Re: [Qemu-devel] [RFC 4/6] virtio-net: implement steering mode feature

[Jason Wang]

On 2018年08月30日 22:27, Sameeh Jubran wrote:
> From: Sameeh Jubran <address@hidden>
>
> Signed-off-by: Sameeh Jubran <address@hidden>
> ---
>   hw/net/virtio-net.c                         | 65 +++++++++++++++++++++++++----
>   include/hw/virtio/virtio-net.h              |  3 ++
>   include/standard-headers/linux/virtio_net.h | 55 ++++++++++++++++++++++++
>   3 files changed, 116 insertions(+), 7 deletions(-)
>
> diff --git a/hw/net/virtio-net.c b/hw/net/virtio-net.c
> index 90502fca7c..e7c4ce6f66 100644
> --- a/hw/net/virtio-net.c
> +++ b/hw/net/virtio-net.c
> @@ -972,13 +972,53 @@ static int virtio_net_handle_mq(VirtIONet *n, uint8_t cmd,
>       return VIRTIO_NET_OK;
>   }
>   
> +static int virtio_net_ctrl_steering_mode(VirtIONet *n, uint8_t cmd,
> +                                struct iovec *iov, unsigned int iov_cnt,
> +                                struct iovec *iov_in, unsigned int iov_cnt_in,
> +        size_t *size_in)
> +{
> +    size_t s;
> +    struct virtio_net_steering_mode sm;
> +
> +    switch (cmd) {
> +    case VIRTIO_NET_CTRL_SM_GET_SUPPORTED_MODES:
> +        if (!size_in) {
> +            return VIRTIO_NET_ERR;
> +        }
> +                  s = iov_from_buf(iov_in, iov_cnt_in, 0,
> +          &n->supported_modes, sizeof(n->supported_modes));
> +        if (s != sizeof(n->supported_modes) ||
> +          !size_in) {

size_in has been checked in the above I think?

> +            return VIRTIO_NET_ERR;
> +        }
> +                  *size_in = s;
> +      break;
> +    case VIRTIO_NET_CTRL_SM_CONTROL:
> +        s = iov_to_buf(iov, iov_cnt, 0, &sm, sizeof(sm) -
> +                sizeof(union command_data));
> +        if (s != sizeof(sm) - sizeof(union command_data)) {
> +            return VIRTIO_NET_ERR;
> +        }
> +        /* switch (cmd)
> +             {
> +                dafault:
> +                return VIRTIO_NET_ERR;
> +         } */
> +      break;
> +    default:
> +                return VIRTIO_NET_ERR;
> +    }
> +
> +    return VIRTIO_NET_OK;
> +}
> +
>   static void virtio_net_handle_ctrl(VirtIODevice *vdev, VirtQueue *vq)
>   {
>       VirtIONet *n = VIRTIO_NET(vdev);
>       struct virtio_net_ctrl_hdr ctrl;
>       virtio_net_ctrl_ack status = VIRTIO_NET_ERR;
>       VirtQueueElement *elem;
> -    size_t s;
> +    size_t s, elem_in_size = 0;
>       struct iovec *iov, *iov2;
>       unsigned int iov_cnt;
>   
> @@ -996,7 +1036,8 @@ static void virtio_net_handle_ctrl(VirtIODevice *vdev, VirtQueue *vq)
>           }
>   
>           iov_cnt = elem->out_num;
> -        iov2 = iov = g_memdup(elem->out_sg, sizeof(struct iovec) * 
> elem->out_num);
> +        iov2 = iov = g_memdup(elem->out_sg, sizeof(struct iovec) *
> +                elem->out_num);

Still looks unnecessary.

>           s = iov_to_buf(iov, iov_cnt, 0, &ctrl, sizeof(ctrl));
>           iov_discard_front(&iov, &iov_cnt, sizeof(ctrl));
>           if (s != sizeof(ctrl)) {
> @@ -1013,12 +1054,20 @@ static void virtio_net_handle_ctrl(VirtIODevice *vdev, 
> VirtQueue *vq)
>               status = virtio_net_handle_mq(n, ctrl.cmd, iov, iov_cnt);
>           } else if (ctrl.class == VIRTIO_NET_CTRL_GUEST_OFFLOADS) {
>               status = virtio_net_handle_offloads(n, ctrl.cmd, iov, iov_cnt);
> +        } else if (ctrl.class == VIRTIO_NET_CTRL_STEERING_MODE) {
> +            size_t size_in = 0;
> +            status = virtio_net_ctrl_steering_mode(n, ctrl.cmd, iov, iov_cnt,
> +                   elem->in_sg, elem->in_num, &size_in);
> +            if (status == VIRTIO_NET_OK  && size_in > 0) {
> +                elem_in_size += size_in;
> +        }
>           }
>   
> -        s = iov_from_buf(elem->in_sg, elem->in_num, 0, &status, sizeof(status));
> +        s = iov_from_buf(elem->in_sg, elem->in_num, elem_in_size, &status,
> +                sizeof(status));
>           assert(s == sizeof(status));
> -
> -        virtqueue_push(vq, elem, sizeof(status));
> +        elem_in_size += s;
> +        virtqueue_push(vq, elem, elem_in_size);
>           virtio_notify(vdev, vq);
>           g_free(iov2);
>           g_free(elem);
> @@ -1375,10 +1424,10 @@ static int32_t virtio_net_flush_tx(VirtIONetQueue *q)
>                                      n->guest_hdr_len, -1);
>                   if (out_num == VIRTQUEUE_MAX_SIZE) {
>                       goto drop;
> -               }
> +    }
>                   out_num += 1;
>                   out_sg = sg2;
> -           }
> +      }

Let's avoid mixing such possible style fixes.

