Skip to contentSkip to conversation

tcp: drop reason additions

PATCH Applied Eric Dumazet 2 participants 12 messages Apr 16, 2022 – Apr 17, 2022

Subsystems (4)
lore
Share & tools
lore
#0 lore raw

cover tcp: drop reason additions

From: Eric Dumazet <[email protected]>

Currently, TCP is either missing drop reasons, or pretending that some useful packets are dropped.

This patch series makes "perf record -a -e skb:kfree_skb" much more usable.

Eric Dumazet (10):
tcp: consume incoming skb leading to a reset
tcp: get rid of rst_seq_match
tcp: add drop reason support to tcp_validate_incoming()
tcp: make tcp_rcv_state_process() drop monitor friendly
tcp: add drop reasons to tcp_rcv_state_process()
tcp: add two drop reasons for tcp_ack()
tcp: add drop reason support to tcp_prune_ofo_queue()
tcp: make tcp_rcv_synsent_state_process() drop monitor friend
tcp: add drop reasons to tcp_rcv_synsent_state_process()
tcp: add drop reason support to tcp_ofo_queue()

 include/linux/skbuff.h     |  13 ++++
 include/trace/events/skb.h |  14 ++++
 net/ipv4/tcp_input.c       | 127 +++++++++++++++++++++----------------
 3 files changed, 100 insertions(+), 54 deletions(-)
Signature · 1 line
2.36.0.rc0.470.gd361397f0d-goog
#1 lore raw

PATCH 1/10 tcp: consume incoming skb leading to a reset

From: Eric Dumazet <[email protected]>

Whenever tcp_validate_incoming() handles a valid RST packet, we should not pretend the packet was dropped.

Create a special section at the end of tcp_validate_incoming() to handle this case.

---
 net/ipv4/tcp_input.c | 28 ++++++++++++++++------------
 1 file changed, 16 insertions(+), 12 deletions(-)
net/ipv4/tcp_input.c +16 -12
index 1595b76ea2bea195b8896f4cd533beb14b56dd96..d4bc717791704795a9a159baf4ccd501a9471609 100644
@@ -5700,7 +5700,7 @@ static bool tcp_validate_incoming(struct sock *sk, struct sk_buff *skb,
 						  &tp->last_oow_ack_time))
 				tcp_send_dupack(sk, skb);
 		} else if (tcp_reset_check(sk, skb)) {
-			tcp_reset(sk, skb);
+			goto reset;
 		}
 		goto discard;
 	}
@@ -5736,17 +5736,16 @@ static bool tcp_validate_incoming(struct sock *sk, struct sk_buff *skb,
 		}
 
 		if (rst_seq_match)
-			tcp_reset(sk, skb);
-		else {
-			/* Disable TFO if RST is out-of-order
-			 * and no data has been received
-			 * for current active TFO socket
-			 */
-			if (tp->syn_fastopen && !tp->data_segs_in &&
-			    sk->sk_state == TCP_ESTABLISHED)
-				tcp_fastopen_active_disable(sk);
-			tcp_send_challenge_ack(sk);
-		}
+			goto reset;
+
+		/* Disable TFO if RST is out-of-order
+		 * and no data has been received
+		 * for current active TFO socket
+		 */
+		if (tp->syn_fastopen && !tp->data_segs_in &&
+		    sk->sk_state == TCP_ESTABLISHED)
+			tcp_fastopen_active_disable(sk);
+		tcp_send_challenge_ack(sk);
 		goto discard;
 	}
 
@@ -5771,6 +5770,11 @@ static bool tcp_validate_incoming(struct sock *sk, struct sk_buff *skb,
 discard:
 	tcp_drop(sk, skb);
 	return false;
+
+reset:
+	tcp_reset(sk, skb);
+	__kfree_skb(skb);
+	return false;
 }
 
 /*
Signature · 1 line
2.36.0.rc0.470.gd361397f0d-goog
#2 lore raw

PATCH 2/10 tcp: get rid of rst_seq_match

In reply to Eric Dumazet

From: Eric Dumazet <[email protected]>

Small cleanup in tcp_validate_incoming(), no need for rst_seq_match setting and testing.

---
 net/ipv4/tcp_input.c | 13 +++++--------
 1 file changed, 5 insertions(+), 8 deletions(-)
net/ipv4/tcp_input.c +5 -8
index d4bc717791704795a9a159baf4ccd501a9471609..b2d5fbef6ce3baa9426b3c9750002317a8915596 100644
@@ -5667,7 +5667,6 @@ static bool tcp_validate_incoming(struct sock *sk, struct sk_buff *skb,
 				  const struct tcphdr *th, int syn_inerr)
 {
 	struct tcp_sock *tp = tcp_sk(sk);
-	bool rst_seq_match = false;
 
