[2/3] net/virtio: merge Rx mergeable and non-mergeable paths
Checks
Commit Message
Signed-off-by: Maxime Coquelin <maxime.coquelin@redhat.com>
---
drivers/net/virtio/virtio_ethdev.c | 5 --
drivers/net/virtio/virtio_ethdev.h | 3 -
drivers/net/virtio/virtio_rxtx.c | 115 ++---------------------------
3 files changed, 7 insertions(+), 116 deletions(-)
Comments
On 12/3/18 4:10 PM, Maxime Coquelin wrote:
> Signed-off-by: Maxime Coquelin <maxime.coquelin@redhat.com>
> ---
> drivers/net/virtio/virtio_ethdev.c | 5 --
> drivers/net/virtio/virtio_ethdev.h | 3 -
> drivers/net/virtio/virtio_rxtx.c | 115 ++---------------------------
> 3 files changed, 7 insertions(+), 116 deletions(-)
Intel STV team ran more tests and found a performance regression.
I managed to reproduce it, and worked on optimizing the Rx path.
I managed to go from -25% to -6% for this use case, but this is still a
significant regression so I'll drop this patch.
The good news is that the optimization is valid for mergeable buffers
case, and it gains 5% so I'll post this optimization instead.
Thanks to the STV team for the testing,
Maxime
@@ -1335,11 +1335,6 @@ set_rxtx_funcs(struct rte_eth_dev *eth_dev)
"virtio: using inorder mergeable buffer Rx path on port %u",
eth_dev->data->port_id);
eth_dev->rx_pkt_burst = &virtio_recv_mergeable_pkts_inorder;
- } else if (vtpci_with_feature(hw, VIRTIO_NET_F_MRG_RXBUF)) {
- PMD_INIT_LOG(INFO,
- "virtio: using mergeable buffer Rx path on port %u",
- eth_dev->data->port_id);
- eth_dev->rx_pkt_burst = &virtio_recv_mergeable_pkts;
} else {
PMD_INIT_LOG(INFO, "virtio: using standard Rx path on port %u",
eth_dev->data->port_id);
@@ -74,9 +74,6 @@ int virtio_dev_tx_queue_setup_finish(struct rte_eth_dev *dev,
uint16_t virtio_recv_pkts(void *rx_queue, struct rte_mbuf **rx_pkts,
uint16_t nb_pkts);
-uint16_t virtio_recv_mergeable_pkts(void *rx_queue, struct rte_mbuf **rx_pkts,
- uint16_t nb_pkts);
-
uint16_t virtio_recv_mergeable_pkts_inorder(void *rx_queue,
struct rte_mbuf **rx_pkts, uint16_t nb_pkts);
@@ -883,111 +883,6 @@ virtio_rx_offload(struct rte_mbuf *m, struct virtio_net_hdr *hdr)
}
#define VIRTIO_MBUF_BURST_SZ 64
-#define DESC_PER_CACHELINE (RTE_CACHE_LINE_SIZE / sizeof(struct vring_desc))
-uint16_t
-virtio_recv_pkts(void *rx_queue, struct rte_mbuf **rx_pkts, uint16_t nb_pkts)
-{
- struct virtnet_rx *rxvq = rx_queue;
- struct virtqueue *vq = rxvq->vq;
- struct virtio_hw *hw = vq->hw;
- struct rte_mbuf *rxm, *new_mbuf;
- uint16_t nb_used, num, nb_rx;
- uint32_t len[VIRTIO_MBUF_BURST_SZ];
- struct rte_mbuf *rcv_pkts[VIRTIO_MBUF_BURST_SZ];
- int error;
- uint32_t i, nb_enqueued;
- uint32_t hdr_size;
- struct virtio_net_hdr *hdr;
-
- nb_rx = 0;
- if (unlikely(hw->started == 0))
