Orinoco testing + ARM (Byte-alignment problems, it seems)

[Pavel Roskin]

Hi, Ben!

> Have you done any lengthy tests...like 24hours+?  I had a .9 driver
> working good enough to last for about 24 hours before screwing up,
> I may have to revert back to that one.

No.  I tried 10000 packets, which took about 1 minute.  It's another sign 
that we are dealing with a race condition.

> > I'm getting this on ARM only with one particular card with Intersil
> > firmware (ZCOMAX AirRunner/XI-300) and only once, when the card is
> > inserted.  I haven't seen this message on i386, but I cannot test that
> > card on i386 - it has a compact flash connector that's too thick for the
> > PCMCIA adaptor.
> 
> Ugggh, I'm using a XI-800 or something like that CF card.

Actually I wrote "ZCOMAX AirRunner/XI-300" because it's the listing in
/etc/pcmcia/config for its manufacturer ID.  Actually it's an unnamed (and
unlabelled) card:

Socket 1:
  product info: " ", "IEEE 802.11 Wireless LAN/PC Card", ""
  manfid: 0xd601, 0x0002
  function: 6 (network)

> What CF cards have you had good luck with on ARM?

This one and Symbol Spectrum24.  But you will need a separate driver for
Spectrum24 - it needs firmware download every time it's inserted.  The
driver is (temporarily) on my homepage:

http://www.red-bean.com/~proski/symbol/

> > Since the problem with the driver seems to show up differently, it's more 
> > likely to be a race than an alignment problem.
> 
> I'm beginning to think maybe tcpdump has alignment issues instead of
> the driver.

Maybe.

> It is my opinion that we should make all structs passed through ioctls
> be 4-byte aligned with no implicit padding (make all padding explicit).

I'm afraid it's too late.  PCMCIA utilities and the kernel can be compiled 
separately and should always work together.

-- 
Regards,
Pavel Roskin