tcp: fix tp->undo_retrans accounting in tcp_sacktag_one()
authorzhenggy <zhenggy@chinatelecom.cn>
Tue, 14 Sep 2021 01:51:15 +0000 (09:51 +0800)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Wed, 22 Sep 2021 09:43:09 +0000 (11:43 +0200)
commitb3507f2bef0795116eb6d7e49025e72ee8e93024
tree5c2d8181de6344054f98f7a51d0d80e041a53ce8
parent69407175c62f56070be65257d196a8ef2b9fabe7
tcp: fix tp->undo_retrans accounting in tcp_sacktag_one()

commit 4f884f3962767877d7aabbc1ec124d2c307a4257 upstream.

Commit 10d3be569243 ("tcp-tso: do not split TSO packets at retransmit
time") may directly retrans a multiple segments TSO/GSO packet without
split, Since this commit, we can no longer assume that a retransmitted
packet is a single segment.

This patch fixes the tp->undo_retrans accounting in tcp_sacktag_one()
that use the actual segments(pcount) of the retransmitted packet.

Before that commit (10d3be569243), the assumption underlying the
tp->undo_retrans-- seems correct.

Fixes: 10d3be569243 ("tcp-tso: do not split TSO packets at retransmit time")
Signed-off-by: zhenggy <zhenggy@chinatelecom.cn>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Acked-by: Yuchung Cheng <ycheng@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
net/ipv4/tcp_input.c