tcp: Fix length tcp_splice_data_recv passes to skb_splice_bits.
authorDimitris Michailidis <dm@chelsio.com>
Tue, 27 Jan 2009 06:15:31 +0000 (22:15 -0800)
committerDavid S. Miller <davem@davemloft.net>
Tue, 27 Jan 2009 06:15:31 +0000 (22:15 -0800)
commit9fa5fdf291c9b58b1cb8b4bb2a0ee57efa21d635
treecfa28bcf674ef224a5ab66b3b8bc4bc1091ccb11
parenta7a41acf99d9150b424839b0d7b4f5ad9d211e2d
tcp: Fix length tcp_splice_data_recv passes to skb_splice_bits.

tcp_splice_data_recv has two lengths to consider: the len parameter it
gets from tcp_read_sock, which specifies the amount of data in the skb,
and rd_desc->count, which is the amount of data the splice caller still
wants.  Currently it passes just the latter to skb_splice_bits, which then
splices min(rd_desc->count, skb->len - offset) bytes.

Most of the time this is fine, except when the skb contains urgent data.
In that case len goes only up to the urgent byte and is less than
skb->len - offset.  By ignoring len tcp_splice_data_recv may a) splice
data tcp_read_sock told it not to, b) return to tcp_read_sock a value > len.

Now, tcp_read_sock doesn't handle used > len and leaves the socket in a
bad state (both sk_receive_queue and copied_seq are bad at that point)
resulting in duplicated data and corruption.

Fix by passing min(rd_desc->count, len) to skb_splice_bits.

Signed-off-by: Dimitris Michailidis <dm@chelsio.com>
Acked-by: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
net/ipv4/tcp.c