Re: [patch 2.6.30-rc3] platform_bus: remove "which platform_data?" confusion

[David Brownell <david-b@pacbell.net>]

On Tuesday 28 April 2009, Greg KH wrote:
> On Tue, Apr 28, 2009 at 02:28:07AM -0700, David Brownell wrote:
> > No comment on the bugfix part of $SUBJECT patch?
> 
> Well, no, I'm assuming it is correct :)
> 
> Should I just revert the original change, if the fact that busses are
> using the platform_data field?

That would be my inclination.


> > On Monday 27 April 2009, Greg KH wrote:
> > > 
> > > > Those patches seem to support what I think is a misguided
> > > > notion:  that somehow device.platform_data might move into
> > > > the platform_device.  The problem with that idea is that it's
> > > > a general purpose hook, and is used by other busses to provide
> > > > board-specific configuration data ... not just for platform_bus.
> > > 
> > > It is?  What other busses do this?
> > 
> > SPI and I2C come quickly to mind...
> > 
> > Basically, *any* bus that could ever be used on an embedded
> > system may need platform_data to explain how each discrete
> > chip has been wired up on that particular board.  Very few
> > such busses can self-enumerate like PCI or USB.  And most of
> > the chips sitting on such busses expect to interface to fairly
> > random external hardware.
> > 
> > And come to think of it, I've seen cases with PCI and USB
> > where board-specific config data is needed.  PCI doesn't
> > always wrap it up in some ACPI bytecode, and sometimes USB
> > devices use "transceiverless link" hookup, so the board
> > can just hook up using a differential pair.
> > 
> > SDIO/MMC doesn't tend to need it though, even for SDIO
> > WLAN or MMC/SD storage links (eMMC, CE-ATA, etc).
> > 
> > 
> > > And why, can't they use their own bus private data pointers?
> > 
> > ENOPATCH.  ;)
> > 
> > Though ... since devices on *any* bus may need this, I
> > don't much see the point of modifying every bus like that.
> 
> Fair enough, no objection from me.
> 
> thanks,
> 
> greg k-h
> 
>