Socket CF driver, the Zaurus, and monitor mode

[Pavel Roskin]

Hi, Ian!

> >   (The plan is to use the WE14 Scanning mechanism.) Also no go.  Any
> >   reference to eth0 sticks a process in D state.  (I think that was the
> >   problem.)

Obviously, the spectrum_cs driver crashes or gets stuck in the middle of 
initialization.  I'll try to figure it out.

> Upon reflection, that was the symptom of a different problem entirely
> (I had initially forgotten to install the modules built in the kernel
> rebuild, and the old usbd modules were causing a hang on references
> to usb0, not eth0.  Sorry.)  The actual symptom was that the modules
> would load, but eth0 wouldn't come into being.  Doing a
> "modprobe -r spectrum_cs" followed by a "modprobe spectrum_cs" caused
> a segfault in modprobe (but the module would load, oddly).

Segfaults in the kernel in the user context are attributed to the process
that owns the context.  Driver initialization is done in the context of
modprobe.  So it's the driver that crashes, not modprobe.  Running "dmesg"
should give you the kernel oops.

Perhaps modprobe could intercept intercept SIGSEGV and give a short 
explanation to avoid confusion.

The kernel is unstable after that.  It's better to reboot it.  Secondary
effects are hardly interesting to anyone here.

-- 
Regards,
Pavel Roskin