tcp: invalidate rate samples during SACK reneging
authorYousuk Seung <ysseung@google.com>
Thu, 7 Dec 2017 21:41:34 +0000 (13:41 -0800)
committerDavid S. Miller <davem@davemloft.net>
Fri, 8 Dec 2017 15:07:02 +0000 (10:07 -0500)
commitd4761754b4fb2ef8d9a1e9d121c4bec84e1fe292
tree5a55983a564530e094904c68883a4ce77b69724b
parent195bd525d5f6e338b948d9a6b25bfaae86291353
tcp: invalidate rate samples during SACK reneging

Mark tcp_sock during a SACK reneging event and invalidate rate samples
while marked. Such rate samples may overestimate bw by including packets
that were SACKed before reneging.

< ack 6001 win 10000 sack 7001:38001
< ack 7001 win 0 sack 8001:38001 // Reneg detected
> seq 7001:8001 // RTO, SACK cleared.
< ack 38001 win 10000

In above example the rate sample taken after the last ack will count
7001-38001 as delivered while the actual delivery rate likely could
be much lower i.e. 7001-8001.

This patch adds a new field tcp_sock.sack_reneg and marks it when we
declare SACK reneging and entering TCP_CA_Loss, and unmarks it after
the last rate sample was taken before moving back to TCP_CA_Open. This
patch also invalidates rate samples taken while tcp_sock.is_sack_reneg
is set.

Fixes: b9f64820fb22 ("tcp: track data delivery rate for a TCP connection")
Signed-off-by: Yousuk Seung <ysseung@google.com>
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Acked-by: Soheil Hassas Yeganeh <soheil@google.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Acked-by: Priyaranjan Jha <priyarjha@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
include/linux/tcp.h
include/net/tcp.h
net/ipv4/tcp.c
net/ipv4/tcp_input.c
net/ipv4/tcp_rate.c