>
> So, my card is good, and the official drivers, with Linuxant, work,
> but the Atheros driver is a no-go.  I'm pretty fluent at kernel
> debugging, so if you want me to poke around at all, just point out
> where you think I should be looking at this point...
>
> Mike
> .___________________________________________________________________.
>                          Michael A. Halcrow
>        Security Software Engineer, IBM Linux Technology Center
> GnuPG Fingerprint: 05B5 08A8 713A 64C1 D35D  2371 2D3C FDDA 3EB6 601D
>
> Who is General Failure and why is he reading my hard disk?

I have stopped using the madwifi driver because the Linuxant wrapper and
the latest D-Link drivers) is much, much more reliable (although not
perfect) and it gives a higher speed (54 Mb/s vs 36 Mb/s). (The
NDISwrapper did not work at all.)

I am using a D-Link DWL-G650 (B3) card with a D-Link DI-624 access
point with Fedora Core 1.

--- Vladimir

------------------------------------------------------------------------
Vladimir G. Ivanovic                        http://leonora.org/~vladimir
2770 Cowper St.                                         vladimir@...
Palo Alto, CA 94306-2447                                 +1 650 678 8014
------------------------------------------------------------------------

>>>>> "mh" == Michael Halcrow <mike@...> writes:

    mh> Hi.  Me again.
    mh> 
    mh> I picked up a Linksys WRT54G AP/router/switch wonder box during my
    mh> lunch break (the Fry's deli in Austin makes some fairly decent buffalo
    mh> wings), and I was unable to get the madwifi Atheros driver to talk to
    mh> it.  My Linksys 802.11b PC-card had no problems communicating with it
    mh> though.  I then decided, just for kicks, to try out the ndiswrapper.
    mh> I grabbed the binary drivers from D-Link's site
    mh> (DWL-G520B_driver_317.zip) and has ndiswrapper do its thing with
    mh> them.  That gave me this message (in case anyone here cares):
    mh> 
    mh> ndiswrapper: Driver is serialized. This will not work!
    mh> 
    mh> As predicted, it did not work (although I did get a network interface
    mh> that I could parameterize).  That I thought I would try out the
    mh> Linuxant thing.  Lo and behold, it works!  54Mbps and everything.  Of
    mh> course, Linuxant will set me back $20.
    mh> 
    mh> So, my card is good, and the official drivers, with Linuxant, work,
    mh> but the Atheros driver is a no-go.  I'm pretty fluent at kernel
    mh> debugging, so if you want me to poke around at all, just point out
    mh> where you think I should be looking at this point...
    mh> 
    mh> Mike
    mh> .___________________________________________________________________.
    mh>                          Michael A. Halcrow                          
    mh>        Security Software Engineer, IBM Linux Technology Center       
    mh> GnuPG Fingerprint: 05B5 08A8 713A 64C1 D35D  2371 2D3C FDDA 3EB6 601D
    mh> 
    mh> Who is General Failure and why is he reading my hard disk?

Can you show me where this "Linuxant" can be had?  I think I heard about
it on /. but I forget. 

On the latest CVS of madwifi, I now get my full 54M speed from all 3 of
my cards (2 DWL-G520 and 1 DWL-G650 connected to a DI-624 Router).
Previous revisions did not enable full speed on the DWL-G650.

Also, have you had any luck enabling the SuperG mode with the Linuxant
wrapper?  That would be the only reason I'd change off of madwifi ( well
or if someone writes a fully open-source driver)

