DPDK patches and discussions
 help / color / mirror / Atom feed
From: Cheng Jiang <cheng1.jiang@intel.com>
To: maxime.coquelin@redhat.com, chenbo.xia@intel.com
Cc: dev@dpdk.org, jiayu.hu@intel.com, xuan.ding@intel.com,
	wenwux.ma@intel.com, yuanx.wang@intel.com,
	yvonnex.yang@intel.com, Cheng Jiang <cheng1.jiang@intel.com>
Subject: [PATCH] add support for async vhost packed ring dequeue
Date: Mon, 13 Jun 2022 08:21:59 +0000	[thread overview]
Message-ID: <20220613082159.40077-1-cheng1.jiang@intel.com> (raw)

This patch implements packed ring dequeue data path for asynchronous
vhost.

Signed-off-by: Cheng Jiang <cheng1.jiang@intel.com>
Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>
---
 lib/vhost/virtio_net.c | 218 ++++++++++++++++++++++++++++++++++++-----
 1 file changed, 192 insertions(+), 26 deletions(-)

diff --git a/lib/vhost/virtio_net.c b/lib/vhost/virtio_net.c
index 68a26eb17d..0d05f7cac3 100644
--- a/lib/vhost/virtio_net.c
+++ b/lib/vhost/virtio_net.c
@@ -3240,7 +3240,7 @@ rte_vhost_dequeue_burst(int vid, uint16_t queue_id,
 }

 static __rte_always_inline uint16_t
-async_poll_dequeue_completed_split(struct virtio_net *dev, struct vhost_virtqueue *vq,
+async_poll_dequeue_completed(struct virtio_net *dev, struct vhost_virtqueue *vq,
 		struct rte_mbuf **pkts, uint16_t count, int16_t dma_id,
 		uint16_t vchan_id, bool legacy_ol_flags)
 {
@@ -3255,7 +3255,7 @@ async_poll_dequeue_completed_split(struct virtio_net *dev, struct vhost_virtqueu
 	from = start_idx;
 	while (vq->async->pkts_cmpl_flag[from] && count--) {
 		vq->async->pkts_cmpl_flag[from] = false;
-		from = (from + 1) & (vq->size - 1);
+		from = (from + 1) % vq->size;
 		nr_cpl_pkts++;
 	}

@@ -3263,7 +3263,7 @@ async_poll_dequeue_completed_split(struct virtio_net *dev, struct vhost_virtqueu
 		return 0;

 	for (i = 0; i < nr_cpl_pkts; i++) {
-		from = (start_idx + i) & (vq->size - 1);
+		from = (start_idx + i) % vq->size;
 		pkts[i] = pkts_info[from].mbuf;

 		if (virtio_net_with_host_offload(dev))
@@ -3272,10 +3272,14 @@ async_poll_dequeue_completed_split(struct virtio_net *dev, struct vhost_virtqueu
 	}

 	/* write back completed descs to used ring and update used idx */
-	write_back_completed_descs_split(vq, nr_cpl_pkts);
-	__atomic_add_fetch(&vq->used->idx, nr_cpl_pkts, __ATOMIC_RELEASE);
-	vhost_vring_call_split(dev, vq);
-
+	if (vq_is_packed(dev)) {
+		write_back_completed_descs_packed(vq, nr_cpl_pkts);
+		vhost_vring_call_packed(dev, vq);
+	} else {
+		write_back_completed_descs_split(vq, nr_cpl_pkts);
+		__atomic_add_fetch(&vq->used->idx, nr_cpl_pkts, __ATOMIC_RELEASE);
+		vhost_vring_call_split(dev, vq);
+	}
 	vq->async->pkts_inflight_n -= nr_cpl_pkts;

 	return nr_cpl_pkts;
@@ -3412,8 +3416,8 @@ virtio_dev_tx_async_split(struct virtio_net *dev, struct vhost_virtqueue *vq,

 out:
 	/* DMA device may serve other queues, unconditionally check completed. */
-	nr_done_pkts = async_poll_dequeue_completed_split(dev, vq, pkts, pkts_size,
-							  dma_id, vchan_id, legacy_ol_flags);
+	nr_done_pkts = async_poll_dequeue_completed(dev, vq, pkts, pkts_size,
+							dma_id, vchan_id, legacy_ol_flags);

 	return nr_done_pkts;
 }
@@ -3440,6 +3444,171 @@ virtio_dev_tx_async_split_compliant(struct virtio_net *dev,
 				pkts, count, dma_id, vchan_id, false);
 }

