[RFC,v11,0/2] add support for VCPU event states

Message

Dongjiu Geng Sept. 27, 2018, 4:55 p.m. UTC

Support KVM_GET/SET_VCPU_EVENTS to get/set the SError exception state, and
support the state migration.

Now the VCPU event only includes the SError exception status, it can be
extended if needed. When do migration, If source machine has serror pending, 
the target machine is also needed to pend this serror regardless of whether
target machine can support to set the serror syndrome.

Note: Because I do not have arm32 environment, I only this patch in the KVM64,
not test it in the KVM32. So I need someone else test it in the 32 bit KVM platform.
Thanks.

Dongjiu Geng (2):
  linux-headers: Update to linux kernel 4.19-rc5
  target/arm: Add support for VCPU event states

 include/standard-headers/linux/input.h |  9 ++---
 linux-headers/asm-arm/kvm.h            | 13 ++++++++
 linux-headers/asm-arm64/kvm.h          | 13 ++++++++
 linux-headers/asm-x86/kvm.h            |  1 +
 linux-headers/linux/kvm.h              |  2 ++
 linux-headers/linux/vhost.h            |  2 +-
 target/arm/cpu.h                       |  7 ++++
 target/arm/kvm.c                       | 60 ++++++++++++++++++++++++++++++++++
 target/arm/kvm32.c                     | 13 ++++++++
 target/arm/kvm64.c                     | 13 ++++++++
 target/arm/kvm_arm.h                   | 24 ++++++++++++++
 target/arm/machine.c                   | 22 +++++++++++++
 12 files changed, 174 insertions(+), 5 deletions(-)

Comments

Peter Maydell Oct. 16, 2018, 12:57 p.m. UTC | #1

On 27 September 2018 at 17:55, Dongjiu Geng <[e-mail]> wrote:
> Support KVM_GET/SET_VCPU_EVENTS to get/set the SError exception state, and
> support the state migration.
>
> Now the VCPU event only includes the SError exception status, it can be
> extended if needed. When do migration, If source machine has serror pending,
> the target machine is also needed to pend this serror regardless of whether
> target machine can support to set the serror syndrome.
>
> Note: Because I do not have arm32 environment, I only this patch in the KVM64,
> not test it in the KVM32. So I need someone else test it in the 32 bit KVM platform.
> Thanks.

Hi; could you clarify why this patchset is tagged RFC, please?
Is it just that it hasn't been tested on 32-bit ?

thanks
-- PMM

Dongjiu Geng Oct. 18, 2018, 9:38 a.m. UTC | #2

Hi peter,

    Thanks very much for this comments and very sorry for my late response due to business trave.

     Yes, the reason that tag the patchset to RFC is that it just  hasn't been tested on 32-bit. if someone else can test it on 32-bit platform, it will be great.

    I do not have 32-bit environment, so only tested it on 64-bit.


On 27 September 2018 at 17:55, Dongjiu Geng <[e-mail]> wrote:
> Support KVM_GET/SET_VCPU_EVENTS to get/set the SError exception state, and
> support the state migration.
>
> Now the VCPU event only includes the SError exception status, it can be
> extended if needed. When do migration, If source machine has serror pending,
> the target machine is also needed to pend this serror regardless of whether
> target machine can support to set the serror syndrome.
>
> Note: Because I do not have arm32 environment, I only this patch in the KVM64,
> not test it in the KVM32. So I need someone else test it in the 32 bit KVM platform.
> Thanks.

Hi; could you clarify why this patchset is tagged RFC, please?
Is it just that it hasn't been tested on 32-bit ?

thanks
-- PMM

Peter Maydell Oct. 18, 2018, 4:31 p.m. UTC | #3

On 18 October 2018 at 10:38, gengdongjiu <[e-mail]> wrote:
> Hi peter,
>
>     Thanks very much for this comments and very sorry for my late response due to business trave.
>
>      Yes, the reason that tag the patchset to RFC is that it just  hasn't been tested on 32-bit. if someone else can test it on 32-bit platform, it will be great.
>
>     I do not have 32-bit environment, so only tested it on 64-bit.

OK, thanks. I gave it a quick 32-bit test, though I don't have
a mechanism for inserting SErrors.

Applied to target-arm.next, thanks.

-- PMM