Skip to contentSkip to conversation

virtio_net GSO/partial csum fixes

PATCH In review Mark McLoughlin 3 participants 8 messages May 27, 2008 – May 29, 2008

Subsystems (4)
lore
Share & tools
lore
#0 lore raw

cover virtio_net GSO/partial csum fixes

Hi Rusty,
Here's a few virtio_net fixes related to partial csums
and GSO. These come about from playing with the tun/tap based GSO patches you and Herbert posted earlier.

3/3 is the signifcant one - I didn't actually reproduce
the problem with virtio_net, but rather observed it with the tun/tap patch. In that case, packets which were sent by the guest with a partial csum were ending up with an incorrect csum when they were forwarded to a physical network.

Cheers,
Mark.

#1 lore raw

PATCH 1/3 virtio: Fix typo in virtio_net_hdr comments

---
 include/linux/virtio_net.h |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)
include/linux/virtio_net.h +1 -1
index 9405aa6..38c0571 100644
@@ -38,7 +38,7 @@ struct virtio_net_hdr
 #define VIRTIO_NET_HDR_GSO_ECN		0x80	// TCP has ECN set
 	__u8 gso_type;
 	__u16 hdr_len;		/* Ethernet + IP + tcp/udp hdrs */
-	__u16 gso_size;		/* Bytes to append to gso_hdr_len per frame */
+	__u16 gso_size;		/* Bytes to append to hdr_len per frame */
 	__u16 csum_start;	/* Position to start checksumming from */
 	__u16 csum_offset;	/* Offset after that to place checksum */
 };
Signature · 1 line
1.5.4.1
#2 lore raw

PATCH 2/3 virtio_net: Trivial coding style fix

---
 drivers/net/virtio_net.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)
drivers/net/virtio_net.c +1 -1
index fe7cdf2..50e6e59 100644
@@ -94,7 +94,7 @@ static void receive_skb(struct net_device *dev, struct sk_buff *skb,
 
 	if (hdr->flags & VIRTIO_NET_HDR_F_NEEDS_CSUM) {
 		pr_debug("Needs csum!\n");
-		if (!skb_partial_csum_set(skb,hdr->csum_start,hdr->csum_offset))
+		if (!skb_partial_csum_set(skb, hdr->csum_start, hdr->csum_offset))
 			goto frame_err;
 	}
 
Signature · 1 line
1.5.4.1
#3 lore raw

PATCH 3/3 virtio_net: Fix skb->csum_start computation

hdr->csum_start is the offset from the start of the ethernet header to the transport layer checksum field. skb->csum_start is the offset from skb->head.

skb_partial_csum_set() assumes that skb->data points to the ethernet header - i.e. it computes skb->csum_start by adding the headroom to hdr->csum_start.

Since eth_type_trans() skb_pull()s the ethernet header, skb_partial_csum_set() should be called before eth_type_trans().

---
 drivers/net/virtio_net.c |    8 +++++---
 1 files changed, 5 insertions(+), 3 deletions(-)
drivers/net/virtio_net.c +5 -3
index 50e6e59..503486f 100644
@@ -86,9 +86,7 @@ static void receive_skb(struct net_device *dev, struct sk_buff *skb,
 	BUG_ON(len > MAX_PACKET_LEN);
 
 	skb_trim(skb, len);
-	skb->protocol = eth_type_trans(skb, dev);
-	pr_debug("Receiving skb proto 0x%04x len %i type %i\n",
-		 ntohs(skb->protocol), skb->len, skb->pkt_type);
+
 	dev->stats.rx_bytes += skb->len;
 	dev->stats.rx_packets++;
 
@@ -98,6 +96,10 @@ static void receive_skb(struct net_device *dev, struct sk_buff *skb,
 			goto frame_err;
 	}
 
+	skb->protocol = eth_type_trans(skb, dev);
+	pr_debug("Receiving skb proto 0x%04x len %i type %i\n",
+		 ntohs(skb->protocol), skb->len, skb->pkt_type);
+
 	if (hdr->gso_type != VIRTIO_NET_HDR_GSO_NONE) {
 		pr_debug("GSO!\n");
 		switch (hdr->gso_type & ~VIRTIO_NET_HDR_GSO_ECN) {
Signature · 1 line
1.5.4.1
#7 lore raw

Re: [PATCH 2/3] virtio_net: Trivial coding style fix

In reply to Mark McLoughlin

drivers/net/virtio_net.c static void receive_skb(struct net_device *dev, struct sk_buff *skb, · line 97

Quoted 2 lines · On Tuesday 27 May 2008 21:20:46 Mark McLoughlin wrote:
drivers/net/virtio_net.c +1 -1
-		if (!skb_partial_csum_set(skb,hdr->csum_start,hdr->csum_offset))
+		if (!skb_partial_csum_set(skb, hdr->csum_start, hdr->csum_offset))

I compressed the whitespace to avoid going over 80 cols. A choice of lesser evils :)

Cheers,
Rusty.

Acked-by on v1 3/3
#4 lore raw

Re: [PATCH 3/3] virtio_net: Fix skb->csum_start computation

In reply to Mark McLoughlin
Quoted 8 lines · Mark McLoughlin

hdr->csum_start is the offset from the start of the ethernet header to the transport layer checksum field. skb->csum_start is the offset from skb->head.

skb_partial_csum_set() assumes that skb->data points to the ethernet header - i.e. it computes skb->csum_start by adding the headroom to hdr->csum_start.

Quoted 5 lines

Since eth_type_trans() skb_pull()s the ethernet header, skb_partial_csum_set() should be called before eth_type_trans().

Good catch! Clearly shows I never ran this across a real Ethernet device :)

Thanks,

Signature · 4 lines
Visit Openswan at http://www.openswan.org/
Email: Herbert Xu ~{PmV>HI~} <[email protected]>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt
#5 lore raw

Re: [PATCH 3/3] virtio_net: Fix skb->csum_start computation

In reply to Mark McLoughlin
Quoted 11 lines · On Tuesday 27 May 2008 21:20:47 Mark McLoughlin wrote:

hdr->csum_start is the offset from the start of the ethernet header to the transport layer checksum field. skb->csum_start is the offset from skb->head.

skb_partial_csum_set() assumes that skb->data points to the ethernet header - i.e. it computes skb->csum_start by adding the headroom to hdr->csum_start.

Since eth_type_trans() skb_pull()s the ethernet header, skb_partial_csum_set() should be called before eth_type_trans().

As should the rx_bytes += skb->len. Thanks for this!

Applied (and 1/3, the typo patch).

I'll backport this to -stable, as well.

Thanks,
Rusty.

#6 lore raw

Re: [PATCH 3/3] virtio_net: Fix skb->csum_start computation

In reply to Mark McLoughlin
Quoted 11 lines · On Tuesday 27 May 2008 21:20:47 Mark McLoughlin wrote:

hdr->csum_start is the offset from the start of the ethernet header to the transport layer checksum field. skb->csum_start is the offset from skb->head.

skb_partial_csum_set() assumes that skb->data points to the ethernet header - i.e. it computes skb->csum_start by adding the headroom to hdr->csum_start.

Since eth_type_trans() skb_pull()s the ethernet header, skb_partial_csum_set() should be called before eth_type_trans().

As should the rx_bytes += skb->len. Thanks for this!

Applied (and 1/3, the typo patch).

I'll backport this to -stable, as well.

Thanks,
Rusty.