Thanks,

Brandon

On Fri, 03/26/04 at 02:10:07 -0800, Vladimir G. Ivanovic wrote:
> I have stopped using the madwifi driver because the Linuxant wrapper and
> the latest D-Link drivers) is much, much more reliable (although not
> perfect) and it gives a higher speed (54 Mb/s vs 36 Mb/s). (The
> NDISwrapper did not work at all.)
> 
> I am using a D-Link DWL-G650 (B3) card with a D-Link DI-624 access
> point with Fedora Core 1.
> 
> --- Vladimir
> 
> ------------------------------------------------------------------------
> Vladimir G. Ivanovic                        http://leonora.org/~vladimir
> 2770 Cowper St.                                         vladimir@...
> Palo Alto, CA 94306-2447                                 +1 650 678 8014
> ------------------------------------------------------------------------
> 
> >>>>> "mh" == Michael Halcrow <mike@...> writes:
> 
>     mh> Hi.  Me again.
>     mh> 
>     mh> I picked up a Linksys WRT54G AP/router/switch wonder box during my
>     mh> lunch break (the Fry's deli in Austin makes some fairly decent buffalo
>     mh> wings), and I was unable to get the madwifi Atheros driver to talk to
>     mh> it.  My Linksys 802.11b PC-card had no problems communicating with it
>     mh> though.  I then decided, just for kicks, to try out the ndiswrapper.
>     mh> I grabbed the binary drivers from D-Link's site
>     mh> (DWL-G520B_driver_317.zip) and has ndiswrapper do its thing with
>     mh> them.  That gave me this message (in case anyone here cares):
>     mh> 
>     mh> ndiswrapper: Driver is serialized. This will not work!
>     mh> 
>     mh> As predicted, it did not work (although I did get a network interface
>     mh> that I could parameterize).  That I thought I would try out the
>     mh> Linuxant thing.  Lo and behold, it works!  54Mbps and everything.  Of
>     mh> course, Linuxant will set me back $20.
>     mh> 
>     mh> So, my card is good, and the official drivers, with Linuxant, work,
>     mh> but the Atheros driver is a no-go.  I'm pretty fluent at kernel
>     mh> debugging, so if you want me to poke around at all, just point out
>     mh> where you think I should be looking at this point...
>     mh> 
>     mh> Mike
>     mh> .___________________________________________________________________.
>     mh>                          Michael A. Halcrow                          
>     mh>        Security Software Engineer, IBM Linux Technology Center       
>     mh> GnuPG Fingerprint: 05B5 08A8 713A 64C1 D35D  2371 2D3C FDDA 3EB6 601D
>     mh> 
>     mh> Who is General Failure and why is he reading my hard disk? 
> 
> 
> -------------------------------------------------------
> This SF.Net email is sponsored by: IBM Linux Tutorials
> Free Linux tutorial presented by Daniel Robbins, President and CEO of
> GenToo technologies. Learn everything from fundamentals to system
> administration.http://ads.osdn.com/?ad_id=1470&alloc_id=3638&op=click
> _______________________________________________
> Madwifi-users mailing list
> Madwifi-users@...
> https://lists.sourceforge.net/lists/listinfo/madwifi-users

Are you using ACPI?  Or do you have the BIOS settings for automatic CPU
powersavings set?

The card does that here on my T40 if the cpu goes into the C3 state.  This
seems to happen on battery even when using APM when I have the BIOS cpu
powersavings set.  My best guess is something improperly handling the
busmastering handling necessary for working with the C3 state, but I haven't
had a chance to fully dig into that.

X might be using enough CPU that it never makes it all the way down into the
C3 state when it's powersaving.

				-- Gerald

On Thu, Mar 25, 2004 at 04:26:49PM -0800, Elliot F. wrote:
> I have madwifi working with 2.6.3 on a T40 (2379) running Debian (Sarge).
> The odd thing about the drivers (when running on my laptop) is that they
> won't work, unless I have X running.  If I try ifup'ing the interface on the
> console (pre X) I get incessant messages about the card being reset.  If I
> start X, I don't get those messages on the console/syslog.  If I quit X
> without ifdown'ing the interface, I get the messages again.  Odd, no?

Gerald,

I am using ACPI, but it is rather broken in some respects (does not turn off laptop, mostly).

I am using the frequency scaling, but not automatically controlled, it's all manual.

Elliot F.

On Thu, 2004-03-25 at 17:11, Gerald Britton wrote:
> Are you using ACPI?  Or do you have the BIOS settings for automatic
> CPU
> powersavings set?
> 
> The card does that here on my T40 if the cpu goes into the C3 state.
> This
> seems to happen on battery even when using APM when I have the BIOS
> cpu
> powersavings set.  My best guess is something improperly handling the
> busmastering handling necessary for working with the C3 state, but I
> haven't
> had a chance to fully dig into that.
> 
> X might be using enough CPU that it never makes it all the way down
> into the
> C3 state when it's powersaving.
> 
> 				-- Gerald
> 
> On Thu, Mar 25, 2004 at 04:26:49PM -0800, Elliot F. wrote:
> > I have madwifi working with 2.6.3 on a T40 (2379) running Debian
> (Sarge).
> > The odd thing about the drivers (when running on my laptop) is that
> they
> > won't work, unless I have X running.  If I try ifup'ing the
> interface on the
> > console (pre X) I get incessant messages about the card being
> reset.  If I
> > start X, I don't get those messages on the console/syslog.  If I
> quit X
> > without ifdown'ing the interface, I get the messages again.  Odd,
> no?
> 
> 
> 
> -------------------------------------------------------
> This SF.Net email is sponsored by: IBM Linux Tutorials
> Free Linux tutorial presented by Daniel Robbins, President and CEO of
> GenToo technologies. Learn everything from fundamentals to system
> administration.http://ads.osdn.com/?ad_id=1470&alloc_id=3638&op=click
> _______________________________________________
> Madwifi-users mailing list
> Madwifi-users@...
> https://lists.sourceforge.net/lists/listinfo/madwifi-users

