eventdev: move Rx adapter to separate function
authorNikhil Rao <nikhil.rao@intel.com>
Mon, 2 Jul 2018 09:11:12 +0000 (14:41 +0530)
committerJerin Jacob <jerin.jacob@caviumnetworks.com>
Fri, 6 Jul 2018 04:54:49 +0000 (06:54 +0200)
Create a separate function that handles eth receive and
enqueue to event buffer. This function will also be called for
interrupt driven receive queues.

Signed-off-by: Nikhil Rao <nikhil.rao@intel.com>
Acked-by: Jerin Jacob <jerin.jacob@caviumnetworks.com>
lib/librte_eventdev/rte_event_eth_rx_adapter.c

index 926f83a..8fe037f 100644 (file)
@@ -616,6 +616,45 @@ rxa_buffer_mbufs(struct rte_event_eth_rx_adapter *rx_adapter,
        }
 }
 
+/* Enqueue packets from  <port, q>  to event buffer */
+static inline uint32_t
+rxa_eth_rx(struct rte_event_eth_rx_adapter *rx_adapter,
+       uint16_t port_id,
+       uint16_t queue_id,
+       uint32_t rx_count,
+       uint32_t max_rx)
+{
+       struct rte_mbuf *mbufs[BATCH_SIZE];
+       struct rte_eth_event_enqueue_buffer *buf =
+                                       &rx_adapter->event_enqueue_buffer;
+       struct rte_event_eth_rx_adapter_stats *stats =
+                                       &rx_adapter->stats;
+       uint16_t n;
+       uint32_t nb_rx = 0;
+
+       /* Don't do a batch dequeue from the rx queue if there isn't
+        * enough space in the enqueue buffer.
+        */
+       while (BATCH_SIZE <= (RTE_DIM(buf->events) - buf->count)) {
+               if (buf->count >= BATCH_SIZE)
+                       rxa_flush_event_buffer(rx_adapter);
+
+               stats->rx_poll_count++;
+               n = rte_eth_rx_burst(port_id, queue_id, mbufs, BATCH_SIZE);
+               if (unlikely(!n))
+                       break;
+               rxa_buffer_mbufs(rx_adapter, port_id, queue_id, mbufs, n);
+               nb_rx += n;
+               if (rx_count + nb_rx > max_rx)
+                       break;
+       }
+
+       if (buf->count >= BATCH_SIZE)
+               rxa_flush_event_buffer(rx_adapter);
+
+       return nb_rx;
+}
+
 /*
  * Polls receive queues added to the event adapter and enqueues received
  * packets to the event device.
@@ -633,17 +672,16 @@ static inline void
 rxa_poll(struct rte_event_eth_rx_adapter *rx_adapter)
 {
        uint32_t num_queue;
-       uint16_t n;
        uint32_t nb_rx = 0;
-       struct rte_mbuf *mbufs[BATCH_SIZE];
        struct rte_eth_event_enqueue_buffer *buf;
        uint32_t wrr_pos;
        uint32_t max_nb_rx;
+       struct rte_event_eth_rx_adapter_stats *stats;
 
        wrr_pos = rx_adapter->wrr_pos;
        max_nb_rx = rx_adapter->max_nb_rx;
        buf = &rx_adapter->event_enqueue_buffer;
-       struct rte_event_eth_rx_adapter_stats *stats = &rx_adapter->stats;
+       stats = &rx_adapter->stats;
 
        /* Iterate through a WRR sequence */
        for (num_queue = 0; num_queue < rx_adapter->wrr_len; num_queue++) {
@@ -658,32 +696,21 @@ rxa_poll(struct rte_event_eth_rx_adapter *rx_adapter)
                        rxa_flush_event_buffer(rx_adapter);
                if (BATCH_SIZE > (ETH_EVENT_BUFFER_SIZE - buf->count)) {
                        rx_adapter->wrr_pos = wrr_pos;
-                       return;
+                       break;
                }
 
-               stats->rx_poll_count++;
-               n = rte_eth_rx_burst(d, qid, mbufs, BATCH_SIZE);
-
-               if (n) {
-                       stats->rx_packets += n;
-                       /* The check before rte_eth_rx_burst() ensures that
-                        * all n mbufs can be buffered
-                        */
-                       rxa_buffer_mbufs(rx_adapter, d, qid, mbufs, n);
-                       nb_rx += n;
-                       if (nb_rx > max_nb_rx) {
-                               rx_adapter->wrr_pos =
+               nb_rx += rxa_eth_rx(rx_adapter, d, qid, nb_rx, max_nb_rx);
+               if (nb_rx > max_nb_rx) {
+                       rx_adapter->wrr_pos =
                                    (wrr_pos + 1) % rx_adapter->wrr_len;
-                               break;
-                       }
+                       break;
                }
 
                if (++wrr_pos == rx_adapter->wrr_len)
                        wrr_pos = 0;
        }
 
-       if (buf->count >= BATCH_SIZE)
-               rxa_flush_event_buffer(rx_adapter);
+       stats->rx_packets += nb_rx;
 }
 
 static int