[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Xen-devel] [PATCH] SVM: re-work VMCB sync-ing


  • To: Jan Beulich <JBeulich@xxxxxxxx>
  • From: Boris Ostrovsky <boris.ostrovsky@xxxxxxxxxx>
  • Date: Fri, 27 Apr 2018 14:29:47 -0400
  • Autocrypt: addr=boris.ostrovsky@xxxxxxxxxx; prefer-encrypt=mutual; keydata= xsFNBFH8CgsBEAC0KiOi9siOvlXatK2xX99e/J3OvApoYWjieVQ9232Eb7GzCWrItCzP8FUV PQg8rMsSd0OzIvvjbEAvaWLlbs8wa3MtVLysHY/DfqRK9Zvr/RgrsYC6ukOB7igy2PGqZd+M MDnSmVzik0sPvB6xPV7QyFsykEgpnHbvdZAUy/vyys8xgT0PVYR5hyvhyf6VIfGuvqIsvJw5 C8+P71CHI+U/IhsKrLrsiYHpAhQkw+Zvyeml6XSi5w4LXDbF+3oholKYCkPwxmGdK8MUIdkM d7iYdKqiP4W6FKQou/lC3jvOceGupEoDV9botSWEIIlKdtm6C4GfL45RD8V4B9iy24JHPlom woVWc0xBZboQguhauQqrBFooHO3roEeM1pxXjLUbDtH4t3SAI3gt4dpSyT3EvzhyNQVVIxj2 FXnIChrYxR6S0ijSqUKO0cAduenhBrpYbz9qFcB/GyxD+ZWY7OgQKHUZMWapx5bHGQ8bUZz2 SfjZwK+GETGhfkvNMf6zXbZkDq4kKB/ywaKvVPodS1Poa44+B9sxbUp1jMfFtlOJ3AYB0WDS Op3d7F2ry20CIf1Ifh0nIxkQPkTX7aX5rI92oZeu5u038dHUu/dO2EcuCjl1eDMGm5PLHDSP 0QUw5xzk1Y8MG1JQ56PtqReO33inBXG63yTIikJmUXFTw6lLJwARAQABzTNCb3JpcyBPc3Ry b3Zza3kgKFdvcmspIDxib3Jpcy5vc3Ryb3Zza3lAb3JhY2xlLmNvbT7CwXgEEwECACIFAlH8 CgsCGwMGCwkIBwMCBhUIAgkKCwQWAgMBAh4BAheAAAoJEIredpCGysGyasEP/j5xApopUf4g 9Fl3UxZuBx+oduuw3JHqgbGZ2siA3EA4bKwtKq8eT7ekpApn4c0HA8TWTDtgZtLSV5IdH+9z JimBDrhLkDI3Zsx2CafL4pMJvpUavhc5mEU8myp4dWCuIylHiWG65agvUeFZYK4P33fGqoaS VGx3tsQIAr7MsQxilMfRiTEoYH0WWthhE0YVQzV6kx4wj4yLGYPPBtFqnrapKKC8yFTpgjaK jImqWhU9CSUAXdNEs/oKVR1XlkDpMCFDl88vKAuJwugnixjbPFTVPyoC7+4Bm/FnL3iwlJVE qIGQRspt09r+datFzPqSbp5Fo/9m4JSvgtPp2X2+gIGgLPWp2ft1NXHHVWP19sPgEsEJXSr9 tskM8ScxEkqAUuDs6+x/ISX8wa5Pvmo65drN+JWA8EqKOHQG6LUsUdJolFM2i4Z0k40BnFU/ kjTARjrXW94LwokVy4x+ZYgImrnKWeKac6fMfMwH2aKpCQLlVxdO4qvJkv92SzZz4538az1T m+3ekJAimou89cXwXHCFb5WqJcyjDfdQF857vTn1z4qu7udYCuuV/4xDEhslUq1+GcNDjAhB nNYPzD+SvhWEsrjuXv+fDONdJtmLUpKs4Jtak3smGGhZsqpcNv8nQzUGDQZjuCSmDqW8vn2o