Merge branch 'tcp_bbr-more-GSO-work'

Eric Dumazet says:

====================
tcp_bbr: more GSO work

Playing with r8152 USB 1Gbit NIC, on both USB2 and USB3 slots, I found
that BBR was performing poorly, because of TSO being limited to 16KB

This patch series makes sure BBR is not under estimating number of
packets that are needed to fill the pipe when a device has suboptimal
TSO limits.
====================

Signed-off-by: David S. Miller <[email protected]>

Author: davem330

include/net/tcp.h

@@ -511,8 +511,6 @@ __u32 cookie_v6_init_sequence(const struct sk_buff *skb, __u16 *mss);
 #endif
 /* tcp_output.c */
 
-u32 tcp_tso_autosize(const struct sock *sk, unsigned int mss_now,
-		     int min_tso_segs);
 void __tcp_push_pending_frames(struct sock *sk, unsigned int cur_mss,
 			       int nonagle);
 int __tcp_retransmit_skb(struct sock *sk, struct sk_buff *skb, int segs);
@@ -981,8 +979,8 @@ struct tcp_congestion_ops {
 	u32  (*undo_cwnd)(struct sock *sk);
 	/* hook for packet ack accounting (optional) */
 	void (*pkts_acked)(struct sock *sk, const struct ack_sample *sample);
-	/* suggest number of segments for each skb to transmit (optional) */
-	u32 (*tso_segs_goal)(struct sock *sk);
+	/* override sysctl_tcp_min_tso_segs */
+	u32 (*min_tso_segs)(struct sock *sk);
 	/* returns the multiplier used in tcp_sndbuf_expand (optional) */
 	u32 (*sndbuf_expand)(struct sock *sk);
 	/* call when packets are delivered to update cwnd and pacing rate,

net/ipv4/tcp_bbr.c

@@ -97,10 +97,9 @@ struct bbr {
 		packet_conservation:1,  /* use packet conservation? */
 		restore_cwnd:1,	     /* decided to revert cwnd to old value */
 		round_start:1,	     /* start of packet-timed tx->ack round? */
-		tso_segs_goal:7,     /* segments we want in each skb we send */
 		idle_restart:1,	     /* restarting after idle? */
 		probe_rtt_round_done:1,  /* a BBR_PROBE_RTT round at 4 pkts? */
-		unused:5,
+		unused:12,
 		lt_is_sampling:1,    /* taking long-term ("LT") samples now? */
 		lt_rtt_cnt:7,	     /* round trips in long-term interval */
 		lt_use_bw:1;	     /* use lt_bw as our bw estimate? */
@@ -261,23 +260,25 @@ static void bbr_set_pacing_rate(struct sock *sk, u32 bw, int gain)
 		sk->sk_pacing_rate = rate;
 }
 
-/* Return count of segments we want in the skbs we send, or 0 for default. */
-static u32 bbr_tso_segs_goal(struct sock *sk)
+/* override sysctl_tcp_min_tso_segs */
+static u32 bbr_min_tso_segs(struct sock *sk)
 {
-	struct bbr *bbr = inet_csk_ca(sk);
-
-	return bbr->tso_segs_goal;
+	return sk->sk_pacing_rate < (bbr_min_tso_rate >> 3) ? 1 : 2;
 }
 
-static void bbr_set_tso_segs_goal(struct sock *sk)
+static u32 bbr_tso_segs_goal(struct sock *sk)
 {
 	struct tcp_sock *tp = tcp_sk(sk);
-	struct bbr *bbr = inet_csk_ca(sk);
-	u32 min_segs;
+	u32 segs, bytes;
+
+	/* Sort of tcp_tso_autosize() but ignoring
+	 * driver provided sk_gso_max_size.
+	 */
+	bytes = min_t(u32, sk->sk_pacing_rate >> sk->sk_pacing_shift,
+		      GSO_MAX_SIZE - 1 - MAX_TCP_HEADER);
+	segs = max_t(u32, bytes / tp->mss_cache, bbr_min_tso_segs(sk));
 
