lk/linux-kernel-test
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

net: skb_shared_info optimization

Browse Source 
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>
This commit is contained in:
Eric Dumazet
2009-06-05 04:04:16 +00:00
committed by David S. Miller
parent eae3f29cc7
commit 042a53a9e4
10 changed files with 30 additions and 29 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
include/linux/skbuff.h
View File

@@ -189,6 +189,9 @@ struct skb_shared_info {
atomic_t dataref;
unsigned short nr_frags;
unsigned short gso_size;
#ifdef CONFIG_HAS_DMA
dma_addr_t dma_head;
#endif
/* Warning: this field is not always filled in (UFO)! */
unsigned short gso_segs;
unsigned short gso_type;
@@ -198,7 +201,7 @@ struct skb_shared_info {
struct skb_shared_hwtstamps hwtstamps;
skb_frag_t frags[MAX_SKB_FRAGS];
#ifdef CONFIG_HAS_DMA
dma_addr_t dma_maps[MAX_SKB_FRAGS + 1];
dma_addr_t dma_maps[MAX_SKB_FRAGS];
#endif
/* Intermediate layers must ensure that destructor_arg
* remains valid until skb destructor */