Re: [Xen-devel] [RFC v3 5/6] xen/arm: Add log_dirty support for ARM

[Ian Campbell <Ian.Campbell@xxxxxxxxxx>]

On Mon, 2014-05-12 at 15:11 +0100, Julien Grall wrote:
> On 05/12/2014 03:00 PM, Wei Huang wrote:
> > On 05/11/2014 10:28 AM, Julien Grall wrote:
> >>   the guest very soon. With this
> >> solution there is a 1GB hole between the 2 banks. Your function will
> >> therefore
> >> stop to work.
> >>
> >> Furthermore, Xen should not assume that the layout of the guest will
> >> always start
> >> at GUEST_RAM_BASE.
> >>
> >> I think you can use max_mapped_pfn and lowest_mapped_pfn here. You may
> >> need to
> >> modify a bit the signification of it in the p2m code or introduce a
> >> new field.
> > These two values don't work for the purpose of dirty tracking. What we
> > need are two fields to track _physical_ memory. lowest_mapped_pfn tracks
> > all memory types, which doesn't apply here.
> > 
> > Will new fields to track physical ram space acceptable for you?
> 
> I'm fine with new fields to track physical RAM space. Don't forget it
> may have non-RAM hole in this range.

FWIW x86 appears to use a rangeset for this purpose, and keeps it up to
date when modifying the p2m. Doesn't seem like a bad plan to me...

Ian.


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel