Re: [Xen-devel] Re: [RFC Patch] Support for making an E820 PCI hole in toolstack (xl + xm)

[Ian Campbell <Ian.Campbell@xxxxxxxxxx>]

On Wed, 2010-11-17 at 11:14 +0000, Gianni Tedesco wrote:
> On Fri, 2010-11-12 at 23:08 +0000, Konrad Rzeszutek Wilk wrote:
> > Hey guys,
> > 
> > Attached is an RFC patch for making a PCI hole in the PV guests. This allows
> > PV guests(*) with 4GB or more to now properly work with or without
> > PCI passthrough cards.
> > 
> > Previously the Linux kernel would not be able to allocate the PCI region
> > underneath the 4GB region as that region was all System RAM. And you would 
> > see
> > this:
> > 
> > [    0.000000] PM: Registered nosave memory: 00000000000a0000 - 
> > 0000000000100000
> > [    0.000000] PCI: Warning: Cannot find a gap in the 32bit address range
> > [    0.000000] PCI: Unassigned devices with 32bit resource registers may 
> > break!
> > [    0.000000] Allocating PCI resources starting at 100100000 (gap: 
> > 100100000:400000)
> > 
> > 
> > This patchset punches an PCI hole in the E820 region and as well fills the 
> > P2M properly,
> > so that now you can see (*):
> > [    0.000000] Allocating PCI resources starting at a0000000 (gap: 
> > a0000000:60000000)
> > 
> > It adds a new option to guest config file, which is "pci_hole". The user can
> > specify the PFN number, such as '0xc0000' or in case of using the xl, '1' 
> > which
> > will automatically figure out the start of the PCI address.
> > 
> > *: This option requires support in the Linux kernel to actually deal with 
> > two
> > entries in the E820 map and P2M space filled with ~0.
> 
> Doesn't this problem affect dom0 also? How is it to be fixed there?

Domain 0 sees the actual host e820 map instead of the paravirtualised
one constructed by the tools for domU. So the BIOS effectively takes the
place of the tools side of this patch for dom0.

Ian.




_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel