ioread/write problem

[Holger Schurig]

> So I'm sure this quite the right forum to ask this, but maybe
> someone here has some ideas which might help me.  In bringing
> up my mv8385 CF module, I've had to update the ioread/write
> functions in if_cs.c to include a udelay(1) between ALL
> ioread/write/rep calls.

The right forum would be the mailing-list for your processor's 
architecture. E.g. if you have an ARM bases processor (this 
includes Intel XScale, Marvell XScale, some Samsungs etc), go to 
the mailing list at http://www.arm.linux.org.uk/mailinglists/

The udelay(1) is ugly, probably a WMB() (write memory barrier) is 
better.

It might even be a hardware problem --- or a setup problem of the 
memory controller. E.g. on Intel PXA 2xx, which I happen to 
know, you have various extra registers that tune delays and pin 
functions for the PCMCIA interface. Those registers are 
different to the normal memory-mapped I/O devices.