Re: [PATCH v4 0/9] toolstack-based approach to pvhvm guest kexec

[Olaf Hering <olaf@aepfle.de>]

On Thu, Dec 11, David Vrabel wrote:

> Nothing special needs to be done with ballooned pages.  If frames are
> not populated in the original domain, they will be unpopulated in the
> new domain.
> 
> It's the responsibility of the guest to ensure it either doesn't kexec
> when it is ballooned or that the kexec kernel can handle this (e.g., by
> using a crash region that is never ballooned out).

There is a difference between kexec and kdump. The kdump kernel does not
care because there is code in /proc/vmcore to handle ballooned pages in
the crashed kernel gracefully.
But a kexec boot will likely access pages which are not backed by RAM.
Unfortunately there is no flag left to mark a page as ballooned.

So what you are saying means that kexec-tools needs to continue to
balloon up before doing the actual kexec. I had hoped this suggested
approach would get rid of that limitation.

Olaf