hWwveNeRTkxh+2x1Qb3GT46uzsFNBFH8CgsBEADGC/yx5ctcLQlB9hbq7KNqCDyZNoYu1HAB Hal3MuxPfoGKObEktawQPQaSTB5vNlDxKihezLnlT/PKjcXC2R1OjSDinlu5XNGc6mnky03q yymUPyiMtWhBBftezTRxWRslPaFWlg/h/Y1iDuOcklhpr7K1h1jRPCrf1yIoxbIpDbffnuyz kuto4AahRvBU4Js4sU7f/btU+h+e0AcLVzIhTVPIz7PM+Gk2LNzZ3/on4dnEc/qd+ZZFlOQ4 KDN/hPqlwA/YJsKzAPX51L6Vv344pqTm6Z0f9M7YALB/11FO2nBB7zw7HAUYqJeHutCwxm7i BDNt0g9fhviNcJzagqJ1R7aPjtjBoYvKkbwNu5sWDpQ4idnsnck4YT6ctzN4I+6lfkU8zMzC gM2R4qqUXmxFIS4Bee+gnJi0Pc3KcBYBZsDK44FtM//5Cp9DrxRQOh19kNHBlxkmEb8kL/pw XIDcEq8MXzPBbxwHKJ3QRWRe5jPNpf8HCjnZz0XyJV0/4M1JvOua7IZftOttQ6KnM4m6WNIZ 2ydg7dBhDa6iv1oKdL7wdp/rCulVWn8R7+3cRK95SnWiJ0qKDlMbIN8oGMhHdin8cSRYdmHK kTnvSGJNlkis5a+048o0C6jI3LozQYD/W9wq7MvgChgVQw1iEOB4u/3FXDEGulRVko6xCBU4 SQARAQABwsFfBBgBAgAJBQJR/AoLAhsMAAoJEIredpCGysGyfvMQAIywR6jTqix6/fL0Ip8G jpt3uk//QNxGJE3ZkUNLX6N786vnEJvc1beCu6EwqD1ezG9fJKMl7F3SEgpYaiKEcHfoKGdh 30B3Hsq44vOoxR6zxw2B/giADjhmWTP5tWQ9548N4VhIZMYQMQCkdqaueSL+8asp8tBNP+TJ PAIIANYvJaD8xA7sYUXGTzOXDh2THWSvmEWWmzok8er/u6ZKdS1YmZkUy8cfzrll/9hiGCTj u3qcaOM6i/m4hqtvsI1cOORMVwjJF4+IkC5ZBoeRs/xW5zIBdSUoC8L+OCyj5JETWTt40+lu qoqAF/AEGsNZTrwHJYu9rbHH260C0KYCNqmxDdcROUqIzJdzDKOrDmebkEVnxVeLJBIhYZUd t3Iq9hdjpU50TA6sQ3mZxzBdfRgg+vaj2DsJqI5Xla9QGKD+xNT6v14cZuIMZzO7w0DoojM4 ByrabFsOQxGvE0w9Dch2BDSI2Xyk1zjPKxG1VNBQVx3flH37QDWpL2zlJikW29Ws86PHdthh Fm5PY8YtX576DchSP6qJC57/eAAe/9ztZdVAdesQwGb9hZHJc75B+VNm4xrh/PJO6c1THqdQ 19WVJ+7rDx3PhVncGlbAOiiiE3NOFPJ1OQYxPKtpBUukAlOTnkKE6QcA4zckFepUkfmBV1wM Jg6OxFYd01z+a+oL
  • Cc: Juergen Gross <jgross@xxxxxxxx>, Andrew Cooper <andrew.cooper3@xxxxxxxxxx>, Razvan Cojocaru <rcojocaru@xxxxxxxxxxxxxxx>, Suravee Suthikulpanit <suravee.suthikulpanit@xxxxxxx>, xen-devel <xen-devel@xxxxxxxxxxxxxxxxxxxx>
  • Delivery-date: Fri, 27 Apr 2018 18:28:30 +0000
  • List-id: Xen developer discussion <xen-devel.lists.xenproject.org>
  • Openpgp: preference=signencrypt