On Thu, Mar 25, 2004 at 06:04:43PM -0800, Elliot F. wrote:
> Gerald,
> 
> I am using ACPI, but it is rather broken in some respects (does not turn off
> laptop, mostly).

If you have the "processor" ACPI kernel module loaded, ACPI will switch the
processor between C-states on demand.  This gains you significant powersavings
(people have been reporting upto an additional hour in normal use ontop of
the typical 4-5h the T40 normally gets).

> I am using the frequency scaling, but not automatically controlled, it's all
> manual.

frequency scaling != powersaving modes, they're independant.  If you're using
ACPI, none of the bios settings should matter, they are still applied though
if you're using APM.  The frequency is never automatically shifted around, but
the powersavings settings still are.  With ACPI, none of this is under bios
control, it's all controlled by Linux.

Anyway, I'll bet the acpi "processor" module is what's killing your card.

				-- Gerald

> Elliot F.
> 
> On Thu, 2004-03-25 at 17:11, Gerald Britton wrote:
> > Are you using ACPI?  Or do you have the BIOS settings for automatic
> > CPU
> > powersavings set?
> > 
> > The card does that here on my T40 if the cpu goes into the C3 state.
> > This
> > seems to happen on battery even when using APM when I have the BIOS
> > cpu
> > powersavings set.  My best guess is something improperly handling the
> > busmastering handling necessary for working with the C3 state, but I
> > haven't
> > had a chance to fully dig into that.
> > 
> > X might be using enough CPU that it never makes it all the way down
> > into the
> > C3 state when it's powersaving.
> > 
> > 				-- Gerald
> > 
> > On Thu, Mar 25, 2004 at 04:26:49PM -0800, Elliot F. wrote:
> > > I have madwifi working with 2.6.3 on a T40 (2379) running Debian
> > (Sarge).
> > > The odd thing about the drivers (when running on my laptop) is that
> > they
> > > won't work, unless I have X running.  If I try ifup'ing the
> > interface on the
> > > console (pre X) I get incessant messages about the card being
> > reset.  If I
> > > start X, I don't get those messages on the console/syslog.  If I
> > quit X
> > > without ifdown'ing the interface, I get the messages again.  Odd,
> > no?
> > 
> > 
> > 
> > -------------------------------------------------------
> > This SF.Net email is sponsored by: IBM Linux Tutorials
> > Free Linux tutorial presented by Daniel Robbins, President and CEO of
> > GenToo technologies. Learn everything from fundamentals to system
> > administration.http://ads.osdn.com/?ad_id=1470&alloc_id=3638&op=click
> > _______________________________________________
> > Madwifi-users mailing list
> > Madwifi-users@...
> > https://lists.sourceforge.net/lists/listinfo/madwifi-users
> 
> 
> 
> -------------------------------------------------------
> This SF.Net email is sponsored by: IBM Linux Tutorials
> Free Linux tutorial presented by Daniel Robbins, President and CEO of
> GenToo technologies. Learn everything from fundamentals to system
> administration.http://ads.osdn.com/?ad_id=1470&alloc_id=3638&op=click
> _______________________________________________
> Madwifi-users mailing list
> Madwifi-users@...
> https://lists.sourceforge.net/lists/listinfo/madwifi-users
>

Gerald,

Thanks for the info, but the thing is that my card works fine, I just need to have X running.  That's the only thing I find odd.

I'll compile the kernel without processor.o next time, see if I still get the issues.

Elliot F.

