On Wed, Aug 03, 2005 at 12:50:04PM +0200, Manuel Bouyer wrote:
> On Tue, Jul 26, 2005 at 01:16:24AM +0000, Wojciech A. Koszek wrote:
> Hi,
> sorry for not replying sooner, I was on vacation.
> From what I remember, we needed (all this relative to the plain i386 port):
> - change the pci_conf_* methods for PCI registers to go though hypervisor
>   calls instead of direct registers access (xen/pci_machdep.c)
> - change the PCI bus detection method to use the informations provided by
>   the hypervisor (xen/hypervisor.c)
> - change the interrupt register routine for PCI and ISA devices to use
>   the appropriate hypervisor event (xen/pci_machdep.c and xen/isa_machdep.c)
> - change the bus_dma(9) methods for deal with physical vs machine addresses
>   (include/bus_private.h + tweaks to arch/x86/x86/bus_dma.c)
> 
> With this, a domain0 kernel should be able to probe and use the hardware.
> Next, you have to write the xbd and network backends. In NetBSD the code is
> in xen/xbdback.c and xen/xennetback.c. You also need to write kernel support
> for domain0 operations though 2 special files:
> /kern/xen/privcmd (xen/privcmd.c)
> /dev/xenevt (xen/xenevt.c)
> then you need to port the xentools to FreeBSD.
> 

Thanks! This response should clarify everything.

Regards,
-- 
* Wojciech A. Koszek && dunstan@FreeBSD.czest.pl