 	/* RFC1323: H1. Apply PAWS check first. */
 	if (tcp_fast_parse_options(sock_net(sk), skb, th, tp) &&
@@ -5717,9 +5716,10 @@ static bool tcp_validate_incoming(struct sock *sk, struct sk_buff *skb,
 		 *     Send a challenge ACK
 		 */
 		if (TCP_SKB_CB(skb)->seq == tp->rcv_nxt ||
-		    tcp_reset_check(sk, skb)) {
-			rst_seq_match = true;
-		} else if (tcp_is_sack(tp) && tp->rx_opt.num_sacks > 0) {
+		    tcp_reset_check(sk, skb))
+			goto reset;
+
+		if (tcp_is_sack(tp) && tp->rx_opt.num_sacks > 0) {
 			struct tcp_sack_block *sp = &tp->selective_acks[0];
 			int max_sack = sp[0].end_seq;
 			int this_sack;
@@ -5732,12 +5732,9 @@ static bool tcp_validate_incoming(struct sock *sk, struct sk_buff *skb,
 			}
 
 			if (TCP_SKB_CB(skb)->seq == max_sack)
-				rst_seq_match = true;
+				goto reset;
 		}
 
-		if (rst_seq_match)
-			goto reset;
-
 		/* Disable TFO if RST is out-of-order
 		 * and no data has been received
 		 * for current active TFO socket
Signature · 1 line
2.36.0.rc0.470.gd361397f0d-goog
#3 lore raw

PATCH 3/10 tcp: add drop reason support to tcp_validate_incoming()

In reply to Eric Dumazet

From: Eric Dumazet <[email protected]>

Creates four new drop reasons for the following cases:

1) packet being rejected by RFC 7323 PAWS check
2) packet being rejected by SEQUENCE check
3) Invalid RST packet
4) Invalid SYN packet

---
 include/linux/skbuff.h     | 6 ++++++
 include/trace/events/skb.h | 5 +++++
 net/ipv4/tcp_input.c       | 7 ++++++-
 3 files changed, 17 insertions(+), 1 deletion(-)
include/linux/skbuff.h +6 -0
index 0ef11df1bc67f26a454a809396bd93299ce787ad..a903da1fa0ed897ba65a3edf6d74d7e5dc575b2e 100644
@@ -381,6 +381,12 @@ enum skb_drop_reason {
 					 * the ofo queue, corresponding to
 					 * LINUX_MIB_TCPOFOMERGE
 					 */
+	SKB_DROP_REASON_TCP_RFC7323_PAWS, /* PAWS check, corresponding to
+					   * LINUX_MIB_PAWSESTABREJECTED
+					   */
+	SKB_DROP_REASON_TCP_INVALID_SEQUENCE, /* Not acceptable SEQ field */
+	SKB_DROP_REASON_TCP_RESET,	/* Invalid RST packet */
+	SKB_DROP_REASON_TCP_INVALID_SYN, /* Incoming packet has unexpected SYN flag */
 	SKB_DROP_REASON_IP_OUTNOROUTES,	/* route lookup failed */
 	SKB_DROP_REASON_BPF_CGROUP_EGRESS,	/* dropped by
 						 * BPF_PROG_TYPE_CGROUP_SKB
include/trace/events/skb.h +5 -0
index 2da72a9a576462bee9f3415141dfffd2eec8c258..820dacd14bad9ecb2b8ff6206cb33b392c0c442c 100644
@@ -37,6 +37,11 @@
 	EM(SKB_DROP_REASON_TCP_OLD_DATA, TCP_OLD_DATA)		\
 	EM(SKB_DROP_REASON_TCP_OVERWINDOW, TCP_OVERWINDOW)	\
 	EM(SKB_DROP_REASON_TCP_OFOMERGE, TCP_OFOMERGE)		\
+	EM(SKB_DROP_REASON_TCP_RFC7323_PAWS, TCP_RFC7323_PAWS)	\
+	EM(SKB_DROP_REASON_TCP_INVALID_SEQUENCE,		\
+	   TCP_INVALID_SEQUENCE)				\
+	EM(SKB_DROP_REASON_TCP_RESET, TCP_RESET)		\
+	EM(SKB_DROP_REASON_TCP_INVALID_SYN, TCP_INVALID_SYN)	\
 	EM(SKB_DROP_REASON_IP_OUTNOROUTES, IP_OUTNOROUTES)	\
 	EM(SKB_DROP_REASON_BPF_CGROUP_EGRESS,			\
 	   BPF_CGROUP_EGRESS)					\
net/ipv4/tcp_input.c +6 -1
index b2d5fbef6ce3baa9426b3c9750002317a8915596..9a1cb3f48c3fb26beac4283001d38828ca15a4d9 100644
@@ -5667,6 +5667,7 @@ static bool tcp_validate_incoming(struct sock *sk, struct sk_buff *skb,
 				  const struct tcphdr *th, int syn_inerr)
 {
 	struct tcp_sock *tp = tcp_sk(sk);
+	SKB_DR(reason);
 
