[NET]: Detect hardware rx checksum faults correctly

Here is the patch that introduces the generic skb_checksum_complete
which also checks for hardware RX checksum faults.  If that happens,
it'll call netdev_rx_csum_fault which currently prints out a stack
trace with the device name.  In future it can turn off RX checksum.

I've converted every spot under net/ that does RX checksum checks to
use skb_checksum_complete or __skb_checksum_complete with the
exceptions of:

* Those places where checksums are done bit by bit.  These will call
netdev_rx_csum_fault directly.

* The following have not been completely checked/converted:

ipmr
ip_vs
netfilter
dccp

This patch is based on patches and suggestions from Stephen Hemminger
and David S. Miller.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>

3 files changed, 41 insertions, 10 deletions

diff --git a/net/core/datagram.c b/net/core/datagram.c
index d219435..1bcfef5 100644
--- a/net/core/datagram.c
+++ b/net/core/datagram.c
@@ -350,6 +350,20 @@ fault:
 	return -EFAULT;
 }
 
+unsigned int __skb_checksum_complete(struct sk_buff *skb)
+{
+	unsigned int sum;
+
+	sum = (u16)csum_fold(skb_checksum(skb, 0, skb->len, skb->csum));
+	if (likely(!sum)) {
+		if (unlikely(skb->ip_summed == CHECKSUM_HW))
+			netdev_rx_csum_fault(skb->dev);
+		skb->ip_summed = CHECKSUM_UNNECESSARY;
+	}
+	return sum;
+}
+EXPORT_SYMBOL(__skb_checksum_complete);
+
 /**
  *	skb_copy_and_csum_datagram_iovec - Copy and checkum skb to user iovec.
  *	@skb: skbuff
@@ -363,7 +377,7 @@ fault:
  *		 -EFAULT - fault during copy. Beware, in this case iovec
  *			   can be modified!
  */
-int skb_copy_and_csum_datagram_iovec(const struct sk_buff *skb,
+int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb,
 				     int hlen, struct iovec *iov)
 {
 	unsigned int csum;
@@ -376,8 +390,7 @@ int skb_copy_and_csum_datagram_iovec(const struct sk_buff *skb,
 		iov++;
 
 	if (iov->iov_len < chunk) {
-		if ((unsigned short)csum_fold(skb_checksum(skb, 0, chunk + hlen,
-							   skb->csum)))
+		if (__skb_checksum_complete(skb))
 			goto csum_error;
 		if (skb_copy_datagram_iovec(skb, hlen, iov, chunk))
 			goto fault;
@@ -388,6 +401,8 @@ int skb_copy_and_csum_datagram_iovec(const struct sk_buff *skb,
 			goto fault;
 		if ((unsigned short)csum_fold(csum))
 			goto csum_error;
+		if (unlikely(skb->ip_summed == CHECKSUM_HW))
+			netdev_rx_csum_fault(skb->dev);
 		iov->iov_len -= chunk;
 		iov->iov_base += chunk;
 	}
diff --git a/net/core/dev.c b/net/core/dev.c
index 8d15415..0b48e29 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c
@@ -1108,6 +1108,18 @@ out:
 	return ret;
 }
 
+/* Take action when hardware reception checksum errors are detected. */
+#ifdef CONFIG_BUG
+void netdev_rx_csum_fault(struct net_device *dev)
+{
+	if (net_ratelimit()) {
+		printk(KERN_ERR "%s: hw csum failure.\n", dev->name);
+		dump_stack();
+	}
+}
+EXPORT_SYMBOL(netdev_rx_csum_fault);
+#endif
+
 #ifdef CONFIG_HIGHMEM
 /* Actually, we should eliminate this check as soon as we know, that:
  * 1. IOMMU is present and allows to map all the memory.
diff --git a/net/core/netpoll.c b/net/core/netpoll.c
index 802fe11..49424a4 100644
--- a/net/core/netpoll.c
+++ b/net/core/netpoll.c
@@ -101,16 +101,20 @@ void netpoll_queue(struct sk_buff *skb)
 static int checksum_udp(struct sk_buff *skb, struct udphdr *uh,
 			     unsigned short ulen, u32 saddr, u32 daddr)
 {
-	if (uh->check == 0)
+	unsigned int psum;
+
+	if (uh->check == 0 || skb->ip_summed == CHECKSUM_UNNECESSARY)
 		return 0;
 
-	if (skb->ip_summed == CHECKSUM_HW)
-		return csum_tcpudp_magic(
-			saddr, daddr, ulen, IPPROTO_UDP, skb->csum);
+	psum = csum_tcpudp_nofold(saddr, daddr, ulen, IPPROTO_UDP, 0);
+
+	if (skb->ip_summed == CHECKSUM_HW &&
+	    !(u16)csum_fold(csum_add(psum, skb->csum)))
+		return 0;
 
-	skb->csum = csum_tcpudp_nofold(saddr, daddr, ulen, IPPROTO_UDP, 0);
+	skb->csum = psum;
 
-	return csum_fold(skb_checksum(skb, 0, skb->len, skb->csum));
+	return __skb_checksum_complete(skb);
 }
 
 /*
@@ -489,7 +493,7 @@ int __netpoll_rx(struct sk_buff *skb)
 
 	if (ulen != len)
 		goto out;
-	if (checksum_udp(skb, uh, ulen, iph->saddr, iph->daddr) < 0)
+	if (checksum_udp(skb, uh, ulen, iph->saddr, iph->daddr))
 		goto out;
 	if (np->local_ip && np->local_ip != ntohl(iph->daddr))
 		goto out;