On Thu, 2004-03-25 at 19:05, Gerald Britton wrote:
> On Thu, Mar 25, 2004 at 06:04:43PM -0800, Elliot F. wrote:
> > Gerald,
> >
> > I am using ACPI, but it is rather broken in some respects (does not
> turn off
> > laptop, mostly).
> 
> If you have the "processor" ACPI kernel module loaded, ACPI will
> switch the
> processor between C-states on demand.  This gains you significant
> powersavings
> (people have been reporting upto an additional hour in normal use
> ontop of
> the typical 4-5h the T40 normally gets).
> 
> > I am using the frequency scaling, but not automatically controlled,
> it's all
> > manual.
> 
> frequency scaling != powersaving modes, they're independant.  If
> you're using
> ACPI, none of the bios settings should matter, they are still applied
> though
> if you're using APM.  The frequency is never automatically shifted
> around, but
> the powersavings settings still are.  With ACPI, none of this is
> under bios
> control, it's all controlled by Linux.
> 
> Anyway, I'll bet the acpi "processor" module is what's killing your
> card.
> 
> 				-- Gerald
> 
> > Elliot F.
> >
> > On Thu, 2004-03-25 at 17:11, Gerald Britton wrote:
> > > Are you using ACPI?  Or do you have the BIOS settings for
> automatic
> > > CPU
> > > powersavings set?
> > >
> > > The card does that here on my T40 if the cpu goes into the C3
> state.
> > > This
> > > seems to happen on battery even when using APM when I have the
> BIOS
> > > cpu
> > > powersavings set.  My best guess is something improperly handling
> the
> > > busmastering handling necessary for working with the C3 state,
> but I
> > > haven't
> > > had a chance to fully dig into that.
> > >
> > > X might be using enough CPU that it never makes it all the way
> down
> > > into the
> > > C3 state when it's powersaving.
> > >
> > > 				-- Gerald
> > >
> > > On Thu, Mar 25, 2004 at 04:26:49PM -0800, Elliot F. wrote:
> > > > I have madwifi working with 2.6.3 on a T40 (2379) running
> Debian
> > > (Sarge).
> > > > The odd thing about the drivers (when running on my laptop) is
> that
> > > they
> > > > won't work, unless I have X running.  If I try ifup'ing the
> > > interface on the
> > > > console (pre X) I get incessant messages about the card being
> > > reset.  If I
> > > > start X, I don't get those messages on the console/syslog.  If
> I
> > > quit X
> > > > without ifdown'ing the interface, I get the messages again.
> Odd,
> > > no?
> > >
> > >
> > >
> > > -------------------------------------------------------
> > > This SF.Net email is sponsored by: IBM Linux Tutorials
> > > Free Linux tutorial presented by Daniel Robbins, President and
> CEO of
> > > GenToo technologies. Learn everything from fundamentals to system
> > > administration.http://ads.osdn.com/
> ?ad_id=1470&alloc_id=3638&op=click
> > > _______________________________________________
> > > Madwifi-users mailing list
> > > Madwifi-users@...
> > > https://lists.sourceforge.net/lists/listinfo/madwifi-users
> >
> >
> >
> > -------------------------------------------------------
> > This SF.Net email is sponsored by: IBM Linux Tutorials
> > Free Linux tutorial presented by Daniel Robbins, President and CEO
> of
> > GenToo technologies. Learn everything from fundamentals to system
> > administration.http://ads.osdn.com/
> ?ad_id=1470&alloc_id=3638&op=click
> > _______________________________________________
> > Madwifi-users mailing list
> > Madwifi-users@...
> > https://lists.sourceforge.net/lists/listinfo/madwifi-users
> >

On Thu, Mar 25, 2004 at 11:03:27PM -0800, Elliot F. wrote:
> Gerald,
> 
> Thanks for the info, but the thing is that my card works fine, I just
> need to have X running.  That's the only thing I find odd.

As I explained before, Your system probably never makes it all the way down
into the C3 state while running X.  On my system if I run something that
generates CPU usage, even something as simple as holding space down in top,
my T40 never makes it down into C3 and my card stops constantly resetting.

				-- Gerald

> I'll compile the kernel without processor.o next time, see if I still get
> the issues.
> 
> Elliot F.

Ramon,

I see a few issues here...

One problem is that you are actually able to associate to the network 
without encryption turned on.  What I have typically seen is that when 
802.1x is enabled, you can ONLY associate when a key (even a bogus one) 
is set.  So it seems odd to me that you can associate without a key 
set.

