Re: [rfc][patch] driverfs multi-node(board) patch [2/2]

[Patrick Mochel]

Ok, I'm finally getting back to you..

> 	Ok..  here are the real changes.  I'd really like to get some feedback on 
> what you (or anyone else) thinks of these proposed changes.  This sets 
> up a generic topology initialization routine which should discover all 
> online nodes (boards), CPUs, and Memory Blocks at boot time.  It also 
> makes the CPUs and memblks it discovers children of the appropriate nodes.

You didn't append the patch, which is annoying, but I'll deal..

The main problem I have is the code placement. I put the CPU stuff under 
arch/ because I anticpate wrapping the cpu structure with an arch-specific 
one, so you can ascertain arch-specific information via the generic 
structure. Moving it out of arch/ precludes that from happening (easily). 

Ditto for memblks, though I'm not really sure what other info you'd want 
in the structures. 

Ditto+ for nodes, or boards. Those are definitely arch-specific 
structures, and shouldn't be in drivers/base/. On top of that, I don't 
think their registration should all be munged together in one file. Maybe 
they should be, in their own play area (under arch/i386/mach-ccnuma/).

	-pat