tcp: skip DSACKs with dubious sequence ranges
authorPriyaranjan Jha <priyarjha@google.com>
Thu, 24 Sep 2020 22:23:14 +0000 (15:23 -0700)
committerDavid S. Miller <davem@davemloft.net>
Fri, 25 Sep 2020 03:15:45 +0000 (20:15 -0700)
commitad2b9b0f8d0107602bdc1f3b1ab90719842ace11
tree13eee63f3542682121acde3c014c7d3423d58456
parent1ec8e74855588cecb2620b28b877c08f45765374
tcp: skip DSACKs with dubious sequence ranges

Currently, we use length of DSACKed range to compute number of
delivered packets. And if sequence range in DSACK is corrupted,
we can get bogus dsacked/acked count, and bogus cwnd.

This patch put bounds on DSACKed range to skip update of data
delivery and spurious retransmission information, if the DSACK
is unlikely caused by sender's action:
- DSACKed range shouldn't be greater than maximum advertised rwnd.
- Total no. of DSACKed segments shouldn't be greater than total
  no. of retransmitted segs. Unlike spurious retransmits, network
  duplicates or corrupted DSACKs shouldn't be counted as delivery.

Signed-off-by: Priyaranjan Jha <priyarjha@google.com>
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
include/uapi/linux/snmp.h
net/ipv4/proc.c
net/ipv4/tcp_input.c