[PATCH 0/5] live migration dirty bitmap support for ARMv7

[PATCH 0/5] live migration dirty bitmap support for ARMv7

[Mario Smarduch]

Revised iteration after initial comments. Still just for ARMv7. I looked
at the ARMv8 code and yes it practically appears to reuse most of 
fault handling in ARMv7, I wasn't aware so much code was in common. 
But before then want to make sure it's reliable on real hardware.

This patch adds support for ARMv7 Live Migration, primarily dirty bit map
management is added. The patch follows the normal migration flow managed by
user space, first write protecting the entire address space and later 
keeping track of dirty pages. In the process of initial write protection,
and first time faults huge pages are broken up into small pages to support
migration on loaded systems.

Mario Smarduch (5):
  add ARMv7 HYP API to flush VM TLBs without address param
  live migration support for initial write protect of VM to manage
    dirty pages
  live migration support for VM dirty log management
  add 2nd stage page fault handling during live migration
  add kvm_arch glogal live migration variable

 arch/arm/include/asm/kvm_asm.h  |    1 +
 arch/arm/include/asm/kvm_host.h |    7 ++
 arch/arm/kvm/arm.c              |   75 ++++++++++++-
 arch/arm/kvm/interrupts.S       |    5 +
 arch/arm/kvm/mmu.c              |  220 ++++++++++++++++++++++++++++++++++++++-
 5 files changed, 305 insertions(+), 3 deletions(-)

-- 
1.7.9.5

Re: [PATCH 0/5] live migration dirty bitmap support for ARMv7

[Marc Zyngier]

Hi Mario,

On Thu, Apr 17 2014 at  2:32:22 am BST, Mario Smarduch <m.smarduch@samsung.com> wrote:
> Revised iteration after initial comments. Still just for ARMv7. I looked
> at the ARMv8 code and yes it practically appears to reuse most of 
> fault handling in ARMv7, I wasn't aware so much code was in common. 
> But before then want to make sure it's reliable on real hardware.
>
> This patch adds support for ARMv7 Live Migration, primarily dirty bit map
> management is added. The patch follows the normal migration flow managed by
> user space, first write protecting the entire address space and later 
> keeping track of dirty pages. In the process of initial write protection,
> and first time faults huge pages are broken up into small pages to support
> migration on loaded systems.

Thanks for respining this. A few words about the patch submission
process though:

- Please add a version number (v2, v3...) to your patch series. It helps
the reviewers keeping track of where we are in the review process

- Add a change log to your cover letter. Clearly state what
changed. Otherwise, reviewers have to start diffing between patches,
which becomes quickly messy, specially considering the lack of
versioning.

- Make sure your patches are part of an email thread starting with your
cover letter. This helps reviewers to locate the various patches in a
busy Inbox, and to follow the evolution of a discussion. Use of "git
send-email" is recommended for posting the patch series.

That being said, I'm off to review the actual patches... ;-)

Thanks,

	M.

> Mario Smarduch (5):
>   add ARMv7 HYP API to flush VM TLBs without address param
>   live migration support for initial write protect of VM to manage
>     dirty pages
>   live migration support for VM dirty log management
>   add 2nd stage page fault handling during live migration
>   add kvm_arch glogal live migration variable
>
>  arch/arm/include/asm/kvm_asm.h  |    1 +
>  arch/arm/include/asm/kvm_host.h |    7 ++
>  arch/arm/kvm/arm.c              |   75 ++++++++++++-
>  arch/arm/kvm/interrupts.S       |    5 +
>  arch/arm/kvm/mmu.c              |  220 ++++++++++++++++++++++++++++++++++++++-
>  5 files changed, 305 insertions(+), 3 deletions(-)

-- 
Jazz is not dead. It just smells funny.