tcp: fix "old stuff" D-SACK causing SACK to be treated as D-SACK
authorPengcheng Yang <yangpc@wangsu.com>
Mon, 30 Dec 2019 09:54:41 +0000 (17:54 +0800)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Sun, 12 Jan 2020 11:17:27 +0000 (12:17 +0100)
commit5ea5b6e3c6e146efeae7df6b0da898868cdb4765
tree38c577e7248d98ec726ec0c9f6a25c762ded6374
parent5f52b9ebd47caf11e5ee11172c1db42ae2f26fe5
tcp: fix "old stuff" D-SACK causing SACK to be treated as D-SACK

[ Upstream commit c9655008e7845bcfdaac10a1ed8554ec167aea88 ]

When we receive a D-SACK, where the sequence number satisfies:
undo_marker <= start_seq < end_seq <= prior_snd_una
we consider this is a valid D-SACK and tcp_is_sackblock_valid()
returns true, then this D-SACK is discarded as "old stuff",
but the variable first_sack_index is not marked as negative
in tcp_sacktag_write_queue().

If this D-SACK also carries a SACK that needs to be processed
(for example, the previous SACK segment was lost), this SACK
will be treated as a D-SACK in the following processing of
tcp_sacktag_write_queue(), which will eventually lead to
incorrect updates of undo_retrans and reordering.

Fixes: fd6dad616d4f ("[TCP]: Earlier SACK block verification & simplify access to them")
Signed-off-by: Pengcheng Yang <yangpc@wangsu.com>
Signed-off-by: Eric Dumazet <edumazet@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