[PATCH] IPoIB: fix skb truesize underestimatiom

Subsystems: infiniband subsystem, the rest

STALE5149d

2 messages, 2 authors, 2012-07-11 · open the first message on its own page

[PATCH] IPoIB: fix skb truesize underestimatiom

From: Eric Dumazet <hidden>
Date: 2012-07-10 20:03:46

From: Eric Dumazet <edumazet@google.com>

Or Gerlitz reported triggering of WARN_ON_ONCE(delta < len); in
skb_try_coalesce()
This warning tracks drivers that incorrectly set skb->truesize

IPoIB indeed allocates a full page to store a fragment, but only
accounts in skb->truesize the used part of the page (frame length)

This patch fixes skb truesize underestimation, and
also fixes a performance issue, because RX skbs have not enough tailroom
to allow IP and TCP stacks to pull their header in skb linear part
without an expensive call to pskb_expand_head()

Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: Or Gerlitz <redacted>
Cc: Erez Shitrit <redacted>
Cc: Shlomo Pongartz <redacted>
Cc: Roland Dreier <redacted>
---
 drivers/infiniband/ulp/ipoib/ipoib_ib.c |   12 ++++++++----
 1 file changed, 8 insertions(+), 4 deletions(-)
diff --git a/drivers/infiniband/ulp/ipoib/ipoib_ib.c b/drivers/infiniband/ulp/ipoib/ipoib_ib.c
index 5c1bc99..f10221f 100644
--- a/drivers/infiniband/ulp/ipoib/ipoib_ib.c
+++ b/drivers/infiniband/ulp/ipoib/ipoib_ib.c
@@ -123,7 +123,7 @@ static void ipoib_ud_skb_put_frags(struct ipoib_dev_priv *priv,
 
 		skb_frag_size_set(frag, size);
 		skb->data_len += size;
-		skb->truesize += size;
+		skb->truesize += PAGE_SIZE;
 	} else
 		skb_put(skb, length);
 
@@ -156,14 +156,18 @@ static struct sk_buff *ipoib_alloc_rx_skb(struct net_device *dev, int id)
 	struct ipoib_dev_priv *priv = netdev_priv(dev);
 	struct sk_buff *skb;
 	int buf_size;
+	int tailroom;
 	u64 *mapping;
 
-	if (ipoib_ud_need_sg(priv->max_ib_mtu))
+	if (ipoib_ud_need_sg(priv->max_ib_mtu)) {
 		buf_size = IPOIB_UD_HEAD_SIZE;
-	else
+		tailroom = 128; /* reserve some tailroom for IP/TCP headers */
+	} else {
 		buf_size = IPOIB_UD_BUF_SIZE(priv->max_ib_mtu);
+		tailroom = 0;
+	}
 
-	skb = dev_alloc_skb(buf_size + 4);
+	skb = dev_alloc_skb(buf_size + tailroom + 4);
 	if (unlikely(!skb))
 		return NULL;
 

Re: [PATCH] IPoIB: fix skb truesize underestimatiom

From: David Miller <davem@davemloft.net>
Date: 2012-07-11 06:37:46

From: Eric Dumazet <redacted>
Date: Tue, 10 Jul 2012 22:03:41 +0200
From: Eric Dumazet <edumazet@google.com>

Or Gerlitz reported triggering of WARN_ON_ONCE(delta < len); in
skb_try_coalesce()
This warning tracks drivers that incorrectly set skb->truesize

IPoIB indeed allocates a full page to store a fragment, but only
accounts in skb->truesize the used part of the page (frame length)

This patch fixes skb truesize underestimation, and
also fixes a performance issue, because RX skbs have not enough tailroom
to allow IP and TCP stacks to pull their header in skb linear part
without an expensive call to pskb_expand_head()

Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: Or Gerlitz <redacted>
Applied to net, thanks.
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help