These patches are version 3 to enalbe Intel MPX for KVM.
Version 1:
* Add some Intel MPX definiation
* Fix a cpuid(0x0d, 0) exposing bug, dynamic per XCR0 features enable/disable
* vmx and msr handle for MPX support at KVM
* enalbe MPX feature for guest
Version 2:
* remove generic MPX definiation, Qiaowei's patch has add the definiation at kernel side
* add MSR_IA32_BNDCFGS to msrs_to_save
Version 3:
* rebase on latest kernel, which include Qiaowei's MPX common definiation pulled from HPA's tree
Thanks,
Jinsong-
Il 22/01/2014 06:29, Liu, Jinsong ha scritto:
> These patches are version 3 to enalbe Intel MPX for KVM.
>
> Version 1:
> * Add some Intel MPX definiation
> * Fix a cpuid(0x0d, 0) exposing bug, dynamic per XCR0 features enable/disable
> * vmx and msr handle for MPX support at KVM
> * enalbe MPX feature for guest
>
> Version 2:
> * remove generic MPX definiation, Qiaowei's patch has add the definiation at kernel side
> * add MSR_IA32_BNDCFGS to msrs_to_save
>
> Version 3:
> * rebase on latest kernel, which include Qiaowei's MPX common definiation pulled from HPA's tree
I am afraid there is still some work to do on these patches, so they
need to be delayed to 3.15.
Patch 1:
this seems mostly separate from the rest of the MPX work. I
commented on the missing "ULL" suffix, but I would also like to
understand why you put this patch in this series.
Patch 2:
As remarked in the reply to this patch:
- the vmx_disable_intercept_for_msr has to be unconditional
- you need a new kvm_x86_ops member mpx_supported, to disable
MPX whenever the two VMX controls are not available.
Patch 3:
this patch needs to be rebased. Apart from that it is fine,
but please move the VMX bits together with patch 2, and the
other bits together with patch 4.
Patch 4:
this patch needs to be rebased and to use the new mpx_supported
member
If you also want to look at nested VMX support for MPX, that would be
nice. It should not be hard. Otherwise we can take care of that later.
Thanks for your work,
Paolo
Il 22/01/2014 13:03, Paolo Bonzini ha scritto:
> Il 22/01/2014 06:29, Liu, Jinsong ha scritto:
>> These patches are version 3 to enalbe Intel MPX for KVM.
>>
>> Version 1:
>> * Add some Intel MPX definiation
>> * Fix a cpuid(0x0d, 0) exposing bug, dynamic per XCR0 features
>> enable/disable
>> * vmx and msr handle for MPX support at KVM
>> * enalbe MPX feature for guest
>>
>> Version 2:
>> * remove generic MPX definiation, Qiaowei's patch has add the
>> definiation at kernel side
>> * add MSR_IA32_BNDCFGS to msrs_to_save
>>
>> Version 3:
>> * rebase on latest kernel, which include Qiaowei's MPX common
>> definiation pulled from HPA's tree
>
> I am afraid there is still some work to do on these patches, so they
> need to be delayed to 3.15.
>
> Patch 1:
> this seems mostly separate from the rest of the MPX work. I
> commented on the missing "ULL" suffix, but I would also like to
> understand why you put this patch in this series.
>
> Patch 2:
> As remarked in the reply to this patch:
> - the vmx_disable_intercept_for_msr has to be unconditional
> - you need a new kvm_x86_ops member mpx_supported, to disable
> MPX whenever the two VMX controls are not available.
>
> Patch 3:
> this patch needs to be rebased. Apart from that it is fine,
> but please move the VMX bits together with patch 2, and the
> other bits together with patch 4.
>
> Patch 4:
> this patch needs to be rebased and to use the new mpx_supported
> member
>
> If you also want to look at nested VMX support for MPX, that would be
> nice. It should not be hard. Otherwise we can take care of that later.
>
> Thanks for your work,
>
> Paolo
Are you going to send v4?
