tcp: lack of available data can also cause TSO defer
authorEric Dumazet <edumazet@google.com>
Thu, 6 Dec 2018 17:58:24 +0000 (09:58 -0800)
committerDavid S. Miller <davem@davemloft.net>
Sat, 8 Dec 2018 00:18:22 +0000 (16:18 -0800)
commitf9bfe4e6a9d08d405fe7b081ee9a13e649c97ecf
treef34a436bece7a3e4da77d6aa900ca8280fd53ab2
parent1b4e5ad5d6b9f15cd0b5121f86d4719165958417
tcp: lack of available data can also cause TSO defer

tcp_tso_should_defer() can return true in three different cases :

 1) We are cwnd-limited
 2) We are rwnd-limited
 3) We are application limited.

Neal pointed out that my recent fix went too far, since
it assumed that if we were not in 1) case, we must be rwnd-limited

Fix this by properly populating the is_cwnd_limited and
is_rwnd_limited booleans.

After this change, we can finally move the silly check for FIN
flag only for the application-limited case.

The same move for EOR bit will be handled in net-next,
since commit 1c09f7d073b1 ("tcp: do not try to defer skbs
with eor mark (MSG_EOR)") is scheduled for linux-4.21

Tested by running 200 concurrent netperf -t TCP_RR -- -r 60000,100
and checking none of them was rwnd_limited in the chrono_stat
output from "ss -ti" command.

Fixes: 41727549de3e ("tcp: Do not underestimate rwnd_limited")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Suggested-by: Neal Cardwell <ncardwell@google.com>
Reviewed-by: Neal Cardwell <ncardwell@google.com>
Acked-by: Soheil Hassas Yeganeh <soheil@google.com>
Reviewed-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
net/ipv4/tcp_output.c