Re: [Qemu-devel] [PATCH V5 2/9] net/filter-mirror.c: Make filter mirror support vnet support.

[Jason Wang <jasowang@redhat.com>]

On 2017年05月25日 20:55, Zhang Chen wrote:
>
>
> On 05/25/2017 02:04 PM, Jason Wang wrote:
>>
>>
>> On 2017年05月23日 22:20, Zhang Chen wrote:
>>> We add the vnet_hdr option for filter-mirror, default is disable.
>>> If you use virtio-net-pci net driver, please enable it.
>>> You can use it for example:
>>> -object 
>>> filter-mirror,id=m0,netdev=hn0,queue=tx,outdev=mirror0,vnet_hdr=on
>>>
>>> If vnet_hdr=on we change the send packet format from
>>> struct {int size; const uint8_t buf[];} to {int size; int 
>>> vnet_hdr_len; const uint8_t buf[];}.
>>> make other module(like colo-compare) know how to parse net packet 
>>> correctly.
>>>
>>> Signed-off-by: Zhang Chen <zhangchen.fnst@cn.fujitsu.com>
>>> ---
>>>   net/filter-mirror.c | 73 
>>> +++++++++++++++++++++++++++++++++++++++++++++++++----
>>>   qemu-options.hx     |  5 ++--
>>>   2 files changed, 70 insertions(+), 8 deletions(-)
>>>
>>> diff --git a/net/filter-mirror.c b/net/filter-mirror.c
>>> index 72fa7c2..8df0be6 100644
>>> --- a/net/filter-mirror.c
>>> +++ b/net/filter-mirror.c
>>> @@ -38,15 +38,17 @@ typedef struct MirrorState {
>>>       NetFilterState parent_obj;
>>>       char *indev;
>>>       char *outdev;
>>> +    bool vnet_hdr;
>>>       CharBackend chr_in;
>>>       CharBackend chr_out;
>>>       SocketReadState rs;
>>>   } MirrorState;
>>>   -static int filter_mirror_send(CharBackend *chr_out,
>>> +static int filter_mirror_send(MirrorState *s,
>>>                                 const struct iovec *iov,
>>>                                 int iovcnt)
>>>   {
>>> +    NetFilterState *nf = NETFILTER(s);
>>>       int ret = 0;
>>>       ssize_t size = 0;
>>>       uint32_t len = 0;
>>> @@ -58,14 +60,42 @@ static int filter_mirror_send(CharBackend *chr_out,
>>>       }
>>>         len = htonl(size);
>>> -    ret = qemu_chr_fe_write_all(chr_out, (uint8_t *)&len, 
>>> sizeof(len));
>>> +    ret = qemu_chr_fe_write_all(&s->chr_out, (uint8_t *)&len, 
>>> sizeof(len));
>>>       if (ret != sizeof(len)) {
>>>           goto err;
>>>       }
>>>   +    if (s->vnet_hdr) {
>>> +        /*
>>> +         * If vnet_hdr = on, we send vnet header len to make other
>>> +         * module(like colo-compare) know how to parse net
>>> +         * packet correctly.
>>> +         */
>>> +        ssize_t vnet_hdr_len;
>>> +
>>> +        /*
>>> +         * In anytime, nf->netdev and nf->netdev->peer both have a 
>>> vnet_hdr_len,
>>> +         * Here we just find out which is we need. When filter set 
>>> RX or TX
>>> +         * that the real vnet_hdr_len are different.
>>> +         */
>>> +        if (nf->netdev->using_vnet_hdr) {
>>> +            vnet_hdr_len = nf->netdev->vnet_hdr_len;
>>> +        } else if (nf->netdev->peer->using_vnet_hdr) {
>>> +            vnet_hdr_len = nf->netdev->peer->vnet_hdr_len;
>>
>> Rethink about this, looks like what we really want is:
>>
>> - For tx direction, only check nf->netdev->vnet_hdr_len,
>> - For rx direction, only check nf->netdev->peer->vnet_hdr_len
>
> Agree, So do you means we should remove the using_vnet_hdr related 
> codes then just
> use the nf->direction to choice which we need check?
>
> Thanks
> Zhang Chen 

Maybe.

Thanks