Paolo
Paolo Bonzini wrote:
> Il 22/01/2014 13:03, Paolo Bonzini ha scritto:
>> Il 22/01/2014 06:29, Liu, Jinsong ha scritto:
>>> These patches are version 3 to enalbe Intel MPX for KVM.
>>>
>>> Version 1:
>>> * Add some Intel MPX definiation
>>> * Fix a cpuid(0x0d, 0) exposing bug, dynamic per XCR0 features
>>> enable/disable
>>> * vmx and msr handle for MPX support at KVM
>>> * enalbe MPX feature for guest
>>>
>>> Version 2:
>>> * remove generic MPX definiation, Qiaowei's patch has add the
>>> definiation at kernel side
>>> * add MSR_IA32_BNDCFGS to msrs_to_save
>>>
>>> Version 3:
>>> * rebase on latest kernel, which include Qiaowei's MPX common
>>> definiation pulled from HPA's tree
>>
>> I am afraid there is still some work to do on these patches, so they
>> need to be delayed to 3.15.
>>
>> Patch 1:
>> this seems mostly separate from the rest of the MPX work. I
>> commented on the missing "ULL" suffix, but I would also like to
>> understand why you put this patch in this series.
>>
>> Patch 2:
>> As remarked in the reply to this patch:
>> - the vmx_disable_intercept_for_msr has to be unconditional
>> - you need a new kvm_x86_ops member mpx_supported, to disable
>> MPX whenever the two VMX controls are not available.
>>
>> Patch 3:
>> this patch needs to be rebased. Apart from that it is fine,
>> but please move the VMX bits together with patch 2, and the
>> other bits together with patch 4.
>>
>> Patch 4:
>> this patch needs to be rebased and to use the new mpx_supported
>> member
>>
>> If you also want to look at nested VMX support for MPX, that would be
>> nice. It should not be hard. Otherwise we can take care of that
>> later.
>>
>> Thanks for your work,
>>
>> Paolo
>
> Are you going to send v4?
>
> Paolo
Yes, I just return from long Chinese Spring Festival, I will send V4 later.
Thanks,
Jinsong-
Paolo Bonzini wrote:
> Il 22/01/2014 13:03, Paolo Bonzini ha scritto:
>> Il 22/01/2014 06:29, Liu, Jinsong ha scritto:
>>> These patches are version 3 to enalbe Intel MPX for KVM.
>>>
>>> Version 1:
>>> * Add some Intel MPX definiation
>>> * Fix a cpuid(0x0d, 0) exposing bug, dynamic per XCR0 features
>>> enable/disable
>>> * vmx and msr handle for MPX support at KVM
>>> * enalbe MPX feature for guest
>>>
>>> Version 2:
>>> * remove generic MPX definiation, Qiaowei's patch has add the
>>> definiation at kernel side
>>> * add MSR_IA32_BNDCFGS to msrs_to_save
>>>
>>> Version 3:
>>> * rebase on latest kernel, which include Qiaowei's MPX common
>>> definiation pulled from HPA's tree
>>
>> I am afraid there is still some work to do on these patches, so they
>> need to be delayed to 3.15.
>>
>> Patch 1:
>> this seems mostly separate from the rest of the MPX work. I
>> commented on the missing "ULL" suffix, but I would also like to
>> understand why you put this patch in this series.
>>
>> Patch 2:
>> As remarked in the reply to this patch:
>> - the vmx_disable_intercept_for_msr has to be unconditional
>> - you need a new kvm_x86_ops member mpx_supported, to disable
>> MPX whenever the two VMX controls are not available.
>>
>> Patch 3:
>> this patch needs to be rebased. Apart from that it is fine,
>> but please move the VMX bits together with patch 2, and the
>> other bits together with patch 4.
>>
>> Patch 4:
>> this patch needs to be rebased and to use the new mpx_supported
>> member
>>
>> If you also want to look at nested VMX support for MPX, that would be
>> nice. It should not be hard. Otherwise we can take care of that
>> later.
>>
>> Thanks for your work,
>>
>> Paolo
>
> Are you going to send v4?
>
> Paolo
Yes, I have just sent out v4 patches per your comments.
Thanks,
Jinsong-