 	/* RFC1323: H1. Apply PAWS check first. */
 	if (tcp_fast_parse_options(sock_net(sk), skb, th, tp) &&
@@ -5678,6 +5679,7 @@ static bool tcp_validate_incoming(struct sock *sk, struct sk_buff *skb,
 						  LINUX_MIB_TCPACKSKIPPEDPAWS,
 						  &tp->last_oow_ack_time))
 				tcp_send_dupack(sk, skb);
+			SKB_DR_SET(reason, TCP_RFC7323_PAWS);
 			goto discard;
 		}
 		/* Reset is accepted even if it did not pass PAWS. */
@@ -5701,6 +5703,7 @@ static bool tcp_validate_incoming(struct sock *sk, struct sk_buff *skb,
 		} else if (tcp_reset_check(sk, skb)) {
 			goto reset;
 		}
+		SKB_DR_SET(reason, TCP_INVALID_SEQUENCE);
 		goto discard;
 	}
 
@@ -5743,6 +5746,7 @@ static bool tcp_validate_incoming(struct sock *sk, struct sk_buff *skb,
 		    sk->sk_state == TCP_ESTABLISHED)
 			tcp_fastopen_active_disable(sk);
 		tcp_send_challenge_ack(sk);
+		SKB_DR_SET(reason, TCP_RESET);
 		goto discard;
 	}
 
@@ -5757,6 +5761,7 @@ static bool tcp_validate_incoming(struct sock *sk, struct sk_buff *skb,
 			TCP_INC_STATS(sock_net(sk), TCP_MIB_INERRS);
 		NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPSYNCHALLENGE);
 		tcp_send_challenge_ack(sk);
+		SKB_DR_SET(reason, TCP_INVALID_SYN);
 		goto discard;
 	}
 
@@ -5765,7 +5770,7 @@ static bool tcp_validate_incoming(struct sock *sk, struct sk_buff *skb,
 	return true;
 
 discard:
-	tcp_drop(sk, skb);
+	tcp_drop_reason(sk, skb, reason);
 	return false;
 
 reset:
Signature · 1 line
2.36.0.rc0.470.gd361397f0d-goog
#4 lore raw

PATCH 4/10 tcp: make tcp_rcv_state_process() drop monitor friendly

In reply to Eric Dumazet

From: Eric Dumazet <[email protected]>

tcp_rcv_state_process() incorrectly drops packets instead of consuming it, making drop monitor very noisy, if not unusable.

Calling tcp_time_wait() or tcp_done() is part of standard behavior, packets triggering these actions were not dropped.

---
 net/ipv4/tcp_input.c | 10 +++++++---
 1 file changed, 7 insertions(+), 3 deletions(-)
net/ipv4/tcp_input.c +7 -3
index 9a1cb3f48c3fb26beac4283001d38828ca15a4d9..f95a8368981d319400d0f46b81ced954d941f718 100644
@@ -6580,7 +6580,7 @@ int tcp_rcv_state_process(struct sock *sk, struct sk_buff *skb)
 			inet_csk_reset_keepalive_timer(sk, tmo);
 		} else {
 			tcp_time_wait(sk, TCP_FIN_WAIT2, tmo);
-			goto discard;
+			goto consume;
 		}
 		break;
 	}
@@ -6588,7 +6588,7 @@ int tcp_rcv_state_process(struct sock *sk, struct sk_buff *skb)
 	case TCP_CLOSING:
 		if (tp->snd_una == tp->write_seq) {
 			tcp_time_wait(sk, TCP_TIME_WAIT, 0);
-			goto discard;
+			goto consume;
 		}
 		break;
 
@@ -6596,7 +6596,7 @@ int tcp_rcv_state_process(struct sock *sk, struct sk_buff *skb)
 		if (tp->snd_una == tp->write_seq) {
 			tcp_update_metrics(sk);
 			tcp_done(sk);
-			goto discard;
+			goto consume;
 		}
 		break;
 	}
@@ -6650,6 +6650,10 @@ int tcp_rcv_state_process(struct sock *sk, struct sk_buff *skb)
 		tcp_drop(sk, skb);
 	}
 	return 0;
+
+consume:
+	__kfree_skb(skb);
+	return 0;
 }
 EXPORT_SYMBOL(tcp_rcv_state_process);
 
Signature · 1 line
2.36.0.rc0.470.gd361397f0d-goog
#5 lore raw

PATCH 5/10 tcp: add drop reasons to tcp_rcv_state_process()

In reply to Eric Dumazet

From: Eric Dumazet <[email protected]>

Add basic support for drop reasons in tcp_rcv_state_process()

