Re: [PATCH net-next v3 4/4] vsock/virtio: MSG_ZEROCOPY flag support

[Stefano Garzarella <sgarzare@redhat.com>]

On Thu, Jul 27, 2023 at 11:32:00AM +0300, Arseniy Krasnov wrote:
>On 25.07.2023 15:28, Stefano Garzarella wrote:
>> On Tue, Jul 25, 2023 at 12:16:11PM +0300, Arseniy Krasnov wrote:
>>> On 25.07.2023 11:46, Arseniy Krasnov wrote:
>>>> On 25.07.2023 11:43, Stefano Garzarella wrote:
>>>>> On Fri, Jul 21, 2023 at 08:09:03AM +0300, Arseniy Krasnov wrote:

[...]

>>>>>>> +    t = vsock_core_get_transport(info->vsk);
>>>>>>>
>>>>>>> -        if (msg_data_left(info->msg) == 0 &&
>>>>>>> -            info->type == VIRTIO_VSOCK_TYPE_SEQPACKET) {
>>>>>>> -            hdr->flags |= cpu_to_le32(VIRTIO_VSOCK_SEQ_EOM);
>>>>>>> +    if (t->msgzerocopy_check_iov &&
>>>>>>> +        !t->msgzerocopy_check_iov(iov_iter))
>>>>>>> +        return false;
>>>>>
>>>>> I'd avoid adding a new transport callback used only internally in virtio
>>>>> transports.
>>>>
>>>> Ok, I see.
>>>>
>>>>>
>>>>> Usually the transport callbacks are used in af_vsock.c, if we need a
>>>>> callback just for virtio transports, maybe better to add it in struct
>>>>> virtio_vsock_pkt_info or struct virtio_vsock_sock.
>>>
>>> Hm, may be I just need to move this callback from 'struct vsock_transport' to parent 'struct virtio_transport',
>>> after 'send_pkt' callback. In this case:
>>> 1) AF_VSOCK part is not touched.
>>> 2) This callback stays in 'virtio_transport.c' and is set also in this file.
>>>   vhost and loopback are unchanged - only 'send_pkt' still enabled in both
>>>   files for these two transports.
>>
>> Yep, this could also work!
>>
>> Stefano
>
>Great! I'll send this implementation when this patchset for MSG_PEEK will be merged
>to net-next as both conflicts with each other.
>
>https://lore.kernel.org/netdev/20230726060150-mutt-send-email-mst@kernel.org/T/#m56f3b850361a412735616145162d2d9df25f6350

Ack!

Thanks,
Stefano