Re: [Xen-devel] [PATCH v2 1/2] SVM: re-work VMCB sync-ing

["Jan Beulich" <JBeulich@xxxxxxxx>]

>>> On 30.04.18 at 17:56, <boris.ostrovsky@xxxxxxxxxx> wrote:
> On 04/30/2018 11:50 AM, Jan Beulich wrote:
>>>>> On 30.04.18 at 17:30, <boris.ostrovsky@xxxxxxxxxx> wrote:
>>> On 04/30/2018 07:37 AM, Jan Beulich wrote:
>>>> @@ -1168,6 +1169,9 @@ static void noreturn svm_do_resume(struc
>>>>  
>>>>      hvm_do_resume(v);
>>>>  
>>>> +    if ( v->arch.hvm_svm.vmcb_sync_state == vmcb_needs_vmload )
>>>> +        svm_sync_vmcb(v, vmcb_needs_vmsave);
>>>
>>> Is it not possible (or advisable) to move the test into svm_sync_vmcb()
>>> (and drop the ASSERT there)?
>> It is possible; I'm not sure myself if it's advisable, but I take you asking 
>> the
>> question as you thinking it is, so I'll change it.
> 
> That was really the main reason for me asking to move svm_vmload into
> svm_sync_vmcb() --- to hide all logic related to the sync state machine
> there.

Well, there's still the code in svm_vmexit_do_vmload(). Depending on
your opinion on the post-commit-message question in patch 2, the one
at the top of svm_vmexit_handler() might go away in that patch.

Jan



_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxxx
https://lists.xenproject.org/mailman/listinfo/xen-devel