bnx2x: uses build_skb() in receive path

bnx2x uses following formula to compute its rx_buf_sz :

dev->mtu + 2*L1_CACHE_BYTES + 14 + 8 + 8 + 2

Then core network adds NET_SKB_PAD and SKB_DATA_ALIGN(sizeof(struct
skb_shared_info))

Final allocated size for skb head on x86_64 (L1_CACHE_BYTES = 64,
MTU=1500) : 2112 bytes : SLUB/SLAB round this to 4096 bytes.

Since skb truesize is then bigger than SK_MEM_QUANTUM, we have lot of
false sharing because of mem_reclaim in UDP stack.

One possible way to half truesize is to reduce the need by 64 bytes
(2112 -> 2048 bytes)

Instead of allocating a full cache line at the end of packet for
alignment, we can use the fact that skb_shared_info sits at the end of
skb->head, and we can use this room, if we convert bnx2x to new
build_skb() infrastructure.

skb_shared_info will be initialized after hardware finished its
transfert, so we can eventually overwrite the final padding.

Using build_skb() also reduces cache line misses in the driver, since we
use cache hot skb instead of cold ones. Number of in-flight sk_buff
structures is lower, they are recycled while still hot.

Performance results :

(820.000 pps on a rx UDP monothread benchmark, instead of 720.000 pps)

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
CC: Eilon Greenstein <eilong@broadcom.com>
CC: Ben Hutchings <bhutchings@solarflare.com>
CC: Tom Herbert <therbert@google.com>
CC: Jamal Hadi Salim <hadi@mojatatu.com>
CC: Stephen Hemminger <shemminger@vyatta.com>
CC: Thomas Graf <tgraf@infradead.org>
CC: Herbert Xu <herbert@gondor.apana.org.au>
CC: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Acked-by: Eilon Greenstein <eilong@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

1 files changed, 17 insertions, 16 deletions

diff --git a/drivers/net/ethernet/broadcom/bnx2x/bnx2x_cmn.h b/drivers/net/ethernet/broadcom/bnx2x/bnx2x_cmn.h
index 260226d09916..41eb17e7720f 100644
--- a/drivers/net/ethernet/broadcom/bnx2x/bnx2x_cmn.h
+++ b/drivers/net/ethernet/broadcom/bnx2x/bnx2x_cmn.h
@@ -910,26 +910,27 @@ static inline int bnx2x_alloc_rx_sge(struct bnx2x *bp,
 	return 0;
 }
 
-static inline int bnx2x_alloc_rx_skb(struct bnx2x *bp,
-				     struct bnx2x_fastpath *fp, u16 index)
+static inline int bnx2x_alloc_rx_data(struct bnx2x *bp,
+				      struct bnx2x_fastpath *fp, u16 index)
 {
-	struct sk_buff *skb;
+	u8 *data;
 	struct sw_rx_bd *rx_buf = &fp->rx_buf_ring[index];
 	struct eth_rx_bd *rx_bd = &fp->rx_desc_ring[index];
 	dma_addr_t mapping;
 
-	skb = netdev_alloc_skb(bp->dev, fp->rx_buf_size);
-	if (unlikely(skb == NULL))
+	data = kmalloc(fp->rx_buf_size + NET_SKB_PAD, GFP_ATOMIC);
+	if (unlikely(data == NULL))
 		return -ENOMEM;
 
-	mapping = dma_map_single(&bp->pdev->dev, skb->data, fp->rx_buf_size,
+	mapping = dma_map_single(&bp->pdev->dev, data + NET_SKB_PAD,
+				 fp->rx_buf_size,
 				 DMA_FROM_DEVICE);
 	if (unlikely(dma_mapping_error(&bp->pdev->dev, mapping))) {
-		dev_kfree_skb_any(skb);
+		kfree(data);
 		return -ENOMEM;
 	}
 
-	rx_buf->skb = skb;
+	rx_buf->data = data;
 	dma_unmap_addr_set(rx_buf, mapping, mapping);
 
 	rx_bd->addr_hi = cpu_to_le32(U64_HI(mapping));
@@ -938,12 +939,12 @@ static inline int bnx2x_alloc_rx_skb(struct bnx2x *bp,
 	return 0;
 }
 
-/* note that we are not allocating a new skb,
+/* note that we are not allocating a new buffer,
  * we are just moving one from cons to prod
  * we are not creating a new mapping,
  * so there is no need to check for dma_mapping_error().
  */
-static inline void bnx2x_reuse_rx_skb(struct bnx2x_fastpath *fp,
+static inline void bnx2x_reuse_rx_data(struct bnx2x_fastpath *fp,
 				      u16 cons, u16 prod)
 {
 	struct sw_rx_bd *cons_rx_buf = &fp->rx_buf_ring[cons];
@@ -953,7 +954,7 @@ static inline void bnx2x_reuse_rx_skb(struct bnx2x_fastpath *fp,
 
 	dma_unmap_addr_set(prod_rx_buf, mapping,
 			   dma_unmap_addr(cons_rx_buf, mapping));
-	prod_rx_buf->skb = cons_rx_buf->skb;
+	prod_rx_buf->data = cons_rx_buf->data;
 	*prod_bd = *cons_bd;
 }
 
@@ -1029,9 +1030,9 @@ static inline void bnx2x_free_tpa_pool(struct bnx2x *bp,
 	for (i = 0; i < last; i++) {
 		struct bnx2x_agg_info *tpa_info = &fp->tpa_info[i];
 		struct sw_rx_bd *first_buf = &tpa_info->first_buf;
-		struct sk_buff *skb = first_buf->skb;
+		u8 *data = first_buf->data;
 
-		if (skb == NULL) {
+		if (data == NULL) {
 			DP(NETIF_MSG_IFDOWN, "tpa bin %d empty on free\n", i);
 			continue;
 		}
@@ -1039,8 +1040,8 @@ static inline void bnx2x_free_tpa_pool(struct bnx2x *bp,
 			dma_unmap_single(&bp->pdev->dev,
 					 dma_unmap_addr(first_buf, mapping),
 					 fp->rx_buf_size, DMA_FROM_DEVICE);
-		dev_kfree_skb(skb);
-		first_buf->skb = NULL;
+		kfree(data);
+		first_buf->data = NULL;
 	}
 }
 
@@ -1148,7 +1149,7 @@ static inline int bnx2x_alloc_rx_bds(struct bnx2x_fastpath *fp,
 	 * fp->eth_q_stats.rx_skb_alloc_failed = 0
 	 */
 	for (i = 0; i < rx_ring_size; i++) {
-		if (bnx2x_alloc_rx_skb(bp, fp, ring_prod) < 0) {
+		if (bnx2x_alloc_rx_data(bp, fp, ring_prod) < 0) {
 			fp->eth_q_stats.rx_skb_alloc_failed++;
 			continue;
 		}