Commit f19fadfc authored by willy tarreau's avatar willy tarreau Committed by David S. Miller
Browse files

net: mvneta: implement rx_copybreak 



calling dma_map_single()/dma_unmap_single() is quite expensive compared
to copying a small packet. So let's copy short frames and keep the buffers
mapped. We set the limit to 256 bytes which seems to give good results both
on the XP-GP board and on the AX3/4.

The Rx small packet rate increased by 16.4% doing this, from 486kpps to
573kpps. It is worth noting that even the call to the function
dma_sync_single_range_for_cpu() is expensive (300 ns) although less
than dma_unmap_single(). Without it, the packet rate raises to 711kpps
(+24% more). Thus on systems where coherency from device to CPU is
guaranteed by a snoop control unit, this patch should provide even more
gains, and probably rx_copybreak could be increased.

Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Cc: Gregory CLEMENT <gregory.clement@free-electrons.com>
Tested-by: default avatarArnaud Ebalard <arno@natisbad.org>
Signed-off-by: default avatarWilly Tarreau <w@1wt.eu>
Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
parent 8ec2cd48

drivers/net/ethernet/marvell/mvneta.c

+38 −6
Original line number Diff line number Diff line
@@ -444,6 +444,8 @@ static int txq_number = 8; 


static int rxq_def;



static int rx_copybreak __read_mostly = 256;



#define MVNETA_DRIVER_NAME "mvneta"

#define MVNETA_DRIVER_VERSION "1.0"
@@ -1463,22 +1465,51 @@ static int mvneta_rx(struct mvneta_port *pp, int rx_todo, 
		rx_done++;

		rx_filled++;

		rx_status = rx_desc->status;

		rx_bytes = rx_desc->data_size - (ETH_FCS_LEN + MVNETA_MH_SIZE);

		data = (unsigned char *)rx_desc->buf_cookie;



		if (!mvneta_rxq_desc_is_first_last(rx_status) ||

		    (rx_status & MVNETA_RXD_ERR_SUMMARY) ||

		    !(skb = build_skb(data, pp->frag_size > PAGE_SIZE ? 0 : pp->frag_size))) {

		    (rx_status & MVNETA_RXD_ERR_SUMMARY)) {

		err_drop_frame:

			dev->stats.rx_errors++;

			mvneta_rx_error(pp, rx_desc);

			/* leave the descriptor untouched */

			continue;

		}



		dma_unmap_single(pp->dev->dev.parent, rx_desc->buf_phys_addr,

		if (rx_bytes <= rx_copybreak) {

			/* better copy a small frame and not unmap the DMA region */

			skb = netdev_alloc_skb_ip_align(dev, rx_bytes);

			if (unlikely(!skb))

				goto err_drop_frame;



			dma_sync_single_range_for_cpu(dev->dev.parent,

			                              rx_desc->buf_phys_addr,

			                              MVNETA_MH_SIZE + NET_SKB_PAD,

			                              rx_bytes,

			                              DMA_FROM_DEVICE);

			memcpy(skb_put(skb, rx_bytes),

			       data + MVNETA_MH_SIZE + NET_SKB_PAD,

			       rx_bytes);



			skb->protocol = eth_type_trans(skb, dev);

			mvneta_rx_csum(pp, rx_status, skb);

			napi_gro_receive(&pp->napi, skb);



			rcvd_pkts++;

			rcvd_bytes += rx_bytes;



			/* leave the descriptor and buffer untouched */

			continue;

		}



		skb = build_skb(data, pp->frag_size > PAGE_SIZE ? 0 : pp->frag_size);

		if (!skb)

			goto err_drop_frame;



		dma_unmap_single(dev->dev.parent, rx_desc->buf_phys_addr,

				 MVNETA_RX_BUF_SIZE(pp->pkt_size), DMA_FROM_DEVICE);



		rx_bytes = rx_desc->data_size -

			(ETH_FCS_LEN + MVNETA_MH_SIZE);

		rcvd_pkts++;

		rcvd_bytes += rx_bytes;
@@ -1495,7 +1526,7 @@ static int mvneta_rx(struct mvneta_port *pp, int rx_todo, 
		/* Refill processing */

		err = mvneta_rx_refill(pp, rx_desc);

		if (err) {

			netdev_err(pp->dev, "Linux processing - Can't refill\n");

			netdev_err(dev, "Linux processing - Can't refill\n");

			rxq->missed++;

			rx_filled--;

		}
@@ -2945,3 +2976,4 @@ module_param(rxq_number, int, S_IRUGO); 
module_param(txq_number, int, S_IRUGO);



module_param(rxq_def, int, S_IRUGO);

module_param(rx_copybreak, int, S_IRUGO | S_IWUSR);

CRA Git | Maintained and supported by SUSTech CRA and CCSE