Re: [PATCH 2/5] kvm/svm: enhance MOV CR intercept handler

[Avi Kivity <avi@redhat.com>]

On 12/07/2010 04:30 PM, Andre Przywara wrote:
> Avi Kivity wrote:
>> On 12/07/2010 12:59 PM, Andre Przywara wrote:
>>> Newer SVM implementations provide the GPR number in the VMCB, so
>>> that the emulation path is no longer necesarry to handle CR
>>> register access intercepts. Implement the handling in svm.c and
>>> use it when the info is provided.
>>>
>>> Signed-off-by: Andre Przywara<andre.przywara@amd.com>
>>> ---
>>>   arch/x86/include/asm/svm.h |    2 +
>>>   arch/x86/kvm/svm.c         |   74 
>>> +++++++++++++++++++++++++++++++++++++++-----
>>>   2 files changed, 68 insertions(+), 8 deletions(-)
>>>
>>> diff --git a/arch/x86/include/asm/svm.h b/arch/x86/include/asm/svm.h
>>> index 11dbca7..589fc25 100644
>>> --- a/arch/x86/include/asm/svm.h
>>> +++ b/arch/x86/include/asm/svm.h
>>> ...
>>> @@ -2864,14 +2922,14 @@ static int pause_interception(struct 
>>> vcpu_svm *svm)
>>>   }
>>>
>>>   static int (*svm_exit_handlers[])(struct vcpu_svm *svm) = {
>>> -    [SVM_EXIT_READ_CR0]            = emulate_on_interception,
>>> -    [SVM_EXIT_READ_CR3]            = emulate_on_interception,
>>> -    [SVM_EXIT_READ_CR4]            = emulate_on_interception,
>>> -    [SVM_EXIT_READ_CR8]            = emulate_on_interception,
>>> +    [SVM_EXIT_READ_CR0]            = cr_interception,
>>> +    [SVM_EXIT_READ_CR3]            = cr_interception,
>>> +    [SVM_EXIT_READ_CR4]            = cr_interception,
>>> +    [SVM_EXIT_READ_CR8]            = cr_interception,
>>>       [SVM_EXIT_CR0_SEL_WRITE]        = emulate_on_interception,
>>>       [SVM_EXIT_WRITE_CR0]            = cr0_write_interception,
>>> -    [SVM_EXIT_WRITE_CR3]            = emulate_on_interception,
>>> -    [SVM_EXIT_WRITE_CR4]            = emulate_on_interception,
>>> +    [SVM_EXIT_WRITE_CR3]            = cr_interception,
>>> +    [SVM_EXIT_WRITE_CR4]            = cr_interception,
>>>       [SVM_EXIT_WRITE_CR8]            = cr8_write_interception,
>>
>> We could move cr[08]_write_interception into cr_interception, but 
>> that takes a bit more thought.  Best done later.
> Yes, I thought about that, too. But since we still have to deal with 
> the emulation code path, it would not make the code easier. But if we 
> overwrite the svm_exit_handlers[] on detecting the SVM feature, this 
> could make more sense. I will check this out.
>

In fact the correct thing is to make sure kvm_set_cr0() and 
kvm_set_cr8() do the right thing, so that the emulation path invoked 
directly (not through a cr intercept) works.

-- 
error compiling committee.c: too many arguments to function