---
 include/linux/skbuff.h     |  3 +++
 include/trace/events/skb.h |  3 +++
 net/ipv4/tcp_input.c       | 24 +++++++++++++++++-------
 3 files changed, 23 insertions(+), 7 deletions(-)
include/linux/skbuff.h +3 -0
index a903da1fa0ed897ba65a3edf6d74d7e5dc575b2e..6f1410b5ff1372d9e1f7a75c303f8d7876c83ef0 100644
@@ -387,6 +387,9 @@ enum skb_drop_reason {
 	SKB_DROP_REASON_TCP_INVALID_SEQUENCE, /* Not acceptable SEQ field */
 	SKB_DROP_REASON_TCP_RESET,	/* Invalid RST packet */
 	SKB_DROP_REASON_TCP_INVALID_SYN, /* Incoming packet has unexpected SYN flag */
+	SKB_DROP_REASON_TCP_CLOSE,	/* TCP socket in CLOSE state */
+	SKB_DROP_REASON_TCP_FASTOPEN,	/* dropped by FASTOPEN request socket */
+	SKB_DROP_REASON_TCP_OLD_ACK,	/* TCP ACK is old, but in window */
 	SKB_DROP_REASON_IP_OUTNOROUTES,	/* route lookup failed */
 	SKB_DROP_REASON_BPF_CGROUP_EGRESS,	/* dropped by
 						 * BPF_PROG_TYPE_CGROUP_SKB
include/trace/events/skb.h +3 -0
index 820dacd14bad9ecb2b8ff6206cb33b392c0c442c..fbe21ad038bc6701ed04cb6be417c544de0dae84 100644
@@ -42,6 +42,9 @@
 	   TCP_INVALID_SEQUENCE)				\
 	EM(SKB_DROP_REASON_TCP_RESET, TCP_RESET)		\
 	EM(SKB_DROP_REASON_TCP_INVALID_SYN, TCP_INVALID_SYN)	\
+	EM(SKB_DROP_REASON_TCP_CLOSE, TCP_CLOSE)		\
+	EM(SKB_DROP_REASON_TCP_FASTOPEN, TCP_FASTOPEN)		\
+	EM(SKB_DROP_REASON_TCP_OLD_ACK, TCP_OLD_ACK)		\
 	EM(SKB_DROP_REASON_IP_OUTNOROUTES, IP_OUTNOROUTES)	\
 	EM(SKB_DROP_REASON_BPF_CGROUP_EGRESS,			\
 	   BPF_CGROUP_EGRESS)					\
net/ipv4/tcp_input.c +17 -7
index f95a8368981d319400d0f46b81ced954d941f718..85fae79c894d4b96820747c658bb4d884981c49e 100644
@@ -6413,21 +6413,26 @@ int tcp_rcv_state_process(struct sock *sk, struct sk_buff *skb)
 	struct request_sock *req;
 	int queued = 0;
 	bool acceptable;
+	SKB_DR(reason);
 
 	switch (sk->sk_state) {
 	case TCP_CLOSE:
+		SKB_DR_SET(reason, TCP_CLOSE);
 		goto discard;
 
 	case TCP_LISTEN:
 		if (th->ack)
 			return 1;
 
-		if (th->rst)
+		if (th->rst) {
+			SKB_DR_SET(reason, TCP_RESET);
 			goto discard;
-
+		}
 		if (th->syn) {
-			if (th->fin)
+			if (th->fin) {
+				SKB_DR_SET(reason, TCP_FLAGS);
 				goto discard;
+			}
 			/* It is possible that we process SYN packets from backlog,
 			 * so we need to make sure to disable BH and RCU right there.
 			 */
@@ -6442,6 +6447,7 @@ int tcp_rcv_state_process(struct sock *sk, struct sk_buff *skb)
 			consume_skb(skb);
 			return 0;
 		}
+		SKB_DR_SET(reason, TCP_FLAGS);
 		goto discard;
 
 	case TCP_SYN_SENT:
@@ -6468,13 +6474,16 @@ int tcp_rcv_state_process(struct sock *sk, struct sk_buff *skb)
 		WARN_ON_ONCE(sk->sk_state != TCP_SYN_RECV &&
 		    sk->sk_state != TCP_FIN_WAIT1);
 
-		if (!tcp_check_req(sk, skb, req, true, &req_stolen))
+		if (!tcp_check_req(sk, skb, req, true, &req_stolen)) {
+			SKB_DR_SET(reason, TCP_FASTOPEN);
 			goto discard;
+		}
 	}
 
-	if (!th->ack && !th->rst && !th->syn)
+	if (!th->ack && !th->rst && !th->syn) {
+		SKB_DR_SET(reason, TCP_FLAGS);
 		goto discard;
-
+	}
 	if (!tcp_validate_incoming(sk, skb, th, 0))
 		return 0;
 
@@ -6487,6 +6496,7 @@ int tcp_rcv_state_process(struct sock *sk, struct sk_buff *skb)
 		if (sk->sk_state == TCP_SYN_RECV)
 			return 1;	/* send one RST */
 		tcp_send_challenge_ack(sk);
+		SKB_DR_SET(reason, TCP_OLD_ACK);
 		goto discard;
 	}
 	switch (sk->sk_state) {
@@ -6647,7 +6657,7 @@ int tcp_rcv_state_process(struct sock *sk, struct sk_buff *skb)
 
 	if (!queued) {
 discard:
-		tcp_drop(sk, skb);
+		tcp_drop_reason(sk, skb, reason);
 	}
 	return 0;
 
Signature · 1 line
2.36.0.rc0.470.gd361397f0d-goog
#6 lore raw

PATCH 6/10 tcp: add two drop reasons for tcp_ack()

In reply to Eric Dumazet

From: Eric Dumazet <[email protected]>

Add TCP_TOO_OLD_ACK and TCP_ACK_UNSENT_DATA drop
reasons so that tcp_rcv_established() can report
them.