You didn't have encryption turned on when you ran xsupplicant.  The CVS 
code will currently turn this on for you (though this will be 
changing), but that causes you to disassociate from the network (this 
is normal), but because you were able to associate without a key turned 
on, it makes me wonder what is going on.

Can you associate to the network with a WEP key set (without 
xsupplicant running)?

You will probably have to issue something like:

iwconfig ath0 key 1234567890
iwconfig ath0 enc on

In order to get encryption working properly since the MADWifi patch 
seems to have broken manual key sets, though that combination works for 
me.  (I'm going to take a stab at figuring out why this is... and 
possibly submit another patch.)

Also, please send me (off madwifi list, or to the xsupplicant list) the 
following:

Your xsupplicant configuration (no passwords please).
Which AP are you using?

More comments inline.

> Hi Terry,
>
> Thanks for your comments. Nevertheless I'm unable to use it (I'm quite
> ignorant regarding all this, so I am not of much help). I applied your
> patch (even the patch in the xsupplicant cvs) but still no luck... 
> Maybe
> you can give me some hints or pointers on what is going on...
>
> I managed to use xsupplicant 0.8b (pre cvs) with TTLS, but affter the 
> keys
> were negotiated, the card moved to scan mode, so I followed your
> suggestion:

Ok... so this doesn't surprise me too much.   Your card associates 
unencrypted, and gets kicked off as soon as the keys are set.  Does 
your AP allow the use of 802.1x without encryption?  Xsupplicant isn't 
going to be able to handle this very well, though we are looking at 
fixing this in CVS.

> Todays Madwifi & xsupplicant cvs (with the new config format)
>
> ath_hal: module license 'Proprietary' taints kernel.
> ath_hal: 0.9.8.6
> wlan: 0.7.3.1 BETA
> ath_pci: 0.8.5.4 BETA
> Setup queue (0) for WME_AC_BK
> Setup queue (1) for WME_AC_BE
> Setup queue (2) for WME_AC_VI
> Setup queue (3) for WME_AC_VO
> ath0: mac 4.2 phy 3.0 5ghz radio 1.7 2ghz radio 2.3
> ath0: 11a rates: 6Mbps 9Mbps 12Mbps 18Mbps 24Mbps 36Mbps 48Mbps 54Mbps
> ath0: 11b rates: 1Mbps 2Mbps 5.5Mbps 11Mbps
> ath0: 802.11 address: 00:05:4e:40:22:0a
> ath0: Atheros 5211: mem=0xc0210000, irq=9

I will play a little bit more with my 5211 and see if I see similar 
problems.  That card was the most troublesome for me in my testing, 
besides the 802.11a issues.

Are you connecting to an 802.11b or 802.11g AP? ( I don't know if this 
is all that important, but it might help me test your problem.)

> Interface is up & ESSID SET
> ath0      IEEE 802.11  ESSID:"EPerm"
>           Mode:Managed  Frequency:2.442GHz  Access Point: 
> 00:0E:38:4E:45:90
>           Bit Rate:11Mb/s   Tx-Power:off   Sensitivity=0/3
>           Retry:off   RTS thr:off   Fragment thr:off
>           Encryption key:off
>           Power Management:off
>           Link Quality:40/94  Signal level:-55 dBm  Noise level:-95 dBm
>           Rx invalid nwid:0  Rx invalid crypt:0  Rx invalid frag:0
>           Tx excessive retries:0  Invalid misc:0   Missed beacon:0
>
> gandalf casellas# xsupplicant -i ath0 -d 12
> "Starting Xsupplicant"

Notice that encryption is off, but you are associated to the AP... this 
seems strange to me, given my testing of 802.1x & associated clients.

>
>
>
> The log shows (see below)
> .....
> [INT] All Fs for dest mac!  Resetting keys!
> .....
>
> which I can check
>
> ath0      IEEE 802.11  ESSID:"EPerm"
>           Mode:Managed  Frequency:2.452GHz  Access Point:
> FF:FF:FF:FF:FF:FF
>           Bit Rate:1Mb/s   Tx-Power:off   Sensitivity=0/3
>           Retry:off   RTS thr:off   Fragment thr:off
>           Encryption key:0000-0000-0000-0000-0000-0000-00   Security
> mode:open
>           Power Management:off
>           Link Quality:0/94  Signal level:-95 dBm  Noise level:-95 dBm
>           Rx invalid nwid:0  Rx invalid crypt:0  Rx invalid frag:0
>           Tx excessive retries:0  Invalid misc:0   Missed beacon:0

xsupplicant has turned encryption on for you, which kicked you off the 
network... you might want to try letting the supplicant run for a few 
minutes and see if it can end up associating for you, or try manually 
associating with a WEP key as mentioned above.  If you have encryption 
turned off on your 802.1x AP then that is probably the problem.

