net: remove skb recycling
authorEric Dumazet <edumazet@google.com>
Fri, 5 Oct 2012 06:23:55 +0000 (06:23 +0000)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Sun, 28 Oct 2012 17:56:16 +0000 (10:56 -0700)
commit541321132ac28c92394b1d57361448d2c6dd5faf
treea7d17762841432532967d94479a435834f72f435
parente9ad2fc50a33c322bf343661851b1f179598b072
net: remove skb recycling

[ Upstream commits acb600def2110b1310466c0e485c0d26299898ae
  and 66eef59f22275002f621ff9d951886b513d011b3. ]

Over time, skb recycling infrastructure got litle interest and
many bugs. Generic rx path skb allocation is now using page
fragments for efficient GRO / TCP coalescing, and recyling
a tx skb for rx path is not worth the pain.

Last identified bug is that fat skbs can be recycled
and it can endup using high order pages after few iterations.

With help from Maxime Bizon, who pointed out that commit
87151b8689d (net: allow pskb_expand_head() to get maximum tailroom)
introduced this regression for recycled skbs.

Instead of fixing this bug, lets remove skb recycling.

Drivers wanting really hot skbs should use build_skb() anyway,
to allocate/populate sk_buff right before netif_receive_skb()

Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Maxime Bizon <mbizon@freebox.fr>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
drivers/net/ethernet/calxeda/xgmac.c
drivers/net/ethernet/freescale/gianfar.c
drivers/net/ethernet/freescale/gianfar.h
drivers/net/ethernet/freescale/ucc_geth.c
drivers/net/ethernet/freescale/ucc_geth.h
drivers/net/ethernet/marvell/mv643xx_eth.c
drivers/net/ethernet/stmicro/stmmac/stmmac.h
drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
include/linux/skbuff.h
net/core/skbuff.c