Re: [PATCH] x86/nSVM: Save L2's CR4 on #VMEXIT, not Xen's

[Jan Beulich <jbeulich@suse.com>]

On 28.09.2026 11:26, Lin Liu wrote:
>>> On 23.09.2026 11:51, Lin Liu wrote:
>>> --- a/xen/arch/x86/hvm/svm/nestedsvm.c
>>> +++ b/xen/arch/x86/hvm/svm/nestedsvm.c
>>> @@ -1082,7 +1082,7 @@ nsvm_vmcb_prepare4vmexit(struct vcpu *v, struct cpu_user_regs *regs)
>>>      ns_vmcb->_efer = n2vmcb->_efer;
>>>  
>>>      /* CRn */
>>> -    ns_vmcb->_cr4 = n2vmcb->_cr4;
>>> +    ns_vmcb->_cr4 = v->arch.hvm.guest_cr[4];
>>>      ns_vmcb->_cr0 = n2vmcb->_cr0;
>>
>> In addition to mirroring the change to CR0 and EFER, doesn't CR2 also
>> need handling the same way? Effectively the inverse direction of anything
>> respective that nsvm_vmcb_prepare4vmrun() does?
> 
> CR2 is different, it store the page fault address on #PF.
> If #PF happens, xen as hardware emulator, should write back the CR2 as part of
> guest state area. so copy from n2vmcb is the right behavior.

But besides by #PF, CR2 may also be changed by the guest writing to it.

Jan