Re: [patch 2/5] KVM: hypercall based pte updates and TLB flushes

[Marcelo Tosatti <marcelo@kvack.org>]

On Sun, Feb 17, 2008 at 03:13:47PM +0200, Avi Kivity wrote:
> Avi Kivity wrote:
> >Marcelo Tosatti wrote:
> >>Hypercall based pte updates are faster than faults, and also allow use
> >>of the lazy MMU mode to batch operations.
> >>
> >>Don't report the feature if two dimensional paging is enabled.
> >>
> >>Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
> >>+/*
> >>+ * We only need to hook operations that are MMU writes.  We hook 
> >>these so that
> >>+ * we can use lazy MMU mode to batch these operations.  We could 
> >>probably
> >>+ * improve the performance of the host code if we used some of the 
> >>information
> >>+ * here to simplify processing of batched writes.
> >>+ */
> >>  
> >
> >One option is, if the guest promises never to write to a page table 
> >directly, is to avoid write protecting guest page tables.  I think the 
> >shadow code can handle it (since the gfn/spte relationship is 
> >maintained by shadow code, and doesn't require reading the guest page 
> >tables), but am not sure.
> >
> 
> In addition to reducing mmu work for write protection, this allows more 
> efficient use of large pages.

Yes, and gets rid of the remote TLB flushing.

Issue is the paravirt_ops code in Linux does not cover all pte updates
(bit updates, ptep_get_and_clear, etc).

The plan is to get the basic infrastructure merged into KVM first (which
is a significant improvement already) and then later have paravirt_ops
cover all updates, disabling write protection.

-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft
Defy all challenges. Microsoft(R) Visual Studio 2008.
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/