- return nb_rx;
-
- nb_used = VIRTQUEUE_NUSED(vq);
-
- virtio_rmb();
-
- num = likely(nb_used <= nb_pkts) ? nb_used : nb_pkts;
- if (unlikely(num > VIRTIO_MBUF_BURST_SZ))
- num = VIRTIO_MBUF_BURST_SZ;
- if (likely(num > DESC_PER_CACHELINE))
- num = num - ((vq->vq_used_cons_idx + num) % DESC_PER_CACHELINE);
-
- num = virtqueue_dequeue_burst_rx(vq, rcv_pkts, len, num);
- PMD_RX_LOG(DEBUG, "used:%d dequeue:%d", nb_used, num);
-
- nb_enqueued = 0;
- hdr_size = hw->vtnet_hdr_size;
-
- for (i = 0; i < num ; i++) {
- rxm = rcv_pkts[i];
-
- PMD_RX_LOG(DEBUG, "packet len:%d", len[i]);
-
- if (unlikely(len[i] < hdr_size + ETHER_HDR_LEN)) {
- PMD_RX_LOG(ERR, "Packet drop");
- nb_enqueued++;
- virtio_discard_rxbuf(vq, rxm);
- rxvq->stats.errors++;
- continue;
- }
-
- rxm->port = rxvq->port_id;
- rxm->data_off = RTE_PKTMBUF_HEADROOM;
- rxm->ol_flags = 0;
- rxm->vlan_tci = 0;
-
- rxm->pkt_len = (uint32_t)(len[i] - hdr_size);
- rxm->data_len = (uint16_t)(len[i] - hdr_size);
-
- hdr = (struct virtio_net_hdr *)((char *)rxm->buf_addr +
- RTE_PKTMBUF_HEADROOM - hdr_size);
-
- if (hw->vlan_strip)
- rte_vlan_strip(rxm);
-
- if (hw->has_rx_offload && virtio_rx_offload(rxm, hdr) < 0) {
- virtio_discard_rxbuf(vq, rxm);
- rxvq->stats.errors++;
- continue;
- }
-
- virtio_rx_stats_updated(rxvq, rxm);
-
- rx_pkts[nb_rx++] = rxm;
- }
-
- rxvq->stats.packets += nb_rx;
-
- /* Allocate new mbuf for the used descriptor */
- while (likely(!virtqueue_full(vq))) {
- new_mbuf = rte_mbuf_raw_alloc(rxvq->mpool);
- if (unlikely(new_mbuf == NULL)) {
- struct rte_eth_dev *dev
- = &rte_eth_devices[rxvq->port_id];
- dev->data->rx_mbuf_alloc_failed++;
- break;
- }
- error = virtqueue_enqueue_recv_refill(vq, new_mbuf);
- if (unlikely(error)) {
- rte_pktmbuf_free(new_mbuf);
- break;
- }
- nb_enqueued++;
- }
-
- if (likely(nb_enqueued)) {
- vq_update_avail_idx(vq);
-
- if (unlikely(virtqueue_kick_prepare(vq))) {
- virtqueue_notify(vq);
- PMD_RX_LOG(DEBUG, "Notified");
- }
- }
-
- return nb_rx;
-}
-
uint16_t
virtio_recv_mergeable_pkts_inorder(void *rx_queue,
struct rte_mbuf **rx_pkts,
@@ -1176,7 +1071,7 @@ virtio_recv_mergeable_pkts_inorder(void *rx_queue,
}
uint16_t
-virtio_recv_mergeable_pkts(void *rx_queue,
+virtio_recv_pkts(void *rx_queue,
struct rte_mbuf **rx_pkts,
uint16_t nb_pkts)
{
@@ -1239,10 +1134,14 @@ virtio_recv_mergeable_pkts(void *rx_queue,
header = (struct virtio_net_hdr_mrg_rxbuf *)((char *)rxm->buf_addr +
RTE_PKTMBUF_HEADROOM - hdr_size);
- seg_num = header->num_buffers;
- if (seg_num == 0)
+ if (vtpci_with_feature(hw, VIRTIO_NET_F_MRG_RXBUF)) {
+ seg_num = header->num_buffers;
+ if (seg_num == 0)
+ seg_num = 1;
+ } else {
seg_num = 1;
+ }
rxm->data_off = RTE_PKTMBUF_HEADROOM;
rxm->nb_segs = seg_num;