Upgrade to new kernel stops wpa

Mårten Thomasson marten.thomasson
Mon Oct 3 12:15:40 PDT 2005

Yeah, I knew about the change to pcmcia support, that was actually the
reason why I changed to 2.6.13, because I had massive problems with kismet
and airodump freezing after a couple of packets captured, when run on my
pcmcia atheros card. I seemed to have stumbled on the solution for that, it
seems it was network manager pulling the card out from monitor mode, so when
I killed network manager, everything seemed ok. Haha, that came out weird

Anyway, I tried modprobe -r ipw, tearing the network down and restarting it,
but no change. My card is internal in this case, intel 2200BG.

I have no problem with the yenta socket, I can get my atheros card working
(pcmcia). It?s a nightmare with wpa, I cannot even compile the software with
support for my atheros card, but that?s another story.

Regards and thanks for the help,



-----Original Message-----
From: hostap-bounces+marten.thomasson=bredband.net at shmoo.com
[mailto:hostap-bounces+marten.thomasson=bredband.net at shmoo.com] On Behalf Of
David P. Reed
Sent: 03 October 2005 16:14
To: hostap at shmoo.com
Subject: RE: Upgrade to new kernel stops wpa

I just had a bunch of problems with the new kernel update in FC4, too.   
Not the one you mention because I'm still waiting for ordinary WEP to work
(no one has responded to my earlier messages about that).

One thing that has changed a lot in 2.6.13 is PCMCIA card support (in 
particular how PCMCIA cards are set up early in the boot process).   
It's very likely that your problem is not with wpa, but with the actual
driver working at all, if you are using a PCMCIA card NIC (mine is internal,
on the USB bus, not PCMCIA).

In my case, since I use an AMD64 laptop computer, the problem is that I get
lots of error messages early in boot and also during the cardmgr setup
because the kernel apparently thinks the Yenta PCMCIA adapter is a 64-bit
PCI interface in the the initial PCI setup process and tries to 
map too much memory.   It hasn't hurt me too much, but makes me 
suspicious of the 2.6.13 release testing process (either at Fedora or
kernel.org, don't know which) because the problem was reported earlier in
the summer by several people and is mentioned on the bugzilla.

One thing I'd try is pulling all of the automagic support out of the boot
process and doing it manually, step-by-step starting with modprobe'ing the

HostAP mailing list
HostAP at shmoo.com

More information about the Hostap mailing list