Belkin Wireless G Plus & Compaq v2417au laptop

[Alex Kavanagh]

At Thu, 20 Apr 2006 19:30:52 +0100,
Christian Bullow wrote:
> 
> [1  <multipart/mixed (7bit)>]
> [1.1  <text/plain; ISO-8859-1 (7bit)>]
> -----BEGIN PGP SIGNED MESSAGE-----

(I must get my PGP setup working again...)

> Hash: SHA1
>  
> 
> > No, the patch to 2.6.16 actually means that you don't need
> > pci=assign-busses as a kernel parameter.  Basically, what I think is
> > going on is that the yenta_socket module isn't correctly picking up
> > the right bus number because of the particular hardware that is
> > installed in our laptops.  I think that HP have done something
> > 'different' (i.e. not previously seen by the Kernel developers) with
> > the TI PCIxx21 cardbus bridge hardware interfacing which was causing
> > yenta_socket and the PCI bridge software to not be able to find the
> > bridge properly.  The very recent patch, mentioned above, fixes this
> > problem and allows the processor to actually access the cardbus card
> > correctly.
> >
> Is this patch in 2.6.16? when will we see it hit the mainstream you
> reckon?

Good question.  It should make it into the mm kernel fairly soon (if
not already).  Otherwise it will be in 2.6.17 whenever that comes off.

> 
> > However, I *think* (but don't know) that there is an additional
> > problem with the TI PCIxx21 interface not being set up properly in our
> > laptops which is why our cards aren't working.
> Fun Fun. I noted that there is some, though limited, mention of this
> particular cardbus controller on the net which mentions issues where
> people have said it is built differently.

Really?  Have you got any references?  It might just help me with the
problems that we are having.

> >
> > Thus, this is probably a step in the right direction, but maybe not
> > the whole solution yet.
> >
> We will see. I'm using the standard distributions kernel, was hopeing
> to avoid compiling my own.

Ah, you may be out of luck there.

> > To answer your problem, you could try rebooting linux with the
> > pci=assign-busses parameter and see if it helps.  If not, then you
> > could try to download the 2.6.16.1 kernel, add the patch mentioned in
> > the previous email, install the new kernel and try that.
> I have downloaded 2.6.16.9 which is listed as most recent stable

Yes, I noticed that it had gone to that.  You will still probably have
to add the patch manually though - it's not too hard to do as it is
small patch.  But now you have the source you can check in
drivers/pcmcia/yenta_socket.c and search for the 'fixup' patch and see
if it is in there.

> >
> > If that sounds too daunting, then I would wait a few days/weeks until
> > the problem can be sorted out.  I'm going to keep looking at
> > datasheets and try to work out what might be different on these
> > laptops.
> 
> What are datasheets?

Unless you are being ironic (it's hard to tell with email), a
datasheet is a document that specifies a piece of hardware;  i.e. it
tells you about the registers, how to program them, electrical
characteristics - basically it is a developer manual for a piece of
hardware (I'm an electronics engineer ...)

> >
> > Incidentally, my card (Novatel Merlin G3 U530) has to have a special
> > driver from Novatel for Windows because Windows, as standard, fails to
> > detect the device as a multifunction device.  The additional fix from
> > Novatel 'fixes' Windows by getting it to detect the additional
> > function and installing a driver for it.  Not sure if this sheds any
> > light on the issue.
> >
> 
> hmm, am I going to have this issue with WiFi cards?

Maybe, maybe not.  Discussions with Bernhard indicate that it might
just be a serial_cs/serial issue that I'm having and that the wifi
driver may be perfectly happy once the bus has been allocated
properly.  On the other hand, if it is a pcixx21 setup problem then
the same problem could be exhibited.  Only one way to find out for
sure ... you have to try it.

Cheers
Alex.