tcp: fix cwnd limited checking to improve congestion control

Yuchung discovered tcp_is_cwnd_limited() was returning false in
slow start phase even if the application filled the socket write queue.

All congestion modules take into account tcp_is_cwnd_limited()
before increasing cwnd, so this behavior limits slow start from
probing the bandwidth at full speed.

The problem is that even if write queue is full (aka we are _not_
application limited), cwnd can be under utilized if TSO should auto
defer or TCP Small queues decided to hold packets.

So the in_flight can be kept to smaller value, and we can get to the
point tcp_is_cwnd_limited() returns false.

With TCP Small Queues and FQ/pacing, this issue is more visible.

We fix this by having tcp_cwnd_validate(), which is supposed to track
such things, take into account unsent_segs, the number of segs that we
are not sending at the moment due to TSO or TSQ, but intend to send
real soon. Then when we are cwnd-limited, remember this fact while we
are processing the window of ACKs that comes back.

For example, suppose we have a brand new connection with cwnd=10; we
are in slow start, and we send a flight of 9 packets. By the time we
have received ACKs for all 9 packets we want our cwnd to be 18.
We implement this by setting tp->lsnd_pending to 9, and
considering ourselves to be cwnd-limited while cwnd is less than
twice tp->lsnd_pending (2*9 -> 18).

This makes tcp_is_cwnd_limited() more understandable, by removing
the GSO/TSO kludge, that tried to work around the issue.

Note the in_flight parameter can be removed in a followup cleanup
patch.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

2 files changed, 22 insertions, 1 deletions

diff --git a/include/linux/tcp.h b/include/linux/tcp.h
index 239946868142..4e37c71ecd74 100644
--- a/include/linux/tcp.h
+++ b/include/linux/tcp.h
@@ -230,6 +230,7 @@ struct tcp_sock {
 	u32	snd_cwnd_clamp; /* Do not allow snd_cwnd to grow above this */
 	u32	snd_cwnd_used;
 	u32	snd_cwnd_stamp;
+	u32	lsnd_pending;	/* packets inflight or unsent since last xmit */
 	u32	prior_cwnd;	/* Congestion window at start of Recovery. */
 	u32	prr_delivered;	/* Number of newly delivered packets to
 				 * receiver in Recovery. */
diff --git a/include/net/tcp.h b/include/net/tcp.h
index 163d2b467d78..a9fe7bc4f4bb 100644
--- a/include/net/tcp.h
+++ b/include/net/tcp.h
@@ -974,7 +974,27 @@ static inline u32 tcp_wnd_end(const struct tcp_sock *tp)
 {
 	return tp->snd_una + tp->snd_wnd;
 }
-bool tcp_is_cwnd_limited(const struct sock *sk, u32 in_flight);
+
+/* We follow the spirit of RFC2861 to validate cwnd but implement a more
+ * flexible approach. The RFC suggests cwnd should not be raised unless
+ * it was fully used previously. But we allow cwnd to grow as long as the
+ * application has used half the cwnd.
+ * Example :
+ *    cwnd is 10 (IW10), but application sends 9 frames.
+ *    We allow cwnd to reach 18 when all frames are ACKed.
+ * This check is safe because it's as aggressive as slow start which already
+ * risks 100% overshoot. The advantage is that we discourage application to
+ * either send more filler packets or data to artificially blow up the cwnd
+ * usage, and allow application-limited process to probe bw more aggressively.
+ *
+ * TODO: remove in_flight once we can fix all callers, and their callers...
+ */
+static inline bool tcp_is_cwnd_limited(const struct sock *sk, u32 in_flight)
+{
+	const struct tcp_sock *tp = tcp_sk(sk);
+
+	return tp->snd_cwnd < 2 * tp->lsnd_pending;
+}
 
 static inline void tcp_check_probe_timer(struct sock *sk)
 {