- Terry

Hi Terry,

Thanks for your answer and patience, please see comments inline.


On Thu, 25 Mar 2004, Terry Simons wrote:

> One problem is that you are actually able to associate to the network 
> without encryption turned on.  What I have typically seen is that when 
> 802.1x is enabled, you can ONLY associate when a key (even a bogus one) 
> is set.  So it seems odd to me that you can associate without a key 
> set.

Maybe I should explain a little further... I'm using a linux laptop in a
windows centric environment. The wifi infrastructure has been set up so
the same access points (cisco aironets 1200) manage both a public
(limited) ssid "GUEST" so guests can use it getting a private address and
a second ssid "STAFF" (hidden). The later works with WPA+Certs (in windows 
people usually install both the CA and personal certs and they are ready 
to go).

(Disclaimer: I changed the names of the essid for clarity purposes)

A iwconfig ath0 scan shows the public ones, e.g.
          Cell 02 - Address: 00:0E:38:4E:47:30
                    Mode:Master
                    Encryption key:off
                    Quality:18/94  Signal level:-77 dBm  Noise level:-95dBm
                    Mode:Master
                    ESSID:"GUEST"



ath_hal: 0.9.8.6
wlan: 0.7.3.1 BETA-802.1x
ath_pci: 0.8.5.4 BETA-802.1x
Setup queue (0) for WME_AC_BK
Setup queue (1) for WME_AC_BE
Setup queue (2) for WME_AC_VI
Setup queue (3) for WME_AC_VO
ath0: mac 4.2 phy 3.0 5ghz radio 1.7 2ghz radio 2.3
ath0: 11a rates: 6Mbps 9Mbps 12Mbps 18Mbps 24Mbps 36Mbps 48Mbps 54Mbps
ath0: 11b rates: 1Mbps 2Mbps 5.5Mbps 11Mbps
ath0: 802.11 address: 00:05:4e:40:22:0a
ath0: Atheros 5211: mem=0xc0210000, irq=9



> Can you associate to the network with a WEP key set (without 
> xsupplicant running)?

Yes, I can (automatically, as soon as I set the interface up) associate to 
the public "GUEST" one...

**However** please note that in my previous email I was associated with 
the "STAFF" one, since I set the essid _before_ bringing the interface up.


ath0      IEEE 802.11  ESSID:"GUEST"
          Mode:Managed  Frequency:2.442GHz  Access Point: 00:0E:38:4E:45:90
          Bit Rate:11Mb/s   Tx-Power:off   Sensitivity=0/3
          Retry:off   RTS thr:off   Fragment thr:off
          Encryption key:off
          Power Management:off
          Link Quality:36/94  Signal level:-59 dBm  Noise level:-95 dBm
          Rx invalid nwid:0  Rx invalid crypt:0  Rx invalid frag:0
          Tx excessive retries:0  Invalid misc:0   Missed beacon:0




> 
> You will probably have to issue something like:
> 
> iwconfig ath0 key 1234567890
> iwconfig ath0 enc on

iwconfig ath0 key 1234567890
iwconfig ath0 enc on

at this moment the assoc. with "GUEST" is lost:

ath0      IEEE 802.11  ESSID:"GUEST"
          Mode:Managed  Frequency:2.462GHz  Access Point: FF:FF:FF:FF:FF:FF
          Bit Rate:1Mb/s   Tx-Power:off   Sensitivity=0/3
          Retry:off   RTS thr:off   Fragment thr:off
          Encryption key:1234-5678-90   Security mode:open
          Power Management:off
          Link Quality:0/94  Signal level:-95 dBm  Noise level:-95 dBm
          Rx invalid nwid:0  Rx invalid crypt:0  Rx invalid frag:0
          Tx excessive retries:0  Invalid misc:0   Missed beacon:0

but it shows that the key has been set.


> In order to get encryption working properly since the MADWifi patch 
> seems to have broken manual key sets, though that combination works for 

the key is set, but as shown below, it does not associate....