---
 include/linux/skbuff.h     | 2 ++
 include/trace/events/skb.h | 3 +++
 net/ipv4/tcp_input.c       | 7 ++++---
 3 files changed, 9 insertions(+), 3 deletions(-)
include/linux/skbuff.h +2 -0
index 6f1410b5ff1372d9e1f7a75c303f8d7876c83ef0..9ff5557b190905f716a25f67113c1db822707959 100644
@@ -390,6 +390,8 @@ enum skb_drop_reason {
 	SKB_DROP_REASON_TCP_CLOSE,	/* TCP socket in CLOSE state */
 	SKB_DROP_REASON_TCP_FASTOPEN,	/* dropped by FASTOPEN request socket */
 	SKB_DROP_REASON_TCP_OLD_ACK,	/* TCP ACK is old, but in window */
+	SKB_DROP_REASON_TCP_TOO_OLD_ACK, /* TCP ACK is too old */
+	SKB_DROP_REASON_TCP_ACK_UNSENT_DATA, /* TCP ACK for data we haven't sent yet */
 	SKB_DROP_REASON_IP_OUTNOROUTES,	/* route lookup failed */
 	SKB_DROP_REASON_BPF_CGROUP_EGRESS,	/* dropped by
 						 * BPF_PROG_TYPE_CGROUP_SKB
include/trace/events/skb.h +3 -0
index fbe21ad038bc6701ed04cb6be417c544de0dae84..eab0b09223f3c66255f930d44be61d45e83ca6e8 100644
@@ -45,6 +45,9 @@
 	EM(SKB_DROP_REASON_TCP_CLOSE, TCP_CLOSE)		\
 	EM(SKB_DROP_REASON_TCP_FASTOPEN, TCP_FASTOPEN)		\
 	EM(SKB_DROP_REASON_TCP_OLD_ACK, TCP_OLD_ACK)		\
+	EM(SKB_DROP_REASON_TCP_TOO_OLD_ACK, TCP_TOO_OLD_ACK)	\
+	EM(SKB_DROP_REASON_TCP_ACK_UNSENT_DATA,			\
+	   TCP_ACK_UNSENT_DATA)					\
 	EM(SKB_DROP_REASON_IP_OUTNOROUTES, IP_OUTNOROUTES)	\
 	EM(SKB_DROP_REASON_BPF_CGROUP_EGRESS,			\
 	   BPF_CGROUP_EGRESS)					\
net/ipv4/tcp_input.c +4 -3
index 85fae79c894d4b96820747c658bb4d884981c49e..8a68785b04053b8e622404035284920e51cd953c 100644
@@ -3766,7 +3766,7 @@ static int tcp_ack(struct sock *sk, const struct sk_buff *skb, int flag)
 		if (before(ack, prior_snd_una - tp->max_window)) {
 			if (!(flag & FLAG_NO_CHALLENGE_ACK))
 				tcp_send_challenge_ack(sk);
-			return -1;
+			return -SKB_DROP_REASON_TCP_TOO_OLD_ACK;
 		}
 		goto old_ack;
 	}
@@ -3775,7 +3775,7 @@ static int tcp_ack(struct sock *sk, const struct sk_buff *skb, int flag)
 	 * this segment (RFC793 Section 3.9).
 	 */
 	if (after(ack, tp->snd_nxt))
-		return -1;
+		return -SKB_DROP_REASON_TCP_ACK_UNSENT_DATA;
 
 	if (after(ack, prior_snd_una)) {
 		flag |= FLAG_SND_UNA_ADVANCED;
@@ -5962,7 +5962,8 @@ void tcp_rcv_established(struct sock *sk, struct sk_buff *skb)
 		return;
 
 step5:
-	if (tcp_ack(sk, skb, FLAG_SLOWPATH | FLAG_UPDATE_TS_RECENT) < 0)
+	reason = tcp_ack(sk, skb, FLAG_SLOWPATH | FLAG_UPDATE_TS_RECENT);
+	if (reason < 0)
 		goto discard;
 
 	tcp_rcv_rtt_measure_ts(sk, skb);
Signature · 1 line
2.36.0.rc0.470.gd361397f0d-goog
#7 lore raw

PATCH 7/10 tcp: add drop reason support to tcp_prune_ofo_queue()

In reply to Eric Dumazet

From: Eric Dumazet <[email protected]>

Add one reason for packets dropped from OFO queue because of memory pressure.

