[Qemu-devel] PING: [RFC PATCH 0/4] GICv3 live migration support

[Qemu-devel] PING: [RFC PATCH 0/4] GICv3 live migration support

[Pavel Fedin]

 Knock-knock!

 PM: I remember we had a talk that we should settle down on migration data format. Isn't it right
time?

Kind regards,
Pavel Fedin
Expert Engineer
Samsung Electronics Research center Russia


> -----Original Message-----
> From: qemu-devel-bounces+p.fedin=samsung.com@nongnu.org [mailto:qemu-devel-
> bounces+p.fedin=samsung.com@nongnu.org] On Behalf Of Pavel Fedin
> Sent: Wednesday, September 30, 2015 5:00 PM
> To: qemu-devel@nongnu.org
> Cc: Peter Maydell; Juan Quintela; Shlomo Pongratz; Shlomo Pongratz; Amit Shah; Diana Craciun
> Subject: [Qemu-devel] [RFC PATCH 0/4] GICv3 live migration support
> 
> This series introduces support for GICv3 live migration. It is based on
> kernel API which is not released yet, therefore i post it as an RFC.
> 
> Kernel patches which implement this functionality are:
> - [PATCH v4 0/7] KVM: arm64: Implement API for vGICv3 live migration
>   http://www.spinics.net/lists/kvm/msg121588.html
> 
> The main purpose of this RFC is to agree on GICv3 state data format,
> because software implementation of GICv3 is also going to use it. In order
> to simplify GICv3 software emulation development, part 1 of this patchset
> can be accepted right now, without waiting for the kernel part.
> 
> The second question which should be addressed is how to correctly describe
> bitfields in vmstate. Bitfields are used by this code in order to reflect
> per-CPU interrupt status. qemu defines bitfields as arrays of 'long',
> therefore element length can be different on different systems. Our vmstate
> macros support only types with fixed size, like UINT64 and UINT32. In order
> to work around this problem, i relied on __SIZEOF_LONG__ definition.
> However, i suppose, it is gcc-specific and this approach is wrong for
> mainstream, therefore i'd like to discuss how this could be done better.
> Since 'long' maps to something, i think that adding a specific code for it
> would be too much anyway. May be add configure test for sizeof(long) ?
> 
> Pavel Fedin (4):
>   hw/intc/arm_gicv3_common: Add state information
>   kernel: Add definitions for GICv3 attributes
>   hw/intc/arm_gicv3_kvm: Implement get/put functions
>   hw/intc/arm_gicv3_common: Add vmstate descriptors
> 
>  hw/intc/arm_gicv3_common.c         | 199 ++++++++++++++++++-
>  hw/intc/arm_gicv3_kvm.c            | 391 ++++++++++++++++++++++++++++++++++++-
>  hw/intc/gicv3_internal.h           | 152 ++++++++++++++
>  include/hw/intc/arm_gicv3_common.h |  76 +++++++
>  linux-headers/asm-arm64/kvm.h      |  10 +-
>  5 files changed, 821 insertions(+), 7 deletions(-)
>  create mode 100644 hw/intc/gicv3_internal.h
> 
> --
> 2.4.4

Re: [Qemu-devel] PING: [RFC PATCH 0/4] GICv3 live migration support

[Peter Maydell]

On 7 October 2015 at 08:57, Pavel Fedin <p.fedin@samsung.com> wrote:
>  Knock-knock!
>
>  PM: I remember we had a talk that we should settle down on migration data format. Isn't it right
> time?

I think Christoffer has a patchset which specifies the
userspace API for KVM for this, which is probably the
right place to start. Christoffer, did you manage to send
that out yet?

-- PMM

Re: [Qemu-devel] PING: [RFC PATCH 0/4] GICv3 live migration support

[Pavel Fedin]

> I think Christoffer has a patchset which specifies the
> userspace API for KVM for this, which is probably the
> right place to start.

 Christoffer? I haven't seen anything from him, BTW, i also posted already 4 versions of this...
http://www.spinics.net/lists/kvm-arm/msg16219.html
http://www.spinics.net/lists/kvm/msg120483.html
http://www.spinics.net/lists/kvm-arm/msg16351.html
http://www.spinics.net/lists/kvm/msg121588.html

Kind regards,
Pavel Fedin
Expert Engineer
Samsung Electronics Research center Russia

Re: [Qemu-devel] PING: [RFC PATCH 0/4] GICv3 live migration support

[Christoffer Dall]

On Wed, Oct 07, 2015 at 09:02:47AM +0100, Peter Maydell wrote:
> On 7 October 2015 at 08:57, Pavel Fedin <p.fedin@samsung.com> wrote:
> >  Knock-knock!
> >
> >  PM: I remember we had a talk that we should settle down on migration data format. Isn't it right
> > time?
> 
> I think Christoffer has a patchset which specifies the
> userspace API for KVM for this, which is probably the
> right place to start. Christoffer, did you manage to send
> that out yet?
> 
Very soon now.

-Christoffer