On 10/15/20 7:46 AM, Marvin Liu wrote:
> When transmitting indirect descriptors, first desc will store net_hdr
> and following descs will be mapped to mbuf segments. Total desc number
> will be seg_num plus one. Meaning of variable needed is the number of
> used descs in packed ring. This value will always be two for indirect
> desc. Now use mbuf segments number for calculating correct desc length.
>
> Fixes: de8b3d238074 ("net/virtio: fix indirect descs in packed datapaths")
> Cc: sta...@dpdk.org
>
> Signed-off-by: Marvin Liu <yong....@intel.com>
>
> diff --git a/drivers/net/virtio/virtqueue.h b/drivers/net/virtio/virtqueue.h
> index 8c8ab9889..42c4c9882 100644
> --- a/drivers/net/virtio/virtqueue.h
> +++ b/drivers/net/virtio/virtqueue.h
> @@ -698,6 +698,7 @@ virtqueue_enqueue_xmit_packed(struct virtnet_tx *txvq,
> struct rte_mbuf *cookie,
> struct virtio_net_hdr *hdr;
> uint16_t prev;
> bool prepend_header = false;
> + uint16_t seg_num = cookie->nb_segs;
>
> id = in_order ? vq->vq_avail_idx : vq->vq_desc_head_idx;
>
> @@ -732,7 +733,7 @@ virtqueue_enqueue_xmit_packed(struct virtnet_tx *txvq,
> struct rte_mbuf *cookie,
> */
> start_dp[idx].addr = txvq->virtio_net_hdr_mem +
> RTE_PTR_DIFF(&txr[idx].tx_packed_indir, txr);
> - start_dp[idx].len = (needed + 1) *
> + start_dp[idx].len = (seg_num + 1) *
> sizeof(struct vring_packed_desc);
> /* reset flags for indirect desc */
> head_flags = VRING_DESC_F_INDIRECT;
>
Applied to dpdk-next-virtio/main.
Thanks,
Maxime