---
 include/linux/skbuff.h     | 1 +
 include/trace/events/skb.h | 2 ++
 net/ipv4/tcp_input.c       | 3 ++-
 3 files changed, 5 insertions(+), 1 deletion(-)
include/linux/skbuff.h +1 -0
index 9ff5557b190905f716a25f67113c1db822707959..ad15ad208b5612357546f23dfe4feaa1535f4982 100644
@@ -392,6 +392,7 @@ enum skb_drop_reason {
 	SKB_DROP_REASON_TCP_OLD_ACK,	/* TCP ACK is old, but in window */
 	SKB_DROP_REASON_TCP_TOO_OLD_ACK, /* TCP ACK is too old */
 	SKB_DROP_REASON_TCP_ACK_UNSENT_DATA, /* TCP ACK for data we haven't sent yet */
+	SKB_DROP_REASON_TCP_OFO_QUEUE_PRUNE, /* pruned from TCP OFO queue */
 	SKB_DROP_REASON_IP_OUTNOROUTES,	/* route lookup failed */
 	SKB_DROP_REASON_BPF_CGROUP_EGRESS,	/* dropped by
 						 * BPF_PROG_TYPE_CGROUP_SKB
include/trace/events/skb.h +2 -0
index eab0b09223f3c66255f930d44be61d45e83ca6e8..73d7a6e594cbc6884148afa473bc08d12d783079 100644
@@ -48,6 +48,8 @@
 	EM(SKB_DROP_REASON_TCP_TOO_OLD_ACK, TCP_TOO_OLD_ACK)	\
 	EM(SKB_DROP_REASON_TCP_ACK_UNSENT_DATA,			\
 	   TCP_ACK_UNSENT_DATA)					\
+	EM(SKB_DROP_REASON_TCP_OFO_QUEUE_PRUNE,			\
+	  TCP_OFO_QUEUE_PRUNE)					\
 	EM(SKB_DROP_REASON_IP_OUTNOROUTES, IP_OUTNOROUTES)	\
 	EM(SKB_DROP_REASON_BPF_CGROUP_EGRESS,			\
 	   BPF_CGROUP_EGRESS)					\
net/ipv4/tcp_input.c +2 -1
index 8a68785b04053b8e622404035284920e51cd953c..a1077adeb1b62d74b5f1c9a6fc34def44ae07790 100644
@@ -5334,7 +5334,8 @@ static bool tcp_prune_ofo_queue(struct sock *sk)
 		prev = rb_prev(node);
 		rb_erase(node, &tp->out_of_order_queue);
 		goal -= rb_to_skb(node)->truesize;
-		tcp_drop(sk, rb_to_skb(node));
+		tcp_drop_reason(sk, rb_to_skb(node),
+				SKB_DROP_REASON_TCP_OFO_QUEUE_PRUNE);
 		if (!prev || goal <= 0) {
 			sk_mem_reclaim(sk);
 			if (atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf &&
Signature · 1 line
2.36.0.rc0.470.gd361397f0d-goog
#8 lore raw

PATCH 8/10 tcp: make tcp_rcv_synsent_state_process() drop monitor friend

In reply to Eric Dumazet

From: Eric Dumazet <[email protected]>

1) A valid RST packet should be consumed, to not confuse drop monitor.

2) Same remark for packet validating cross syn setup, even if we might ignore part of it.

3) When third packet of 3WHS is delayed, do not pretend the SYNACK was dropped.

---
 net/ipv4/tcp_input.c | 17 ++++++++---------
 1 file changed, 8 insertions(+), 9 deletions(-)
net/ipv4/tcp_input.c +8 -9
index a1077adeb1b62d74b5f1c9a6fc34def44ae07790..cd9f5c39f85a042751ef78132860a2a6cc96bccc 100644
@@ -6186,7 +6186,9 @@ static int tcp_rcv_synsent_state_process(struct sock *sk, struct sk_buff *skb,
 
 		if (th->rst) {
 			tcp_reset(sk, skb);
-			goto discard;
+consume:
+			__kfree_skb(skb);
+			return 0;
 		}
 
 		/* rfc793:
@@ -6275,13 +6277,9 @@ static int tcp_rcv_synsent_state_process(struct sock *sk, struct sk_buff *skb,
 			tcp_enter_quickack_mode(sk, TCP_MAX_QUICKACKS);
 			inet_csk_reset_xmit_timer(sk, ICSK_TIME_DACK,
 						  TCP_DELACK_MAX, TCP_RTO_MAX);
-
-discard:
-			tcp_drop(sk, skb);
-			return 0;
-		} else {
-			tcp_send_ack(sk);
+			goto consume;
 		}
+		tcp_send_ack(sk);
 		return -1;
 	}
 
@@ -6350,7 +6348,7 @@ static int tcp_rcv_synsent_state_process(struct sock *sk, struct sk_buff *skb,
 		 */
 		return -1;
 #else
-		goto discard;
+		goto consume;
 #endif
 	}
 	/* "fifth, if neither of the SYN or RST bits is set then
@@ -6360,7 +6358,8 @@ static int tcp_rcv_synsent_state_process(struct sock *sk, struct sk_buff *skb,
 discard_and_undo:
 	tcp_clear_options(&tp->rx_opt);
 	tp->rx_opt.mss_clamp = saved_clamp;
-	goto discard;
+	tcp_drop(sk, skb);
+	return 0;
 
 reset_and_undo:
 	tcp_clear_options(&tp->rx_opt);
Signature · 1 line
2.36.0.rc0.470.gd361397f0d-goog
#9 lore raw

PATCH 9/10 tcp: add drop reasons to tcp_rcv_synsent_state_process()

In reply to Eric Dumazet

From: Eric Dumazet <[email protected]>

Re-use existing reasons.