On 04/27/2018 11:52 AM, Jan Beulich wrote:
>>>> On 26.04.18 at 19:27, <boris.ostrovsky@xxxxxxxxxx> wrote:
>> On 04/26/2018 11:55 AM, Jan Beulich wrote:
>>>>>> On 26.04.18 at 17:20, <boris.ostrovsky@xxxxxxxxxx> wrote:
>>>> On 04/26/2018 09:33 AM, Jan Beulich wrote:
>>>>>>> -static void svm_sync_vmcb(struct vcpu *v)
>>>>>>> +static void svm_sync_vmcb(struct vcpu *v, enum vmcb_sync_state 
>>>>>>> new_state)
>>>>>>>  {
>>>>>>>      struct arch_svm_struct *arch_svm = &v->arch.hvm_svm;
>>>>>>>  
>>>>>>> -    if ( arch_svm->vmcb_in_sync )
>>>>>>> -        return;
>>>>>>> -
>>>>>>> -    arch_svm->vmcb_in_sync = 1;
>>>>>>> +    if ( arch_svm->vmcb_sync_state == vmcb_needs_vmsave )
>>>>>>> +        svm_vmsave(arch_svm->vmcb);
>>>>>>>  
>>>>>>> -    svm_vmsave(arch_svm->vmcb);
>>>>>>> +    if ( arch_svm->vmcb_sync_state != vmcb_needs_vmload )
>>>>>>> +        arch_svm->vmcb_sync_state = new_state;
>>>>>> This is slightly awkward for a couple of reasons.  First, passing
>>>>>> vmcb_in_sync in forget the fact that a vmload is needed.
>>>>> Certainly not - that's the purpose of the if() around it.
>>>>>
>>>>>> In my patch, I introduced svm_sync_vmcb_for_update(), rather than
>>>>>> requiring a parameter to be passed in.  I think this is a better API,
>>>>>> and it shrinks the size of the patch.
>>>>> I'm not convinced of the "better", and even less so of the "shrinks". But
>>>>> I'll wait to see what the SVM maintainers say.
>>>> I think a single function is better. In fact, I was wondering whether
>>>> svm_vmload() could also be folded into svm_sync_vmcb() since it is also
>>>> a syncing operation.
>>> That doesn't look like it would produce a usable interface: How would
>>> you envision the state transition to be specified by the caller? Right
>>> now the intended new state gets passed in, but in your model
>>> vmcb_in_sync could mean either vmload or vmsave is needed. The
>>> two svm_vmload() uses right now would pass that value in addition
>>> to the svm_sync_vmcb() calls already doing so. And the function
>>> couldn't tell what to do from the current state (if it's
>>> vmcb_needs_vmload, a load is only needed in the cases where
>>> svm_vmload() is called right now). Adding a 3rd parameter or a
>>> second enum 
>> I was thinking about another enum value, e.g. sync_to_cpu (and
>> sync_to_vmcb replacing vmcb_needs_vmsave).
>>
>> This will allow us to hide (v->arch.hvm_svm.vmcb_sync_state ==
>> vmcb_needs_vmload) test.
> I'm still not entirely clear how you want that to look like. At the example
> of svm_get_segment_register() and svm_set_segment_register(), how
> would the svm_sync_vmcb() calls look like? I.e. how do you distinguish
> the sync_to_vmcb/transition-to-clean case from the
> sync_to_vmcb/transition-to-dirty one?


Something like

static void svm_sync_vmcb(struct vcpu *v, enum vmcb_sync_state new_state)
{
     struct arch_svm_struct *arch_svm = &v->arch.hvm_svm;

     
    if ( new_state == vmcb_sync_to_cpu )
    {
        if (v->arch.hvm_svm.vmcb_sync_state == vmcb_needs_vmload )
        {     
            svm_vmload(vmcb);
            v->arch.hvm_svm.vmcb_sync_state = vmcb_in_sync;
        }
        return;
    }

    if ( arch_svm->vmcb_sync_state == vmcb_needs_vmsave )
        svm_vmsave(arch_svm->vmcb);
 
    if ( arch_svm->vmcb_sync_state != vmcb_needs_vmload )
        arch_svm->vmcb_sync_state = new_state;
 }


-boris


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

 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.