>           }
>           /*
>            * If host wants to see the guest header as is, we can
> @@ -1957,6 +2006,8 @@ static void virtio_net_device_realize(DeviceState *dev, 
> Error **errp)
>           n->host_features |= (1ULL << VIRTIO_NET_F_MTU);
>       }
>   
> +        n->host_features |= (1ULL << VIRTIO_NET_F_CTRL_STEERING_MODE);

I think we need a property for virtio-net and compat this for old 
machine types.

> +
>       if (n->net_conf.duplex_str) {
>           if (strncmp(n->net_conf.duplex_str, "half", 5) == 0) {
>               n->net_conf.duplex = DUPLEX_HALF;
> diff --git a/include/hw/virtio/virtio-net.h b/include/hw/virtio/virtio-net.h
> index a7b53edc96..809e85481c 100644
> --- a/include/hw/virtio/virtio-net.h
> +++ b/include/hw/virtio/virtio-net.h
> @@ -102,6 +102,9 @@ typedef struct VirtIONet {
>       int announce_counter;
>       bool needs_vnet_hdr_swap;
>       bool mtu_bypass_backend;
> +    struct virtio_net_steering_modes supported_modes;
> +    struct virtio_net_steering_modes current_mode;
> +    struct virtio_net_rss_conf *rss_conf;
>   } VirtIONet;
>   
>   void virtio_net_set_netclient_name(VirtIONet *n, const char *name,
> diff --git a/include/standard-headers/linux/virtio_net.h 
> b/include/standard-headers/linux/virtio_net.h
> index e9f255ea3f..fa399b97ab 100644
> --- a/include/standard-headers/linux/virtio_net.h
> +++ b/include/standard-headers/linux/virtio_net.h
> @@ -258,4 +258,59 @@ struct virtio_net_ctrl_mq {
>   #define VIRTIO_NET_CTRL_GUEST_OFFLOADS   5
>   #define VIRTIO_NET_CTRL_GUEST_OFFLOADS_SET        0
>   
> +
> +#define RSS_HASH_FUNCTION_TOEPLITZ       0x1
> +#define RSS_HASH_FUNCTION_SYMMETRIC      0x2
> +
> +// Hash function fields
> +#define RSS_HASH_FIELDS_IPV4             0x00000100
> +#define RSS_HASH_FIELDS_TCP_IPV4         0x00000200
> +#define RSS_HASH_FIELDS_IPV6             0x00000400
> +#define RSS_HASH_FIELDS_IPV6_EX          0x00000800
> +#define RSS_HASH_FIELDS_TCP_IPV6         0x00001000
> +#define RSS_HASH_FIELDS_TCP_IPV6_EX      0x00002000
> +
> +struct virtio_net_rss_supported_hash{
> +uint32_t hash_function;
> +};
> +
> +struct virtio_net_rss_conf_ptrs {
> +    uint8_t *hash_key;
> +    uint32_t  *indirection_table;
> +};
> +
> +struct virtio_net_rss_conf {
> +    uint32_t hash_function;
> +    uint32_t hash_function_flags;
> +    uint32_t hash_key_length;
> +    uint32_t indirection_table_length;
> +    struct virtio_net_rss_conf_ptrs ptrs;
> +};

I think the RSS stuffs should go with an independent patch?

> +
> +#define VIRTIO_NET_SM_CTRL_RSS_GET_SUPPORTED_FUNCTIONS   0
> +#define VIRTIO_NET_SM_CTRL_RSS_SET                       1
> +
> +
> +struct virtio_net_steering_modes {
> +    uint32_t steering_modes;
> +};
> +
> +union command_data {
> +    struct virtio_net_rss_conf rss;
> +};
> +
> +struct virtio_net_steering_mode {
> +    uint32_t steering_mode;
> +    uint32_t command;
> +};
> +
> +#define VIRTIO_NET_F_CTRL_STEERING_MODE  60
> +
> +#define VIRTIO_NET_CTRL_STEERING_MODE             7
> +#define VIRTIO_NET_CTRL_SM_GET_SUPPORTED_MODES    0
> +#define VIRTIO_NET_CTRL_SM_CONTROL                1

SM_CONTROL sounds pretty generic, maybe just SET_SUPPORTED_MODES?

Thanks

> +
> +#define STEERING_MODE_AUTO          0x1
> +#define STEERING_MODE_RSS           0x2
> +
>   #endif /* _LINUX_VIRTIO_NET_H */