Re: [Qemu-devel] [virtio-dev] Re: [PATCH v2 02/15] crypto: introduce crypto queue handler

[Ola Liljedahl]

On 14/09/2016, 03:07, "Gonglei (Arei)" <address@hidden> wrote:

>Hi Ola,
>
>
>
>> -----Original Message-----
>> From: Ola Liljedahl [mailto:address@hidden
>> Sent: Tuesday, September 13, 2016 7:53 PM
>> To: Paolo Bonzini; Daniel P. Berrange; Gonglei (Arei)
>> Cc: address@hidden; address@hidden; Huangpeng
>> (Peter); Luonengjun; address@hidden; address@hidden; Huangweidong
>> (C); address@hidden; address@hidden; address@hidden; Claudio
>> Fontana; address@hidden; address@hidden
>> Subject: Re: [virtio-dev] Re: [PATCH v2 02/15] crypto: introduce crypto
>>queue
>> handler
>>
>>
>>
>>
>>
>> On 13/09/2016, 12:58, "address@hidden on behalf of
>>Paolo
>> Bonzini" <address@hidden on behalf of
>> address@hidden> wrote:
>>
>> >
>> >
>> >On 13/09/2016 11:20, Daniel P. Berrange wrote:
>> >>> > +typedef struct CryptoPacket CryptoPacket;
>> >>> > +typedef struct CryptoQueue CryptoQueue;
>> >>> > +typedef struct CryptoPacketBuf CryptoPacketBuf;
>> >>> > +
>> >>> > +typedef void (CryptoPacketSent) (CryptoClientState *, int);
>> >> As previously, I'd expect naming of
>> >>
>> >>  QCryptoCryptodevPacket
>> >>  QCryptoCryptodevPacketBuf
>> >>  QCryptoCryptodevQueue
>> >>
>> >
>> >Gonglei,
>> >
>> >you are copying a lot of code from network backends.
>> >
>> >I am not sure why you would need a queue for virtio-crypto rather than
>>a
>> >direct connection between frontend and backend (and the backend would
>>be
>> >QEMU crypto APIs, like Daniel suggested).
>>
>> What about backends implemented directly in HW? Bypass the middle man.
>> Make crypto offload meaningful for small size blocks.
>>
>As I said in other reply, I only provide interfaces, you can realize them
>according
>to different backend cryptodevs.
Interfaces imposes contraints on performance and implementation. A poor
interface could make certain implementations unnecessarily difficult.
Since these virtio-* interfaces will live for a long time, we should
ensure that they do not impose unnecessary constraints. For functionality
like crypto and IPsec, I can see the benefit of implementing the backend
in HW (for the datapath).

>
>Regards,
>-Gonglei
>
>> < Ola
>>
>> >
>> >Paolo
>> >
>> >---------------------------------------------------------------------
>> >To unsubscribe, e-mail: address@hidden
>> >For additional commands, e-mail: address@hidden
>> >
>>
>> IMPORTANT NOTICE: The contents of this email and any attachments are
>> confidential and may also be privileged. If you are not the intended
>>recipient,
>> please notify the sender immediately and do not disclose the contents
>>to any
>> other person, use it for any purpose, or store or copy the information
>>in any
>> medium. Thank you.
>

IMPORTANT NOTICE: The contents of this email and any attachments are 
confidential and may also be privileged. If you are not the intended recipient, 
please notify the sender immediately and do not disclose the contents to any 
other person, use it for any purpose, or store or copy the information in any 
medium. Thank you.