RE: ohci laptop ?

[Mark K. <mk...@co...>]

I haven't looked into this in awhile, but a couple of years ago it was my
understanding that most of the machines that implement cardbus are doing
this behind what is essentially a PCI-PCI bridge. Could this be an issue
here?

Some of the early designs had trouble doing DMA behind these bridges, or at
least the performance they could get was marginal at that time. Hopefully
things have improved.



-----Original Message-----
From: Ben Collins [mailto:bco...@de...]
Sent: Friday, June 02, 2000 7:09 AM
To: Olivier Daigle
Cc: Seb...@sy...; Linux1394-devel
Subject: Re: ohci laptop ?


On Fri, Jun 02, 2000 at 09:54:42AM -0400, Olivier Daigle wrote:
> Sebastien,
>   I'm currently working on a project that will have a laptop linked to
> other devices (including other computers) by FireWire. We bought a Ratoc
> IEEE1394 CardBus card (CBFW1 model) that is supposed to be OHCI 1.0
> compliant (compatible with built-in Winslows98 drivers). I'm waiting for
> some specs about this card. I tried to load standard ohci drivers to
> make it work but, as I expected, it didn't work. I think that I will
> have to (at least) rewrite the probing part of the driver to make it
> compatible with CardBus interface. I should begin reading on the CardBus
> standard (and the programming interface of the Linux PCMCIA Card
> Service) by next week.
> 
>   Stay tuned for more info. If anybody in there have any experience with
> any type of CardBus devices, please raise your hand.

I have the same card from Ratoc and it works[1] with the ohci1394 driver.
However, you have to reload the driver whenever you insert the card so it
will detect it.

I have a patch to avoid this by using the PCI API of the 2.3/2.4 kernels,
but it isn't very backward compatible with 2.2.x kernels, so it most
likely wouldn't be put into the main source.

For PCI interface for 2.3/2.4 is completey compatible with CardBus cards.
One driver for all should work, with autoprobing initiated by he kernel
against a PCI vendor/make masks. Thing that is critical is that it means
the driver/core wont initiate the probe. That means the core needs to be
reworked a bit to allow the driver to handle the probe and register single
cards with the core, or the driver will have to register the core to
handle it's PCI probes. Either way, the change wont allow things to work
with 2.2.x (atleast not in a way that the code looks clean).

Ben

[1] Atleast it detected the card and raw1394 testlibraw was able to probe
it for testing.

-- 
 -----------=======-=-======-=========-----------=====------------=-=------
/  Ben Collins  --  ...on that fantastic voyage...  --  Debian GNU/Linux   \
`  bco...@de...  --  bco...@op...  --  bco...@li...  '
 `---=========------=======-------------=-=-----=-===-======-------=--=---'

_______________________________________________
mailing list lin...@li...
http://lists.sourceforge.net/mailman/listinfo/linux1394-devel