]> git.droids-corp.org - dpdk.git/commitdiff
vhost: simplify async IO vectors
authorMaxime Coquelin <maxime.coquelin@redhat.com>
Tue, 26 Oct 2021 16:28:52 +0000 (18:28 +0200)
committerMaxime Coquelin <maxime.coquelin@redhat.com>
Fri, 29 Oct 2021 10:32:30 +0000 (12:32 +0200)
IO vectors implementation is unnecessarily complex, mixing
source and destinations vectors in the same array.

This patch declares two arrays, one for the source and one
for the destination. It also gets rid of seg_awaits variable
in both packed and split implementation, which is the same
as iovec_idx.

Signed-off-by: Maxime Coquelin <maxime.coquelin@redhat.com>
Reviewed-by: Chenbo Xia <chenbo.xia@intel.com>
Reviewed-by: Jiayu Hu <jiayu.hu@intel.com>
lib/vhost/vhost.h
lib/vhost/virtio_net.c

index fa92a0d33d800c0db850460565c4ad2e02b9bfd8..b0bad495e1efbd987da70aa64cdc707b2827b27f 100644 (file)
@@ -49,7 +49,7 @@
 #define MAX_PKT_BURST 32
 
 #define VHOST_MAX_ASYNC_IT (MAX_PKT_BURST * 2)
-#define VHOST_MAX_ASYNC_VEC (BUF_VECTOR_MAX * 4)
+#define VHOST_MAX_ASYNC_VEC (BUF_VECTOR_MAX * 2)
 
 #define PACKED_DESC_ENQUEUE_USED_FLAG(w)       \
        ((w) ? (VRING_DESC_F_AVAIL | VRING_DESC_F_USED | VRING_DESC_F_WRITE) : \
@@ -133,7 +133,8 @@ struct vhost_async {
        struct rte_vhost_async_channel_ops ops;
 
        struct rte_vhost_iov_iter it_pool[VHOST_MAX_ASYNC_IT];
-       struct iovec vec_pool[VHOST_MAX_ASYNC_VEC];
+       struct iovec src_iovec[VHOST_MAX_ASYNC_VEC];
+       struct iovec dst_iovec[VHOST_MAX_ASYNC_VEC];
 
        /* data transfer status */
        struct async_inflight_info *pkts_info;
index d9ac85f82979578320698d05676e78d935449c46..2a243701c06c0dc91318f55457c1e37ebe8923ed 100644 (file)
@@ -1512,14 +1512,12 @@ virtio_dev_rx_async_submit_split(struct virtio_net *dev,
 
        struct vhost_async *async = vq->async;
        struct rte_vhost_iov_iter *it_pool = async->it_pool;
-       struct iovec *vec_pool = async->vec_pool;
        struct rte_vhost_async_desc tdes[MAX_PKT_BURST];
-       struct iovec *src_iovec = vec_pool;
-       struct iovec *dst_iovec = vec_pool + (VHOST_MAX_ASYNC_VEC >> 1);
+       struct iovec *src_iovec = async->src_iovec;
+       struct iovec *dst_iovec = async->dst_iovec;
        struct async_inflight_info *pkts_info = async->pkts_info;
        uint32_t n_pkts = 0, pkt_err = 0;
        int32_t n_xfer;
-       uint16_t segs_await = 0;
        uint16_t iovec_idx = 0, it_idx = 0, slot_idx = 0;
 
