Re: [PATCH 13/15] KVM: SVM: Add and use svm_register_cache_reset()

[Lai Jiangshan <laijs@linux.alibaba.com>]

On 2021/11/18 23:37, Paolo Bonzini wrote:
> On 11/8/21 13:44, Lai Jiangshan wrote:
>> From: Lai Jiangshan <laijs@linux.alibaba.com>
>>
>> It resets all the appropriate bits like vmx.
>>
>> Signed-off-by: Lai Jiangshan <laijs@linux.alibaba.com>
>> ---
>>   arch/x86/kvm/svm/svm.c |  3 +--
>>   arch/x86/kvm/svm/svm.h | 26 ++++++++++++++++++++++++++
>>   2 files changed, 27 insertions(+), 2 deletions(-)
>>
>> diff --git a/arch/x86/kvm/svm/svm.c b/arch/x86/kvm/svm/svm.c
>> index b7da66935e72..ba9cfddd2875 100644
>> --- a/arch/x86/kvm/svm/svm.c
>> +++ b/arch/x86/kvm/svm/svm.c
>> @@ -3969,8 +3969,7 @@ static __no_kcsan fastpath_t svm_vcpu_run(struct kvm_vcpu *vcpu)
>>       svm->vmcb->control.tlb_ctl = TLB_CONTROL_DO_NOTHING;
>>       vmcb_mark_all_clean(svm->vmcb);
>> -
>> -    kvm_register_clear_available(vcpu, VCPU_EXREG_PDPTR);
>> +    svm_register_cache_reset(vcpu);
>>       /*
>>        * We need to handle MC intercepts here before the vcpu has a chance to
>> diff --git a/arch/x86/kvm/svm/svm.h b/arch/x86/kvm/svm/svm.h
>> index 0d7bbe548ac3..1cf5d5e2d0cd 100644
>> --- a/arch/x86/kvm/svm/svm.h
>> +++ b/arch/x86/kvm/svm/svm.h
>> @@ -274,6 +274,32 @@ static inline bool vmcb_is_dirty(struct vmcb *vmcb, int bit)
>>           return !test_bit(bit, (unsigned long *)&vmcb->control.clean);
>>   }
>> +static inline void svm_register_cache_reset(struct kvm_vcpu *vcpu)
>> +{
>> +/*
>> + * SVM_REGS_AVAIL_SET - The set of registers that will be updated in cache on
>> + *            demand.  Other registers not listed here are synced to
>> + *            the cache immediately after VM-Exit.
>> + *
>> + * SVM_REGS_DIRTY_SET - The set of registers that might be outdated in
>> + *            architecture. Other registers not listed here are synced
>> + *            to the architecture immediately when modifying.
>> + *
>> + *            Special case: VCPU_EXREG_CR3 should be in this set due
>> + *            to the fact.  But KVM_REQ_LOAD_MMU_PGD is always
>> + *            requested when the cache vcpu->arch.cr3 is changed and
>> + *            svm_load_mmu_pgd() always syncs the new CR3 value into
>> + *            the architecture.  So the dirty information of
>> + *            VCPU_EXREG_CR3 is not used which means VCPU_EXREG_CR3
>> + *            isn't required to be put in this set.
>> + */
>> +#define SVM_REGS_AVAIL_SET    (1 << VCPU_EXREG_PDPTR)
>> +#define SVM_REGS_DIRTY_SET    (0)
>> +
>> +    vcpu->arch.regs_avail &= ~SVM_REGS_AVAIL_SET;
>> +    vcpu->arch.regs_dirty &= ~SVM_REGS_DIRTY_SET;
>> +}
> 
> I think touching regs_dirty is confusing here, so I'd go with this:

It makes the code the same as vmx by clearing all the SVM_REGS_DIRTY_SET
bits.  And the compiler will remove this line of code since
SVM_REGS_DIRTY_SET is (0), and regs_dirty is not touched.

Using VMX_REGS_DIRTY_SET and SVM_REGS_DIRTY_SET and making the code
similar is my intent for patch12,13.  If it causes confusing, I would
like to make a second thought.  SVM_REGS_DIRTY_SET does be special
in svm where VCPU_EXREG_CR3 is in it by definition, but it is not
added into SVM_REGS_DIRTY_SET in the patch just for optimization to allow
the compiler optimizes the line of code out.

I'm Ok to not queue patch12,13,14.

> 
>          vcpu->arch.regs_avail &= ~SVM_REGS_LAZY_LOAD_SET;
> 
>          /*
>           * SVM does not use vcpu->arch.regs_dirty.  The only register that
>           * might be out of date in the VMCB is CR3, but KVM_REQ_LOAD_MMU_PGD
>           * is always requested when the cache vcpu->arch.cr3 is changed and
>           * svm_load_mmu_pgd() always syncs the new CR3 value into the VMCB.
>           */
> 
> (VMX instead needs VCPU_EXREG_CR3 mostly because it does not want to
> update it unconditionally on exit).
> 
> Paolo