+static __rte_always_inline void
+vhost_async_shadow_dequeue_single_packed(struct vhost_virtqueue *vq, uint16_t buf_id)
+{
+	struct vhost_async *async = vq->async;
+	uint16_t idx = async->buffer_idx_packed;
+
+	async->buffers_packed[idx].id = buf_id;
+	async->buffers_packed[idx].len = 0;
+	async->buffers_packed[idx].count = 1;
+
+	async->buffer_idx_packed++;
+	if (async->buffer_idx_packed >= vq->size)
+		async->buffer_idx_packed -= vq->size;
+
+}
+
+static __rte_always_inline int
+virtio_dev_tx_async_single_packed(struct virtio_net *dev,
+			struct vhost_virtqueue *vq,
+			struct rte_mempool *mbuf_pool,
+			struct rte_mbuf *pkts,
+			uint16_t slot_idx,
+			bool legacy_ol_flags)
+{
+	int err;
+	uint16_t buf_id, desc_count = 0;
+	uint16_t nr_vec = 0;
+	uint32_t buf_len;
+	struct buf_vector buf_vec[BUF_VECTOR_MAX];
+	static bool allocerr_warned;
+
+	if (unlikely(fill_vec_buf_packed(dev, vq, vq->last_avail_idx, &desc_count,
+					 buf_vec, &nr_vec, &buf_id, &buf_len,
+					 VHOST_ACCESS_RO) < 0))
+		return -1;
+
+	if (unlikely(virtio_dev_pktmbuf_prep(dev, pkts, buf_len))) {
+		if (!allocerr_warned) {
+			VHOST_LOG_DATA(ERR, "(%s) Failed mbuf alloc of size %d from %s.\n",
+				dev->ifname, buf_len, mbuf_pool->name);
+
+			allocerr_warned = true;
+		}
+		return -1;
+	}
+
+	err = desc_to_mbuf(dev, vq, buf_vec, nr_vec, pkts, mbuf_pool,
+		legacy_ol_flags, slot_idx, true);
+	if (unlikely(err)) {
+		rte_pktmbuf_free(pkts);
+		if (!allocerr_warned) {
+			VHOST_LOG_DATA(ERR, "(%s) Failed to copy desc to mbuf on.\n", dev->ifname);
+			allocerr_warned = true;
+		}
+		return -1;
+	}
+
+	/* update async shadow packed ring */
+	vhost_async_shadow_dequeue_single_packed(vq, buf_id);
+
+	return err;
+}
+
+static __rte_always_inline uint16_t
+virtio_dev_tx_async_packed(struct virtio_net *dev, struct vhost_virtqueue *vq,
+		struct rte_mempool *mbuf_pool, struct rte_mbuf **pkts,
+		uint16_t count, uint16_t dma_id, uint16_t vchan_id, bool legacy_ol_flags)
+{
+	uint16_t pkt_idx;
+	uint16_t slot_idx = 0;
+	uint16_t nr_done_pkts = 0;
+	uint16_t pkt_err = 0;
+	uint32_t n_xfer;
+	struct vhost_async *async = vq->async;
+	struct async_inflight_info *pkts_info = async->pkts_info;
+	struct rte_mbuf *pkts_prealloc[MAX_PKT_BURST];
+
+	VHOST_LOG_DATA(DEBUG, "(%d) about to dequeue %u buffers\n", dev->vid, count);
+
+	async_iter_reset(async);
+
+	if (rte_pktmbuf_alloc_bulk(mbuf_pool, pkts_prealloc, count))
+		goto out;
+
+	for (pkt_idx = 0; pkt_idx < count; pkt_idx++) {
+		struct rte_mbuf *pkt = pkts_prealloc[pkt_idx];
+
+		rte_prefetch0(&vq->desc_packed[vq->last_avail_idx]);
+
+		slot_idx = (async->pkts_idx + pkt_idx) % vq->size;
+		if (unlikely(virtio_dev_tx_async_single_packed(dev, vq, mbuf_pool, pkt,
+				slot_idx, legacy_ol_flags))) {
+			rte_pktmbuf_free_bulk(&pkts_prealloc[pkt_idx], count - pkt_idx);
+			break;
+		}
+
+		pkts_info[slot_idx].mbuf = pkt;
+
+		vq_inc_last_avail_packed(vq, 1);
+
+	}
+
+	n_xfer = vhost_async_dma_transfer(dev, vq, dma_id, vchan_id, async->pkts_idx,
+					async->iov_iter, pkt_idx);
+
+	async->pkts_inflight_n += n_xfer;
+
+	pkt_err = pkt_idx - n_xfer;
+
+	if (unlikely(pkt_err)) {
+		pkt_idx -= pkt_err;
+
+		/**
+		 * recover DMA-copy related structures and free pktmbuf for DMA-error pkts.
+		 */
+		if (async->buffer_idx_packed >= pkt_err)
+			async->buffer_idx_packed -= pkt_err;
+		else
+			async->buffer_idx_packed += vq->size - pkt_err;
+
+		while (pkt_err-- > 0) {
+			rte_pktmbuf_free(pkts_info[slot_idx % vq->size].mbuf);
+			slot_idx--;
+		}
+
+		/* recover available ring */
+		if (vq->last_avail_idx >= pkt_err) {
+			vq->last_avail_idx -= pkt_err;
+		} else {
+			vq->last_avail_idx += vq->size - pkt_err;
+			vq->avail_wrap_counter ^= 1;
+		}
+	}
+
+	async->pkts_idx += pkt_idx;
+	if (async->pkts_idx >= vq->size)
+		async->pkts_idx -= vq->size;
+
+out:
+	nr_done_pkts = async_poll_dequeue_completed(dev, vq, pkts, count,
+					dma_id, vchan_id, legacy_ol_flags);
+
+	return nr_done_pkts;
+}
+
+__rte_noinline
+static uint16_t
+virtio_dev_tx_async_packed_legacy(struct virtio_net *dev, struct vhost_virtqueue *vq,
+		struct rte_mempool *mbuf_pool, struct rte_mbuf **pkts,
+		uint16_t count, uint16_t dma_id, uint16_t vchan_id)
+{
+	return virtio_dev_tx_async_packed(dev, vq, mbuf_pool,
+				pkts, count, dma_id, vchan_id, true);
+}
+
+__rte_noinline
+static uint16_t
+virtio_dev_tx_async_packed_compliant(struct virtio_net *dev, struct vhost_virtqueue *vq,
+		struct rte_mempool *mbuf_pool, struct rte_mbuf **pkts,
+		uint16_t count, uint16_t dma_id, uint16_t vchan_id)
+{
+	return virtio_dev_tx_async_packed(dev, vq, mbuf_pool,
+				pkts, count, dma_id, vchan_id, false);
+}
+
 uint16_t
 rte_vhost_async_try_dequeue_burst(int vid, uint16_t queue_id,
 	struct rte_mempool *mbuf_pool, struct rte_mbuf **pkts, uint16_t count,
@@ -3542,25 +3711,22 @@ rte_vhost_async_try_dequeue_burst(int vid, uint16_t queue_id,
 		count -= 1;
 	}

