net: NET_SKB_PAD should depend on L1_CACHE_BYTES

In old kernels, NET_SKB_PAD was defined to 16.

Then commit d6301d3dd1c2 (net: Increase default NET_SKB_PAD to 32), and
commit 18e8c134f4e9 (net: Increase NET_SKB_PAD to 64 bytes) increased it
to 64.

While first patch was governed by network stack needs, second was more
driven by performance issues on current hardware. Real intent was to
align data on a cache line boundary.

So use max(32, L1_CACHE_BYTES) instead of 64, to be more generic.

Remove microblaze and powerpc own NET_SKB_PAD definitions.

Thanks to Alexander Duyck and David Miller for their comments.

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

1 files changed, 5 insertions, 3 deletions

diff --git a/include/linux/skbuff.h b/include/linux/skbuff.h
index 122d08396e5..ac74ee085d7 100644
--- a/include/linux/skbuff.h
+++ b/include/linux/skbuff.h
@@ -1414,12 +1414,14 @@ static inline int skb_network_offset(const struct sk_buff *skb)
  *
  * Various parts of the networking layer expect at least 32 bytes of
  * headroom, you should not reduce this.
- * With RPS, we raised NET_SKB_PAD to 64 so that get_rps_cpus() fetches span
- * a 64 bytes aligned block to fit modern (>= 64 bytes) cache line sizes
+ *
+ * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
+ * to reduce average number of cache lines per packet.
+ * get_rps_cpus() for example only access one 64 bytes aligned block :
  * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
  */
 #ifndef NET_SKB_PAD
-#define NET_SKB_PAD	64
+#define NET_SKB_PAD	max(32, L1_CACHE_BYTES)
 #endif
 
 extern int ___pskb_trim(struct sk_buff *skb, unsigned int len);