-	min_segs = sk->sk_pacing_rate < (bbr_min_tso_rate >> 3) ? 1 : 2;
-	bbr->tso_segs_goal = min(tcp_tso_autosize(sk, tp->mss_cache, min_segs),
-				 0x7FU);
+	return min(segs, 0x7FU);
 }
 
 /* Save "last known good" cwnd so we can restore it after losses or PROBE_RTT */
@@ -348,7 +349,7 @@ static u32 bbr_target_cwnd(struct sock *sk, u32 bw, int gain)
 	cwnd = (((w * gain) >> BBR_SCALE) + BW_UNIT - 1) / BW_UNIT;
 
 	/* Allow enough full-sized skbs in flight to utilize end systems. */
-	cwnd += 3 * bbr->tso_segs_goal;
+	cwnd += 3 * bbr_tso_segs_goal(sk);
 
 	/* Reduce delayed ACKs by rounding up cwnd to the next even number. */
 	cwnd = (cwnd + 1) & ~1U;
@@ -824,7 +825,6 @@ static void bbr_main(struct sock *sk, const struct rate_sample *rs)
 
 	bw = bbr_bw(sk);
 	bbr_set_pacing_rate(sk, bw, bbr->pacing_gain);
-	bbr_set_tso_segs_goal(sk);
 	bbr_set_cwnd(sk, rs, rs->acked_sacked, bw, bbr->cwnd_gain);
 }
 
@@ -834,7 +834,6 @@ static void bbr_init(struct sock *sk)
 	struct bbr *bbr = inet_csk_ca(sk);
 
 	bbr->prior_cwnd = 0;
-	bbr->tso_segs_goal = 0;	 /* default segs per skb until first ACK */
 	bbr->rtt_cnt = 0;
 	bbr->next_rtt_delivered = 0;
 	bbr->prev_ca_state = TCP_CA_Open;
@@ -936,7 +935,7 @@ static struct tcp_congestion_ops tcp_bbr_cong_ops __read_mostly = {
 	.undo_cwnd	= bbr_undo_cwnd,
 	.cwnd_event	= bbr_cwnd_event,
 	.ssthresh	= bbr_ssthresh,
-	.tso_segs_goal	= bbr_tso_segs_goal,
+	.min_tso_segs	= bbr_min_tso_segs,
 	.get_info	= bbr_get_info,
 	.set_state	= bbr_set_state,
 };

net/ipv4/tcp_output.c

@@ -1703,8 +1703,8 @@ static bool tcp_nagle_check(bool partial, const struct tcp_sock *tp,
 /* Return how many segs we'd like on a TSO packet,
  * to send one TSO packet per ms
  */
-u32 tcp_tso_autosize(const struct sock *sk, unsigned int mss_now,
-		     int min_tso_segs)
+static u32 tcp_tso_autosize(const struct sock *sk, unsigned int mss_now,
+			    int min_tso_segs)
 {
 	u32 bytes, segs;
 
@@ -1720,19 +1720,20 @@ u32 tcp_tso_autosize(const struct sock *sk, unsigned int mss_now,
 
 	return segs;
 }
-EXPORT_SYMBOL(tcp_tso_autosize);
 
 /* Return the number of segments we want in the skb we are transmitting.
  * See if congestion control module wants to decide; otherwise, autosize.
  */
 static u32 tcp_tso_segs(struct sock *sk, unsigned int mss_now)
 {
 	const struct tcp_congestion_ops *ca_ops = inet_csk(sk)->icsk_ca_ops;
-	u32 tso_segs = ca_ops->tso_segs_goal ? ca_ops->tso_segs_goal(sk) : 0;
+	u32 min_tso, tso_segs;
 
-	if (!tso_segs)
-		tso_segs = tcp_tso_autosize(sk, mss_now,
-				sock_net(sk)->ipv4.sysctl_tcp_min_tso_segs);
+	min_tso = ca_ops->min_tso_segs ?
+			ca_ops->min_tso_segs(sk) :
+			sock_net(sk)->ipv4.sysctl_tcp_min_tso_segs;
+
+	tso_segs = tcp_tso_autosize(sk, mss_now, min_tso);
 	return min_t(u32, tso_segs, sk->sk_gso_max_segs);
 }