-	if (unlikely(vq_is_packed(dev))) {
-		static bool not_support_pack_log;
-		if (!not_support_pack_log) {
-			VHOST_LOG_DATA(ERR,
-				"(%s) %s: async dequeue does not support packed ring.\n",
-				dev->ifname, __func__);
-			not_support_pack_log = true;
-		}
-		count = 0;
-		goto out;
+	if (vq_is_packed(dev)) {
+		if (dev->flags & VIRTIO_DEV_LEGACY_OL_FLAGS)
+			count = virtio_dev_tx_async_packed_legacy(dev, vq, mbuf_pool,
+					pkts, count, dma_id, vchan_id);
+		else
+			count = virtio_dev_tx_async_packed_compliant(dev, vq, mbuf_pool,
+					pkts, count, dma_id, vchan_id);
+	} else {
+		if (dev->flags & VIRTIO_DEV_LEGACY_OL_FLAGS)
+			count = virtio_dev_tx_async_split_legacy(dev, vq, mbuf_pool,
+					pkts, count, dma_id, vchan_id);
+		else
+			count = virtio_dev_tx_async_split_compliant(dev, vq, mbuf_pool,
+					pkts, count, dma_id, vchan_id);
 	}

-	if (dev->flags & VIRTIO_DEV_LEGACY_OL_FLAGS)
-		count = virtio_dev_tx_async_split_legacy(dev, vq, mbuf_pool, pkts,
-							 count, dma_id, vchan_id);
-	else
-		count = virtio_dev_tx_async_split_compliant(dev, vq, mbuf_pool, pkts,
-							    count, dma_id, vchan_id);
-
 	*nr_inflight = vq->async->pkts_inflight_n;

 out:
--
2.35.1


             reply	other threads:[~2022-06-13  8:57 UTC|newest]

Thread overview: 3+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2022-06-13  8:21 Cheng Jiang [this message]
2022-06-17 13:33 ` Maxime Coquelin
2022-06-20 12:33   ` Jiang, Cheng1

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20220613082159.40077-1-cheng1.jiang@intel.com \
    --to=cheng1.jiang@intel.com \
    --cc=chenbo.xia@intel.com \
    --cc=dev@dpdk.org \
    --cc=jiayu.hu@intel.com \
    --cc=maxime.coquelin@redhat.com \
    --cc=wenwux.ma@intel.com \
    --cc=xuan.ding@intel.com \
    --cc=yuanx.wang@intel.com \
    --cc=yvonnex.yang@intel.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).