---
 net/ipv4/tcp_input.c | 15 +++++++++------
 1 file changed, 9 insertions(+), 6 deletions(-)
net/ipv4/tcp_input.c +9 -6
index cd9f5c39f85a042751ef78132860a2a6cc96bccc..339cc3d40745a0ea2a9f66b03dfda5aa6800d4a2 100644
@@ -6144,6 +6144,7 @@ static int tcp_rcv_synsent_state_process(struct sock *sk, struct sk_buff *skb,
 	struct tcp_fastopen_cookie foc = { .len = -1 };
 	int saved_clamp = tp->rx_opt.mss_clamp;
 	bool fastopen_fail;
+	SKB_DR(reason);
 
 	tcp_parse_options(sock_net(sk), skb, &tp->rx_opt, 0, &foc);
 	if (tp->rx_opt.saw_tstamp && tp->rx_opt.rcv_tsecr)
@@ -6198,9 +6199,10 @@ static int tcp_rcv_synsent_state_process(struct sock *sk, struct sk_buff *skb,
 		 *    See note below!
 		 *                                        --ANK(990513)
 		 */
-		if (!th->syn)
+		if (!th->syn) {
+			SKB_DR_SET(reason, TCP_FLAGS);
 			goto discard_and_undo;
-
+		}
 		/* rfc793:
 		 *   "If the SYN bit is on ...
 		 *    are acceptable then ...
@@ -6291,15 +6293,16 @@ static int tcp_rcv_synsent_state_process(struct sock *sk, struct sk_buff *skb,
 		 *
 		 *      Otherwise (no ACK) drop the segment and return."
 		 */
-
+		SKB_DR_SET(reason, TCP_RESET);
 		goto discard_and_undo;
 	}
 
 	/* PAWS check. */
 	if (tp->rx_opt.ts_recent_stamp && tp->rx_opt.saw_tstamp &&
-	    tcp_paws_reject(&tp->rx_opt, 0))
+	    tcp_paws_reject(&tp->rx_opt, 0)) {
+		SKB_DR_SET(reason, TCP_RFC7323_PAWS);
 		goto discard_and_undo;
-
+	}
 	if (th->syn) {
 		/* We see SYN without ACK. It is attempt of
 		 * simultaneous connect with crossed SYNs.
@@ -6358,7 +6361,7 @@ static int tcp_rcv_synsent_state_process(struct sock *sk, struct sk_buff *skb,
 discard_and_undo:
 	tcp_clear_options(&tp->rx_opt);
 	tp->rx_opt.mss_clamp = saved_clamp;
-	tcp_drop(sk, skb);
+	tcp_drop_reason(sk, skb, reason);
 	return 0;
 
 reset_and_undo:
Signature · 1 line
2.36.0.rc0.470.gd361397f0d-goog
#10 lore raw

PATCH 10/10 tcp: add drop reason support to tcp_ofo_queue()

In reply to Eric Dumazet

From: Eric Dumazet <[email protected]>

packets in OFO queue might be redundant, and dropped.

tcp_drop() is no longer needed.

