From mboxrd@z Thu Jan  1 00:00:00 1970
From: Tiwei Bie <tiwei.bie@intel.com>
Subject: Re: [PATCH v8 5/8] net/virtio: implement transmit path
 for packed queues
Date: Thu, 18 Oct 2018 12:43:20 +0800
Message-ID: <20181018044320.GA15631@debian>
References: <20181016090134.29448-1-jfreimann@redhat.com>
 <20181016090134.29448-6-jfreimann@redhat.com>
Mime-Version: 1.0
Content-Type: text/plain; charset=utf-8
Cc: dev@dpdk.org, maxime.coquelin@redhat.com, Gavin.Hu@arm.com,
 zhihong.wang@intel.com
To: Jens Freimann <jfreimann@redhat.com>
Return-path: <dev-bounces@dpdk.org>
Received: from mga05.intel.com (mga05.intel.com [192.55.52.43])
 by dpdk.org (Postfix) with ESMTP id 32B5B4C96
 for <dev@dpdk.org>; Thu, 18 Oct 2018 06:44:43 +0200 (CEST)
Content-Disposition: inline
In-Reply-To: <20181016090134.29448-6-jfreimann@redhat.com>
List-Id: DPDK patches and discussions <dev.dpdk.org>
List-Unsubscribe: <https://mails.dpdk.org/options/dev>,
 <mailto:dev-request@dpdk.org?subject=unsubscribe>
List-Archive: <http://mails.dpdk.org/archives/dev/>
List-Post: <mailto:dev@dpdk.org>
List-Help: <mailto:dev-request@dpdk.org?subject=help>
List-Subscribe: <https://mails.dpdk.org/listinfo/dev>,
 <mailto:dev-request@dpdk.org?subject=subscribe>
Errors-To: dev-bounces@dpdk.org
Sender: "dev" <dev-bounces@dpdk.org>

On Tue, Oct 16, 2018 at 11:01:31AM +0200, Jens Freimann wrote:
[...]
> +void
> +vq_ring_free_chain_packed(struct virtqueue *vq, uint16_t desc_idx)
> +{
> +	struct vring_desc_packed *dp;
> +	struct vq_desc_extra *dxp = NULL, *dxp_tail = NULL;
> +	uint16_t desc_idx_last = desc_idx;
> +
> +	dp  = &vq->vq_ring.desc_packed[desc_idx];
> +	dxp = &vq->vq_descx[desc_idx];

In the out-of-order case, we can't assume the desc and desc_extra
have the same index.

> +	vq->vq_free_cnt = (uint16_t)(vq->vq_free_cnt + dxp->ndescs);
> +	if ((dp->flags & VRING_DESC_F_INDIRECT) == 0) {
> +		while (dp->flags & VRING_DESC_F_NEXT) {

https://github.com/oasis-tcs/virtio-spec/blob/89dd55f5e606/packed-ring.tex#L222-L226
The device only writes out a single used descriptor for the whole
list. It then skips forward according to the number of
descriptors in the list. The driver needs to keep track of the size
of the list corresponding to each buffer ID, to be able to skip
to where the next used descriptor is written by the device.

IOW, we can't use the VRING_DESC_F_NEXT here.


> +			desc_idx_last = dxp->next;
> +			dp = &vq->vq_ring.desc_packed[dxp->next];
> +			dxp = &vq->vq_descx[dxp->next];
> +		}
> +	}