Re: [Xen-devel] difference between xen hypervisor and common kernel on handling BIOS's e820 map ?

[Ian Campbell <Ian.Campbell@xxxxxxxxxx>]

On Tue, 2011-08-30 at 17:10 +0100, Konrad Rzeszutek Wilk wrote:
> On Tue, Aug 30, 2011 at 08:30:32AM -0700, Lin-bao Zhang wrote:
> > 
> > 
> > >> I diff them in a picture :
> > 
> > >Huh? Can you just do a diff of the two texts or just
> > >do <=== on the email to point to it?
> > orry for lost that differing result of picture(I just paste it here ,it 
> > seems it can't work),
> 
> I am having a hard time believing you can't paste it (which you did in the 
> previous email)
> and adding this in what you pasted:
> 
>  Xen: 00000000fff80000 - 0000000100000000 (reserved)
>  Xen: 0000000100000000 - 0000000144b15000 (usable)   <==== THIS ONE?
> 
> > so , I paste this picture 
> > link:https://lh3.googleusercontent.com/-SfaWAj7_hzQ/TlyUQhgoUcI/AAAAAAAAAVk/Ax_ucBYVbD4/s800/e820%252520map.jpg
> 
> Ah, That. It is an extra 8MB phantom area that is used to setup mappings with
> other guests.

It's described as being to "balance backend allocations", which AIUI
means it's there to provide an extra 8M of (empty) PFN space for the
backend drivers to use for the page arrays where they map foreign pages.

This might have made sense in the old days when netback/blkback would
allocate there array and return the backing MFNs to Xen but in mainline
we only shadow those MFNs rather than returning them to the hypervisor,
don't we? So the slack doesn't really make sense any more.

I fully expect 8M bears no relationship to the actual size used by the
backends these days.

Ian.

> 
> > 
> > I also upload my machine's all booting messages including 
> > BIOS/grub/linuxkernel.
> 
> There is no need for all of that.
> 
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@xxxxxxxxxxxxxxxxxxx
> http://lists.xensource.com/xen-devel



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