I set the ESSID to "STAFF"

and it simply never associates... keeps on scanning. So I cannot start 
xsupplicant (or if I start it, it complains as expected...)
[ALL] Processing interface ath0...
[INT] All Fs for dest mac!  Resetting keys!
[INT] Successfully set WEP key [1]
[INT] Successfully set the WEP transmit key [1]
[ALL] Clock tick! authWhile=0 heldWhile=18 startWhen=18 
curState=CONNECTING

etc....
etc...

So my next test is as follows: remove modules, shutdown everything wifi 
related, modprobe ath_pci, and *without* bringing the interface up, set a 
key, set enc on and set directly the ESSID to "STAFF", bring it up (the 
main idea is to avoid associating with "GUEST")


(Interface is down)

ath0      IEEE 802.11  ESSID:"STAFF"
          Mode:Managed  Frequency:2.412GHz  Access Point: 00:00:00:00:00:00
          Bit Rate:0kb/s   Tx-Power:off   Sensitivity=0/3
          Retry:off   RTS thr:off   Fragment thr:off
          Encryption key:1234-5678-90   Security mode:open
          Power Management:off
          Link Quality:0/94  Signal level:-95 dBm  Noise level:-95 dBm
          Rx invalid nwid:0  Rx invalid crypt:0  Rx invalid frag:0
          Tx excessive retries:0  Invalid misc:0   Missed beacon:0

ifconfig ath0 up

and it keeps on scanning and never associates.




xsupplicant bugs
------------------------------------------------------------------
an extra } at the end of the config file causes xsupplicant to segfault
(lexer.l)

launching xsupplicant with no key set when the card is associated with the 
GUEST essid causes xsupplicant to segfault. I recompiled with symbols on:

Program received signal SIGSEGV, Segmentation fault.
0x0804bd54 in sendframe (thisint=0x806ba50, sendframe=0xbfffe140 "", 
sendsize=18) at os_cardif.c:591
591       if (memcmp(nomac, (char *)&thisint->userdata->dest_mac[0], 6) != 
0)
(gdb) bt
#0  0x0804bd54 in sendframe (thisint=0x806ba50, sendframe=0xbfffe140 "", 
sendsize=18) at os_cardif.c:591
#1  0x0804ef72 in eapol_execute (workint=0x806ba50) at eapol.c:215
#2  0x0804aa2a in main (argc=5, argv=0xbfffeea4) at xsup_driver.c:814









madwifi bugs
-------------------------------------------------------------------
Badness in remove_proc_entry at fs/proc/generic.c:660
Call Trace:
 [<c018dc0d>] remove_proc_entry+0x11d/0x160
 [<f9b5bed7>] ieee80211_proc_remove+0x47/0x60 [wlan]
 [<f9b549f7>] ieee80211_ifdetach+0x57/0xd0 [wlan]
 [<f9b63778>] ath_detach+0x48/0x70 [ath_pci]
 [<f9b68a84>] ath_pci_remove+0x24/0xb0 [ath_pci]
 [<c01e089b>] pci_device_remove+0x3b/0x40
 [<c023be76>] device_release_driver+0x66/0x70
 [<c023beab>] driver_detach+0x2b/0x40
 [<c023c0ed>] bus_remove_driver+0x3d/0x80
 [<c023c4f3>] driver_unregister+0x13/0x28
 [<c01e0a76>] pci_unregister_driver+0x16/0x30
 [<f9b68bd4>] exit_ath_pci+0x14/0x2d [ath_pci]
 [<c013699e>] sys_delete_module+0x13e/0x180
 [<c014eeda>] do_munmap+0x14a/0x190
 [<c033a34e>] sysenter_past_esp+0x43/0x65

ath_pci: driver unloaded


ieee80211_ifdetach(struct net_device *dev)
{
        struct ieee80211com *ic = dev->priv;

	/**** Is the device still valid at this point ?? */
....
#ifdef CONFIG_PROC_FS
        ieee80211_proc_remove(ic);
#endif

static void
ieee80211_proc_remove(struct ieee80211com *ic)
{
        if (ic->i

I really don't have the time to fix it, but I report it nevertheless...
----------------------------------------------------------------------





> 
> Also, please send me (off madwifi list, or to the xsupplicant list) the 
> following:
> 
> Your xsupplicant configuration (no passwords please).

Sure, by private mail. I've changed the password in the file, but what is 
the role of the password with certs and TLS ? (sorry, this must be a FAQ 
question).



Thanks,

Ramon