---
 include/linux/skbuff.h     | 1 +
 include/trace/events/skb.h | 1 +
 net/ipv4/tcp_input.c       | 9 ++-------
 3 files changed, 4 insertions(+), 7 deletions(-)
include/linux/skbuff.h +1 -0
index ad15ad208b5612357546f23dfe4feaa1535f4982..84d78df60453955a8eaf05847f6e2145176a727a 100644
@@ -393,6 +393,7 @@ enum skb_drop_reason {
 	SKB_DROP_REASON_TCP_TOO_OLD_ACK, /* TCP ACK is too old */
 	SKB_DROP_REASON_TCP_ACK_UNSENT_DATA, /* TCP ACK for data we haven't sent yet */
 	SKB_DROP_REASON_TCP_OFO_QUEUE_PRUNE, /* pruned from TCP OFO queue */
+	SKB_DROP_REASON_TCP_OFO_DROP,	/* data already in receive queue */
 	SKB_DROP_REASON_IP_OUTNOROUTES,	/* route lookup failed */
 	SKB_DROP_REASON_BPF_CGROUP_EGRESS,	/* dropped by
 						 * BPF_PROG_TYPE_CGROUP_SKB
include/trace/events/skb.h +1 -0
index 73d7a6e594cbc6884148afa473bc08d12d783079..a477bf9074982cde8d1025ed18086fecafae8807 100644
@@ -37,6 +37,7 @@
 	EM(SKB_DROP_REASON_TCP_OLD_DATA, TCP_OLD_DATA)		\
 	EM(SKB_DROP_REASON_TCP_OVERWINDOW, TCP_OVERWINDOW)	\
 	EM(SKB_DROP_REASON_TCP_OFOMERGE, TCP_OFOMERGE)		\
+	EM(SKB_DROP_REASON_TCP_OFO_DROP, TCP_OFO_DROP)		\
 	EM(SKB_DROP_REASON_TCP_RFC7323_PAWS, TCP_RFC7323_PAWS)	\
 	EM(SKB_DROP_REASON_TCP_INVALID_SEQUENCE,		\
 	   TCP_INVALID_SEQUENCE)				\
net/ipv4/tcp_input.c +2 -7
index 339cc3d40745a0ea2a9f66b03dfda5aa6800d4a2..cf2dc19bb8c766c1d33406053fd61c0873f15489 100644
@@ -4674,7 +4674,7 @@ static bool tcp_ooo_try_coalesce(struct sock *sk,
 {
 	bool res = tcp_try_coalesce(sk, to, from, fragstolen);
 
-	/* In case tcp_drop() is called later, update to->gso_segs */
+	/* In case tcp_drop_reason() is called later, update to->gso_segs */
 	if (res) {
 		u32 gso_segs = max_t(u16, 1, skb_shinfo(to)->gso_segs) +
 			       max_t(u16, 1, skb_shinfo(from)->gso_segs);
@@ -4691,11 +4691,6 @@ static void tcp_drop_reason(struct sock *sk, struct sk_buff *skb,
 	kfree_skb_reason(skb, reason);
 }
 
-static void tcp_drop(struct sock *sk, struct sk_buff *skb)
-{
-	tcp_drop_reason(sk, skb, SKB_DROP_REASON_NOT_SPECIFIED);
-}
-
 /* This one checks to see if we can put data from the
  * out_of_order queue into the receive_queue.
  */
@@ -4723,7 +4718,7 @@ static void tcp_ofo_queue(struct sock *sk)
 		rb_erase(&skb->rbnode, &tp->out_of_order_queue);
 
 		if (unlikely(!after(TCP_SKB_CB(skb)->end_seq, tp->rcv_nxt))) {
-			tcp_drop(sk, skb);
+			tcp_drop_reason(sk, skb, SKB_DROP_REASON_TCP_OFO_DROP);
 			continue;
 		}
 
Signature · 1 line
2.36.0.rc0.470.gd361397f0d-goog
#11 lore raw
In reply to Eric Dumazet
This series was applied to netdev/net-next.git (master) by David S. Miller <[email protected]>:

Hello:

This series was applied to netdev/net-next.git (master) by David S. Miller <[email protected]>:

Quoted 9 lines · On Fri, 15 Apr 2022 17:10:38 -0700 you wrote:

From: Eric Dumazet <[email protected]>

Currently, TCP is either missing drop reasons, or pretending that some useful packets are dropped.

This patch series makes "perf record -a -e skb:kfree_skb" much more usable.

[...]

Here is the summary with links:
- [net-next,01/10] tcp: consume incoming skb leading to a reset

    https://git.kernel.org/netdev/net-next/c/d9d024f96609
  - [net-next,02/10] tcp: get rid of rst_seq_match
    https://git.kernel.org/netdev/net-next/c/b5ec1e6205a1
  - [net-next,03/10] tcp: add drop reason support to tcp_validate_incoming()
    https://git.kernel.org/netdev/net-next/c/da40b613f89c
  - [net-next,04/10] tcp: make tcp_rcv_state_process() drop monitor friendly
    https://git.kernel.org/netdev/net-next/c/37fd4e842391
  - [net-next,05/10] tcp: add drop reasons to tcp_rcv_state_process()
    https://git.kernel.org/netdev/net-next/c/669da7a71890
  - [net-next,06/10] tcp: add two drop reasons for tcp_ack()
    https://git.kernel.org/netdev/net-next/c/4b506af9c5b8
  - [net-next,07/10] tcp: add drop reason support to tcp_prune_ofo_queue()
    https://git.kernel.org/netdev/net-next/c/e7c89ae4078e
  - [net-next,08/10] tcp: make tcp_rcv_synsent_state_process() drop monitor friend
    https://git.kernel.org/netdev/net-next/c/c337578a6592
  - [net-next,09/10] tcp: add drop reasons to tcp_rcv_synsent_state_process()
    https://git.kernel.org/netdev/net-next/c/659affdb5140
  - [net-next,10/10] tcp: add drop reason support to tcp_ofo_queue()
    https://git.kernel.org/netdev/net-next/c/8fbf195798b5

You are awesome, thank you!

Signature · 2 lines
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html