SCIENTIFIC-LINUX-USERS Archives

January 2006

SCIENTIFIC-LINUX-USERS@LISTSERV.FNAL.GOV

Options: Use Monospaced Font
Show Text Part by Default
Show All Mail Headers

Message: [<< First] [< Prev] [Next >] [Last >>]
Topic: [<< First] [< Prev] [Next >] [Last >>]
Author: [<< First] [< Prev] [Next >] [Last >>]

Print Reply
Subject:
From:
"Alan J. Flavell" <[log in to unmask]>
Reply To:
Alan J. Flavell
Date:
Mon, 23 Jan 2006 11:36:09 +0000
Content-Type:
TEXT/PLAIN
Parts/Attachments:
TEXT/PLAIN (95 lines)
Background:-

I've written here before about ndiswrapper, which I've deployed 
successfully to support several different wireless cards under SL3.  
Primarily, in our group I've deployed an inexpensive card based on the 
rt2500 chipset, which works well on Windows, as well as in SL3 under 
ndiswrapper.

However, under SL4 there is a standoff situation: the available 
kernels are built with the 4KSTACKS option, but, as warned in the 
ndiswrapper notes, quite a few of the Win32 drivers will crash under 
these conditions; ndiswrapper recommends rebuilding the kernel without 
the 4KSTACKS option, but this seems like overkill just to get a 
wireless card to work, and would turn into a maintenance nightmare for 
the users.

Consequently I've also been keeping an eye on native linux drivers for 
the rt2500 chipset.  This note records some initial success.


Native rt2500 drivers:-

The chipset vendor offers source code from their support page
at http://www.ralinktech.com/supp-1.htm - unfortunately I've not
been successful in building any of those under SL.

The rt2x00 (originally "rt2400") sourceforge project offers several 
source code packages developed from ralinktech released source.  See: 
http://rt2x00.serialmonkey.com/wiki/index.php/Downloads

Looking at rt2x00-v2.0.0-b3, it said it was for kernel versions
of at least 2.6.13: attempting to build it for SL4 was not successful.

Looking at rt2500-v1.1.0-b3 on the other hand, I *have* been 
successful in building this, both for SL3.0.3 and for SL4 (4.2 to be 
exact).  Both of the systems on which it had been tried had previously 
been set up to use ndiswrapper, but I commented that out in 
modules.conf (resp. modprobe.conf).

In SL4, it appeared to me to be necessary to run "depmod" manually, 
after installing the kernel module.

One thing I haven't yet understood is that the kernel module, although 
claiming to support a device name of wlan0 when built for Fedora Core, 
or ra0 for everything else, in fact seems to come up "all by itself" 
as eth1 when I try it on SL ("eth0" being the wire Ethernet).  I 
haven't yet understood how/why it does this without having to be given 
an explict "options" setting.

Building the GUI "utility" RaConfig2500 (in the quaintly-named 
"Utilitys/" subdirectory) requires "qt" and "qt-devel" to be 
installed.  In my case, I had to do "yum install qt-devel", and 
afterwards open a fresh shell window, before following the recipe for 
building the GUI utility.

When executed without modification, the utility consistently alerts
that the device driver could not be found, and then exits.

I modified the source code for raconfig.cpp to change the hard-coded 
reference to "ra%d" to be "eth%d", and rebuilt it.  It then works, 
provided that the rt2500 kernel module has been previously loaded.  
It looks very similar to the configuration utility provided by 
Ralinktech in their Windows package.

I'm successfully using it with WPA-PSK security, at least from "root".  
Note that, unlike ndiswrapper, this does not require the 
wpa_supplicant package - the WPA-PSK support is somehow built-in.

The RaConfig2500 GUI utility is not the only way to configure the 
setup. For example, it says it honours a configuration file, or can be 
configured via iwconfig/iwpriv low-level commands.

In fact, the GUI utility contains a hard-coded test for uid zero, so 
the utility can only be successfully executed by "root".  We would 
want to be able to run the wireless card from an unprivileged user 
account, so I'd be looking at one of the other configuration methods 
(that's on my TODO list).

There seems to be a bit of incompatibility between the available 
network configuration features in the standard SL setup (gnome network 
control panel, hot-plugging support etc.), and the features which come 
with this source-code package.  I haven't entirely got these under 
control yet.

Hope that's useful to somebody.  If anyone can fill in some of the 
open questions above, please go ahead.


We nevertheless have a few users whose laptops have built-in wireless 
facilities for which we haven't found a native driver compatible with 
SL.  So there is still a "market" for ndiswrapper, if it ever can be 
got working safely in SL4.  But I don't think I can do anything about 
that personally - I'm doing no more than keeping a watching brief on 
developments, in the hope that something emerges.

ATOM RSS1 RSS2