]> bbs.cooldavid.org Git - net-next-2.6.git/commit - net/core/dev.c
[NET]: Use csum_start offset instead of skb_transport_header
authorHerbert Xu <herbert@gondor.apana.org.au>
Mon, 9 Apr 2007 18:59:07 +0000 (11:59 -0700)
committerDavid S. Miller <davem@sunset.davemloft.net>
Thu, 26 Apr 2007 05:28:40 +0000 (22:28 -0700)
commit663ead3bb8d5b561e70fc3bb3861c9220b5a77eb
tree551d031fc9c904487ced4d8d3363e5ed8f369794
parentac758e3c55c529714354fc268892ca4d23ca1e99
[NET]: Use csum_start offset instead of skb_transport_header

The skb transport pointer is currently used to specify the start
of the checksum region for transmit checksum offload.  Unfortunately,
the same pointer is also used during receive side processing.

This creates a problem when we want to retransmit a received
packet with partial checksums since the skb transport pointer
would be overwritten.

This patch solves this problem by creating a new 16-bit csum_start
offset value to replace the skb transport header for the purpose
of checksums.  This offset is calculated from skb->head so that
it does not have to change when skb->data changes.

No extra space is required since csum_offset itself fits within
a 16-bit word so we can use the other 16 bits for csum_start.

For backwards compatibility, just before we push a packet with
partial checksums off into the device driver, we set the skb
transport header to what it would have been under the old scheme.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
include/linux/skbuff.h
net/core/dev.c
net/core/skbuff.c
net/ipv4/tcp_ipv4.c
net/ipv4/udp.c
net/ipv6/tcp_ipv6.c