lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
Open Source and information security mailing list archives
| ||
|
Message-ID: <68a0e4baf2ace47c04874f1918a3bd5f1e5784dc.1346775479.git.nicolas.ferre@atmel.com> Date: Wed, 5 Sep 2012 11:04:06 +0200 From: Nicolas Ferre <nicolas.ferre@...el.com> To: <netdev@...r.kernel.org> CC: <linux-arm-kernel@...ts.infradead.org>, <davem@...emloft.net>, <havard@...nnemoen.net>, <nicolas.ferre@...el.com>, <plagnioj@...osoft.com>, <jamie@...ieiles.com>, <linux-kernel@...r.kernel.org>, <patrice.vilchez@...el.com> Subject: [PATCH 10/10] net/macb: Offset first RX buffer by two bytes From: Havard Skinnemoen <havard@...nnemoen.net> Make the ethernet frame payload word-aligned, possibly making the memcpy into the skb a bit faster. This will be even more important after we eliminate the copy altogether. Also eliminate the redundant RX_OFFSET constant -- it has the same definition and purpose as NET_IP_ALIGN. Signed-off-by: Havard Skinnemoen <havard@...nnemoen.net> [nicolas.ferre@...el.com: adapt to newer kernel] Signed-off-by: Nicolas Ferre <nicolas.ferre@...el.com> --- drivers/net/ethernet/cadence/macb.c | 23 ++++++++++++++++------- 1 file changed, 16 insertions(+), 7 deletions(-) diff --git a/drivers/net/ethernet/cadence/macb.c b/drivers/net/ethernet/cadence/macb.c index f31c0a7..f7716b6 100644 --- a/drivers/net/ethernet/cadence/macb.c +++ b/drivers/net/ethernet/cadence/macb.c @@ -33,9 +33,6 @@ #define RX_RING_SIZE 512 #define RX_RING_BYTES (sizeof(struct macb_dma_desc) * RX_RING_SIZE) -/* Make the IP header word-aligned (the ethernet header is 14 bytes) */ -#define RX_OFFSET 2 - #define TX_RING_SIZE 128 #define TX_RING_BYTES (sizeof(struct macb_dma_desc) * TX_RING_SIZE) @@ -466,7 +463,7 @@ static int macb_rx_frame(struct macb *bp, unsigned int first_frag, { unsigned int len; unsigned int frag; - unsigned int offset = 0; + unsigned int offset; struct sk_buff *skb; struct macb_dma_desc *desc; @@ -477,7 +474,16 @@ static int macb_rx_frame(struct macb *bp, unsigned int first_frag, macb_rx_ring_wrap(first_frag), macb_rx_ring_wrap(last_frag), len); - skb = netdev_alloc_skb(bp->dev, len + RX_OFFSET); + /* + * The ethernet header starts NET_IP_ALIGN bytes into the + * first buffer. Since the header is 14 bytes, this makes the + * payload word-aligned. + * + * Instead of calling skb_reserve(NET_IP_ALIGN), we just copy + * the two padding bytes into the skb so that we avoid hitting + * the slowpath in memcpy(), and pull them off afterwards. + */ + skb = netdev_alloc_skb(bp->dev, len + NET_IP_ALIGN); if (!skb) { bp->stats.rx_dropped++; for (frag = first_frag; ; frag++) { @@ -493,7 +499,8 @@ static int macb_rx_frame(struct macb *bp, unsigned int first_frag, return 1; } - skb_reserve(skb, RX_OFFSET); + offset = 0; + len += NET_IP_ALIGN; skb_checksum_none_assert(skb); skb_put(skb, len); @@ -517,10 +524,11 @@ static int macb_rx_frame(struct macb *bp, unsigned int first_frag, /* Make descriptor updates visible to hardware */ wmb(); + __skb_pull(skb, NET_IP_ALIGN); skb->protocol = eth_type_trans(skb, bp->dev); bp->stats.rx_packets++; - bp->stats.rx_bytes += len; + bp->stats.rx_bytes += skb->len; netdev_vdbg(bp->dev, "received skb of length %u, csum: %08x\n", skb->len, skb->csum); netif_receive_skb(skb); @@ -985,6 +993,7 @@ static void macb_init_hw(struct macb *bp) __macb_set_hwaddr(bp); config = macb_mdc_clk_div(bp); + config |= MACB_BF(RBOF, NET_IP_ALIGN); /* Make eth data aligned */ config |= MACB_BIT(PAE); /* PAuse Enable */ config |= MACB_BIT(DRFCS); /* Discard Rx FCS */ config |= MACB_BIT(BIG); /* Receive oversized frames */ -- 1.7.10 -- To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to majordomo@...r.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Powered by blists - more mailing lists