On Thu, Feb 22, 2007 at 12:37:58PM +0900, Isaku Yamahata wrote:
> On Sat, Feb 17, 2007 at 09:47:53AM +0900, Akio Takebe wrote:
> Content-Description: Mail message body
> > paravirtualize mmap hundler of /proc/bus/pci
> >
> > Signed-off-by: Akio Takebe <takebe_akio@xxxxxxxxxxxxxx>
> >
> > Best Regards,
> >
> > Akio Takebe
>
> Hi Akio.
> It seems reasonable to use X as short term solution.
> However this is hacky work around of each device drivers.
> Such a work around would be necessary for each device drivers which
> implements mmap. It is impossible to paravirtualize all of them.
>
> add_window(), which registers io region and memory region,
> calls ioremap with ACPI_IO_RANGE regions, but it doesn't
> with ACPI_MEMORY_RANGE regions.
> If the ioremap hypercall is called with ACPI_MEMORY_RANGE regions,
> all pci devices are free from such paravirtualization.
> Especially we can get rid of /dev/mem and pci mmap
> paravirtualization to use X.
Would be nice, but isn't the ACPI_MEMORY_RANGE very large on some machine ?
> Presumably we don't want to map pci bridge which may be large
> so that we need to detect it somehow in add_window().
Pci root bridge or pci-to-pci bridges ?
Tristan.
_______________________________________________
Xen-ia64-devel mailing list
Xen-ia64-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-ia64-devel
|
Home