2.6.x kernels & PCI-to-PCMCIA bridges [again]

[Pavel Roskin]

On Wed, 5 May 2004, jar wrote:

> I have already do that. The PD6729 driver _is_ _working_ very well and
> it should be added to the kernel sources I think.
>
> Only issue I have is the pcmcia system restarting. I get couple of times
>  "cs: unable to map card memory!" messages. This is most probably
> because  the rmmod can't unload the pd6729 module, even if the driver
> for the pcmcia card and cardmanager is unloaded. This is not happen with
> i82365 driver. Maybe this is just cosmetic.

It may be caused by the problem I described recently:
http://lists.infradead.org/pipermail/linux-pcmcia/2004-April/000771.html

Essentially, pcmcia_core uses static variables where it shouldn't.

> When I try to add "/sbin/cardctl eject" command just before rmmods
> (rmmod ds, rmmod pd6729, rmmod pcmcia_core) to free the pd6729 module,
> kernel panics after cardmgr exiting.
>
> Can you run /sbin/cardctl eject successfully?
> Can you unload all those modules?

I haven't seen any panics in the latest 2.6 kernels (at least in 2.6.5).
It may be a problem specific to pd6729.

Do you kill cardmgr?  If I do "cardctl eject", I can remove the socket
driver (yenta_socket in my case) without killing cardmgr, which doesn't
seem right.  It confuses the kernel, but doesn't cause kernel panic for
me.  If I restart cardmgr after that it fails to adjust resources.

You may want to describe the exact sequence of commands and post the stack
trace.

-- 
Regards,
Pavel Roskin