Thread (12 messages) flat view 12 messages, 2 authors, 2011-10-18

Re: [net-next 4/7] stmmac: allow mtu bigger than 1500 in case of normal desc (V3)

From: Giuseppe CAVALLARO <hidden>
Date: 2011-10-18 09:26:18

Hello Eric,

On 10/18/2011 11:03 AM, Eric Dumazet wrote:
Le mardi 18 octobre 2011 à 10:42 +0200, Giuseppe CAVALLARO a écrit :
quoted
This patch allows to set the mtu bigger than 1500
in case of normal descriptors.
This is helping some SPEAr customers.

Signed-off-by: Deepak SIKRI <redacted>
Signed-off-by: Giuseppe Cavallaro <redacted>
---
 drivers/net/ethernet/stmicro/stmmac/norm_desc.c   |    8 +++++++-
 drivers/net/ethernet/stmicro/stmmac/stmmac_main.c |    4 ++--
 2 files changed, 9 insertions(+), 3 deletions(-)
diff --git a/drivers/net/ethernet/stmicro/stmmac/norm_desc.c b/drivers/net/ethernet/stmicro/stmmac/norm_desc.c
index 029c2a2..e13226b 100644
--- a/drivers/net/ethernet/stmicro/stmmac/norm_desc.c
+++ b/drivers/net/ethernet/stmicro/stmmac/norm_desc.c
@@ -126,6 +126,7 @@ static void ndesc_init_rx_desc(struct dma_desc *p, unsigned int ring_size,
 	for (i = 0; i < ring_size; i++) {
 		p->des01.rx.own = 1;
 		p->des01.rx.buffer1_size = BUF_SIZE_2KiB - 1;
+		p->des01.rx.buffer2_size = BUF_SIZE_2KiB - 1;
 		if (i == ring_size - 1)
 			p->des01.rx.end_ring = 1;
 		if (disable_rx_ic)
@@ -183,7 +184,12 @@ static void ndesc_prepare_tx_desc(struct dma_desc *p, int is_fs, int len,
 				  int csum_flag)
 {
 	p->des01.tx.first_segment = is_fs;
-	p->des01.tx.buffer1_size = len;
+
+	if (unlikely(len > BUF_SIZE_2KiB)) {
+		p->des01.etx.buffer1_size = BUF_SIZE_2KiB - 1;
+		p->des01.etx.buffer2_size = len - p->des01.etx.buffer1_size;
+	} else
+		p->des01.tx.buffer1_size = len;
 }
 
 static void ndesc_clear_tx_ic(struct dma_desc *p)
diff --git a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
index d0af002..622b7ac 100644
--- a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
+++ b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
@@ -1412,10 +1412,10 @@ static int stmmac_change_mtu(struct net_device *dev, int new_mtu)
 		return -EBUSY;
 	}
 
-	if (priv->plat->has_gmac)
+	if (priv->plat->enh_desc)
 		max_mtu = JUMBO_LEN;
 	else
-		max_mtu = ETH_DATA_LEN;
+		max_mtu = BUF_SIZE_4KiB;
 
 	if ((new_mtu < 46) || (new_mtu > max_mtu)) {
 		pr_err("%s: invalid MTU, max MTU is: %d\n", dev->name, max_mtu);
Problem using big mtu around 4096 bytes is you end allocating (4096
+NET_SKB_PAD + NET_IP_ALIGN + sizeof(struct skb_shared_info) bytes ->
8192 bytes : order-1 pages
yes.
Maybe it would be better to limit your mtu to SKB_MAX_HEAD(NET_SKB_PAD),
to have no more than one page per skb ?
ok! ;-)
quoted hunk ↗ jump to hunk
I would suggest changing netdev_alloc_skb_ip_align() done in
init_dma_desc_rings() to use a variant allowing GFP_KERNEL allocations
and allow your driver to load even in case of memory pressure.
diff --git a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
index c0ee6b6..8ec8057 100644
--- a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
+++ b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
@@ -458,11 +458,12 @@ static void init_dma_desc_rings(struct net_device *dev)
 	for (i = 0; i < rxsize; i++) {
 		struct dma_desc *p = priv->dma_rx + i;
 
-		skb = netdev_alloc_skb_ip_align(dev, bfsize);
+		skb = __netdev_alloc_skb(dev, bfsize + NET_IP_ALIGN, GFP_KERNEL);
 		if (unlikely(skb == NULL)) {
 			pr_err("%s: Rx init fails; skb is NULL\n", __func__);
 			break;
 		}
+		skb_reserve(skb, NET_IP_ALIGN);
 		priv->rx_skbuff[i] = skb;
 		priv->rx_skbuff_dma[i] = dma_map_single(priv->device, skb->data,
 						bfsize, DMA_FROM_DEVICE);


Agree! I'm reviewing the patches and send all again.

Many thanks
Regards
Peppe
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help