On Mon, Jun 25, 2007 at 08:39:09PM +0100, Christoph Hellwig wrote:

quoted

On Mon, Jun 25, 2007 at 12:12:21PM -0500, Olof Johansson wrote:

quoted

Driver for the CompactFlash slot on the PA Semi Electra eval board. 
It's
a simple device sitting on localbus, with interrupts and 
detect/voltage
control over GPIO.

The driver is implemented as an of_platform driver, and adds localbus
as a bus being probed by the of_platform framework.

Looks good to me. The only odd thing is the inconsistant use of
#ifdef CONFIG_PCMCIA in setup.c.

The ifdef is needed since for CONFIG_PCMCIA=n builds, the bus notifier
isn't available. I wanted to do the bus notifier registration 
explicitly
before the of_platform bus probe to avoid later surprises due to 
reordered
initcalls in case it was split up in it's own initcall.

If CONFIG_PCMCIA=m then your notifier is not registered.  The modprobe 
of your of_driver loads ds.ko, registers the bus, then registers your 
driver.  When the socket driver tries to dma, the BUG in dma_64 for no 
archdata.dma_ops triggers.

It seems like we need

(1) a notifier that a bus is registered, run before allowing any 
devices, so that platforms can register bus notifiers by bus name 
before the devices and drivers are registered.

(2) a powerpc64 generic pcmcia bus notifier that copys the dma ops from 
the parent socket.

(3) something to set the dma_ops to direct_dma_ops on the of device.

If we don't want (3) to be in the driver (as Christoph previosly 
mentioned), then it needs to be a seperate bus that reuses the of 
matching.  This would be similar to how ibmebus is setup.   If I 
remember the discussion, ibmebus is to provide the alternate dma ops 
and steals match etc code from the of_platform bus type.


Oh, is this why you have depends on PCMCIA=y ?

I could add the code under ifdef as well, but it didn't seem too
critical. Once the second major board comes along I'll probably move it
out to a per-board file, there's no real need for it just yet.

milton