net: skb_shared_info optimization

skb_dma_unmap() is quite expensive for small packets,
because we use two different cache lines from skb_shared_info.

One to access nr_frags, one to access dma_maps[0]

Instead of dma_maps being an array of MAX_SKB_FRAGS + 1 elements,
let dma_head alone in a new dma_head field, close to nr_frags,
to reduce cache lines misses.

Tested on my dev machine (bnx2 & tg3 adapters), nice speedup !

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

1 files changed, 5 insertions, 5 deletions

diff --git a/drivers/net/tg3.c b/drivers/net/tg3.c
index a39b534fb43..46a3f86125b 100644
--- a/drivers/net/tg3.c
+++ b/drivers/net/tg3.c
@@ -5021,7 +5021,7 @@ static int tigon3_dma_hwbug_workaround(struct tg3 *tp, struct sk_buff *skb,
                 /* New SKB is guaranteed to be linear. */
                 entry = *start;
                 ret = skb_dma_map(&tp->pdev->dev, new_skb, DMA_TO_DEVICE);
-                new_addr = skb_shinfo(new_skb)->dma_maps[0];
+                new_addr = skb_shinfo(new_skb)->dma_head;
 
                 /* Make sure new skb does not cross any 4G boundaries.
                  * Drop the packet if it does.
@@ -5155,7 +5155,7 @@ static int tg3_start_xmit(struct sk_buff *skb, struct net_device *dev)
 
         sp = skb_shinfo(skb);
 
-        mapping = sp->dma_maps[0];
+        mapping = sp->dma_head;
 
         tp->tx_buffers[entry].skb = skb;
 
@@ -5173,7 +5173,7 @@ static int tg3_start_xmit(struct sk_buff *skb, struct net_device *dev)
                         skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
 
                         len = frag->size;
-                        mapping = sp->dma_maps[i + 1];
+                        mapping = sp->dma_maps[i];
                         tp->tx_buffers[entry].skb = NULL;
 
                         tg3_set_txd(tp, entry, mapping, len,
@@ -5331,7 +5331,7 @@ static int tg3_start_xmit_dma_bug(struct sk_buff *skb, struct net_device *dev)
 
         sp = skb_shinfo(skb);
 
-        mapping = sp->dma_maps[0];
+        mapping = sp->dma_head;
 
         tp->tx_buffers[entry].skb = skb;
 
@@ -5356,7 +5356,7 @@ static int tg3_start_xmit_dma_bug(struct sk_buff *skb, struct net_device *dev)
                         skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
 
                         len = frag->size;
-                        mapping = sp->dma_maps[i + 1];
+                        mapping = sp->dma_maps[i];
 
                         tp->tx_buffers[entry].skb = NULL;