        /*
@@ -1562,7 +1560,6 @@ virtio_dev_rx_async_submit_split(struct virtio_net *dev,
                pkts_info[slot_idx].mbuf = pkts[pkt_idx];
 
                iovec_idx += it_pool[it_idx].nr_segs;
-               segs_await += it_pool[it_idx].nr_segs;
                it_idx += 2;
 
                vq->last_avail_idx += num_buffers;
@@ -1573,8 +1570,7 @@ virtio_dev_rx_async_submit_split(struct virtio_net *dev,
                 * - unused async iov number is less than max vhost vector
                 */
                if (unlikely(pkt_burst_idx >= VHOST_ASYNC_BATCH_THRESHOLD ||
-                       ((VHOST_MAX_ASYNC_VEC >> 1) - segs_await <
-                       BUF_VECTOR_MAX))) {
+                       (VHOST_MAX_ASYNC_VEC - iovec_idx < BUF_VECTOR_MAX))) {
                        n_xfer = async->ops.transfer_data(dev->vid,
                                        queue_id, tdes, 0, pkt_burst_idx);
                        if (likely(n_xfer >= 0)) {
@@ -1588,7 +1584,6 @@ virtio_dev_rx_async_submit_split(struct virtio_net *dev,
 
                        iovec_idx = 0;
                        it_idx = 0;
-                       segs_await = 0;
 
                        if (unlikely(n_pkts < pkt_burst_idx)) {
                                /*
@@ -1745,8 +1740,11 @@ vhost_enqueue_async_packed(struct virtio_net *dev,
                if (unlikely(++tries > max_tries))
                        return -1;
 
-               if (unlikely(fill_vec_buf_packed(dev, vq, avail_idx, &desc_count, buf_vec, &nr_vec,
-                                               &buf_id, &len, VHOST_ACCESS_RW) < 0))
+               if (unlikely(fill_vec_buf_packed(dev, vq,
+                                               avail_idx, &desc_count,
+                                               buf_vec, &nr_vec,
+                                               &buf_id, &len,
+                                               VHOST_ACCESS_RW) < 0))
                        return -1;
 
                len = RTE_MIN(len, size);
@@ -1832,14 +1830,12 @@ virtio_dev_rx_async_submit_packed(struct virtio_net *dev,
 
        struct vhost_async *async = vq->async;
        struct rte_vhost_iov_iter *it_pool = async->it_pool;
-       struct iovec *vec_pool = async->vec_pool;
        struct rte_vhost_async_desc tdes[MAX_PKT_BURST];
-       struct iovec *src_iovec = vec_pool;
-       struct iovec *dst_iovec = vec_pool + (VHOST_MAX_ASYNC_VEC >> 1);
+       struct iovec *src_iovec = async->src_iovec;
+       struct iovec *dst_iovec = async->dst_iovec;
        struct async_inflight_info *pkts_info = async->pkts_info;
        uint32_t n_pkts = 0, pkt_err = 0;
        uint16_t slot_idx = 0;
-       uint16_t segs_await = 0;
        uint16_t iovec_idx = 0, it_idx = 0;
 
        do {
@@ -1861,7 +1857,6 @@ virtio_dev_rx_async_submit_packed(struct virtio_net *dev,
                pkts_info[slot_idx].nr_buffers = num_buffers;
                pkts_info[slot_idx].mbuf = pkts[pkt_idx];
                iovec_idx += it_pool[it_idx].nr_segs;
-               segs_await += it_pool[it_idx].nr_segs;
                it_idx += 2;
 
                pkt_idx++;
@@ -1874,7 +1869,7 @@ virtio_dev_rx_async_submit_packed(struct virtio_net *dev,
                 * - unused async iov number is less than max vhost vector
                 */
                if (unlikely(pkt_burst_idx >= VHOST_ASYNC_BATCH_THRESHOLD ||
-                       ((VHOST_MAX_ASYNC_VEC >> 1) - segs_await < BUF_VECTOR_MAX))) {
+                       (VHOST_MAX_ASYNC_VEC - iovec_idx < BUF_VECTOR_MAX))) {
                        n_xfer = async->ops.transfer_data(dev->vid,
                                        queue_id, tdes, 0, pkt_burst_idx);
                        if (likely(n_xfer >= 0)) {
@@ -1888,7 +1883,6 @@ virtio_dev_rx_async_submit_packed(struct virtio_net *dev,
 
                        iovec_idx = 0;
                        it_idx = 0;
-                       segs_await = 0;
 
                        if (unlikely(n_pkts < pkt_burst_idx)) {
                                /*