net: mvneta: convert to build_skb()

Make use of build_skb() to allocate frags on the RX path. When frag size
is lower than a page size, we can use netdev_alloc_frag(), and we fall back
to kmalloc() for larger sizes. The frag size is stored into the mvneta_port
struct. The alloc/free functions check the frag size to decide what alloc/
free method to use. MTU changes are safe because the MTU change function
stops the device and clears the queues before applying the change.

With this patch, I observed a reproducible 2% performance improvement on
HTTP-based benchmarks, and 5% on small packet RX rate.

Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Cc: Gregory CLEMENT <gregory.clement@free-electrons.com>
Tested-by: Arnaud Ebalard <arno@natisbad.org>
Signed-off-by: Willy Tarreau <w@1wt.eu>
Signed-off-by: David S. Miller <davem@davemloft.net>

diff --git a/drivers/net/ethernet/marvell/mvneta.c b/drivers/net/ethernet/marvell/mvneta.c
index c7b37e0..726a8d2 100644 (file)
--- a/drivers/net/ethernet/marvell/mvneta.c
+++ b/drivers/net/ethernet/marvell/mvneta.c
@@ -268,6 +268,7 @@ struct mvneta_pcpu_stats {
 
 struct mvneta_port {
        int pkt_size;
+       unsigned int frag_size;
        void __iomem *base;
        struct mvneta_rx_queue *rxqs;
        struct mvneta_tx_queue *txqs;
@@ -1332,28 +1333,43 @@ static int mvneta_txq_done(struct mvneta_port *pp,
        return tx_done;
 }
 
+static void *mvneta_frag_alloc(const struct mvneta_port *pp)
+{
+       if (likely(pp->frag_size <= PAGE_SIZE))
+               return netdev_alloc_frag(pp->frag_size);
+       else
+               return kmalloc(pp->frag_size, GFP_ATOMIC);
+}
+
+static void mvneta_frag_free(const struct mvneta_port *pp, void *data)
+{
+       if (likely(pp->frag_size <= PAGE_SIZE))
+               put_page(virt_to_head_page(data));
+       else
+               kfree(data);
+}
+
 /* Refill processing */
 static int mvneta_rx_refill(struct mvneta_port *pp,
                            struct mvneta_rx_desc *rx_desc)
 
 {
        dma_addr_t phys_addr;
-       struct sk_buff *skb;
+       void *data;
 
-       skb = netdev_alloc_skb(pp->dev, pp->pkt_size);
-       if (!skb)
+       data = mvneta_frag_alloc(pp);
+       if (!data)
                return -ENOMEM;
 
-       phys_addr = dma_map_single(pp->dev->dev.parent, skb->head,
+       phys_addr = dma_map_single(pp->dev->dev.parent, data,
                                   MVNETA_RX_BUF_SIZE(pp->pkt_size),
                                   DMA_FROM_DEVICE);
        if (unlikely(dma_mapping_error(pp->dev->dev.parent, phys_addr))) {
-               dev_kfree_skb(skb);
+               mvneta_frag_free(pp, data);
                return -ENOMEM;
        }
 
-       mvneta_rx_desc_fill(rx_desc, phys_addr, (u32)skb);
-
+       mvneta_rx_desc_fill(rx_desc, phys_addr, (u32)data);
        return 0;
 }
 
@@ -1407,9 +1423,9 @@ static void mvneta_rxq_drop_pkts(struct mvneta_port *pp,
        rx_done = mvneta_rxq_busy_desc_num_get(pp, rxq);
        for (i = 0; i < rxq->size; i++) {
                struct mvneta_rx_desc *rx_desc = rxq->descs + i;
-               struct sk_buff *skb = (struct sk_buff *)rx_desc->buf_cookie;
+               void *data = (void *)rx_desc->buf_cookie;
 
-               dev_kfree_skb_any(skb);
+               mvneta_frag_free(pp, data);
                dma_unmap_single(pp->dev->dev.parent, rx_desc->buf_phys_addr,
                                 MVNETA_RX_BUF_SIZE(pp->pkt_size), DMA_FROM_DEVICE);
        }
@@ -1440,20 +1456,21 @@ static int mvneta_rx(struct mvneta_port *pp, int rx_todo,
        while (rx_done < rx_todo) {
                struct mvneta_rx_desc *rx_desc = mvneta_rxq_next_desc_get(rxq);
                struct sk_buff *skb;
+               unsigned char *data;
                u32 rx_status;
                int rx_bytes, err;
 
                rx_done++;
                rx_filled++;
                rx_status = rx_desc->status;
-               skb = (struct sk_buff *)rx_desc->buf_cookie;
+               data = (unsigned char *)rx_desc->buf_cookie;
 
                if (!mvneta_rxq_desc_is_first_last(rx_status) ||
-                   (rx_status & MVNETA_RXD_ERR_SUMMARY)) {
+                   (rx_status & MVNETA_RXD_ERR_SUMMARY) ||
+                   !(skb = build_skb(data, pp->frag_size > PAGE_SIZE ? 0 : pp->frag_size))) {
                        dev->stats.rx_errors++;
                        mvneta_rx_error(pp, rx_desc);
-                       mvneta_rx_desc_fill(rx_desc, rx_desc->buf_phys_addr,
-                                           (u32)skb);
+                       /* leave the descriptor untouched */
                        continue;
                }
 
@@ -1466,7 +1483,7 @@ static int mvneta_rx(struct mvneta_port *pp, int rx_todo,
                rcvd_bytes += rx_bytes;
 
                /* Linux processing */
-               skb_reserve(skb, MVNETA_MH_SIZE);
+               skb_reserve(skb, MVNETA_MH_SIZE + NET_SKB_PAD);
                skb_put(skb, rx_bytes);
 
                skb->protocol = eth_type_trans(skb, dev);
@@ -2276,6 +2293,8 @@ static int mvneta_change_mtu(struct net_device *dev, int mtu)
        mvneta_cleanup_rxqs(pp);
 
        pp->pkt_size = MVNETA_RX_PKT_SIZE(pp->dev->mtu);
+       pp->frag_size = SKB_DATA_ALIGN(MVNETA_RX_BUF_SIZE(pp->pkt_size)) +
+                       SKB_DATA_ALIGN(sizeof(struct skb_shared_info));
 
        ret = mvneta_setup_rxqs(pp);
        if (ret) {
@@ -2423,6 +2442,8 @@ static int mvneta_open(struct net_device *dev)
        mvneta_mac_addr_set(pp, dev->dev_addr, rxq_def);
 
        pp->pkt_size = MVNETA_RX_PKT_SIZE(pp->dev->mtu);
+       pp->frag_size = SKB_DATA_ALIGN(MVNETA_RX_BUF_SIZE(pp->pkt_size)) +
+                       SKB_DATA_ALIGN(sizeof(struct skb_shared_info));
 
        ret = mvneta_setup_rxqs(pp);
        if (ret)