Skip to contentSkip to conversation

tcp: completely support active reset

PATCH Superseded Jason Xing 2 participants 14 messages Aug 1, 2024 – Aug 2, 2024

Subsystems (3)

v3 of 4 Superseded Latest is v4 →

lore
Share & tools
lore

Jason Xing

author
#0 lore raw

v3 cover tcp: completely support active reset

From: Jason Xing <[email protected]>

This time the patch series finally covers all the cases in the active reset logic. After this, we can know the related exact reason(s).

v3

1. introduce TCP_DISCONNECT_WITH_DATA reason (Eric)
2. use a better name 'TCP_KEEPALIVE_TIMEOUT' (Eric)
3. add three reviewed-by tags (Eric)

v2

1. use RFC 9293 in the comment and changelog instead of old RFC 793
2. correct the comment and changelog in patch 5

Jason Xing (7):
tcp: rstreason: introduce SK_RST_REASON_TCP_ABORT_ON_CLOSE for active
reset
tcp: rstreason: introduce SK_RST_REASON_TCP_ABORT_ON_LINGER for active
reset
tcp: rstreason: introduce SK_RST_REASON_TCP_ABORT_ON_MEMORY for active
reset
tcp: rstreason: introduce SK_RST_REASON_TCP_STATE for active reset
tcp: rstreason: introduce SK_RST_REASON_TCP_KEEPALIVE_TIMEOUT for
active reset
tcp: rstreason: introduce SK_RST_REASON_TCP_DISCONNECT_WITH_DATA for
active reset
tcp: rstreason: let it work finally in tcp_send_active_reset()

 include/net/rstreason.h | 39 +++++++++++++++++++++++++++++++++++++++
 net/ipv4/tcp.c          | 19 +++++++++++--------
 net/ipv4/tcp_output.c   |  2 +-
 net/ipv4/tcp_timer.c    |  6 +++---
 4 files changed, 54 insertions(+), 12 deletions(-)
Signature · 1 line
2.37.3
author
#1 lore raw

PATCH v3 1/7 tcp: rstreason: introduce SK_RST_REASON_TCP_ABORT_ON_CLOSE for active reset

From: Jason Xing <[email protected]>

Introducing a new type TCP_ABORT_ON_CLOSE for tcp reset reason to handle the case where more data is unread in closing phase.

---
 include/net/rstreason.h | 6 ++++++
 net/ipv4/tcp.c          | 2 +-
 2 files changed, 7 insertions(+), 1 deletion(-)
include/net/rstreason.h +6 -0
index 2575c85d7f7a..fa6bfd0d7d69 100644
@@ -17,6 +17,7 @@
 	FN(TCP_ABORT_ON_DATA)		\
 	FN(TCP_TIMEWAIT_SOCKET)		\
 	FN(INVALID_SYN)			\
+	FN(TCP_ABORT_ON_CLOSE)		\
 	FN(MPTCP_RST_EUNSPEC)		\
 	FN(MPTCP_RST_EMPTCP)		\
 	FN(MPTCP_RST_ERESOURCE)		\
@@ -84,6 +85,11 @@ enum sk_rst_reason {
 	 * an error, send a reset"
 	 */
 	SK_RST_REASON_INVALID_SYN,
+	/**
+	 * @SK_RST_REASON_TCP_ABORT_ON_CLOSE: abort on close
+	 * corresponding to LINUX_MIB_TCPABORTONCLOSE
+	 */
+	SK_RST_REASON_TCP_ABORT_ON_CLOSE,
 
 	/* Copy from include/uapi/linux/mptcp.h.
 	 * These reset fields will not be changed since they adhere to
net/ipv4/tcp.c +1 -1
index e03a342c9162..2e010add0317 100644
@@ -2833,7 +2833,7 @@ void __tcp_close(struct sock *sk, long timeout)
 		NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPABORTONCLOSE);
 		tcp_set_state(sk, TCP_CLOSE);
 		tcp_send_active_reset(sk, sk->sk_allocation,
-				      SK_RST_REASON_NOT_SPECIFIED);
+				      SK_RST_REASON_TCP_ABORT_ON_CLOSE);
 	} else if (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime) {
 		/* Check zero linger _after_ checking for unread data. */
 		sk->sk_prot->disconnect(sk, 0);
Signature · 1 line
2.37.3
author
#2 lore raw

PATCH v3 2/7 tcp: rstreason: introduce SK_RST_REASON_TCP_ABORT_ON_LINGER for active reset

In reply to Jason Xing

From: Jason Xing <[email protected]>

Introducing a new type TCP_ABORT_ON_LINGER for tcp reset reason to handle negative linger value case.

---
 include/net/rstreason.h | 6 ++++++
 net/ipv4/tcp.c          | 2 +-
 2 files changed, 7 insertions(+), 1 deletion(-)
include/net/rstreason.h +6 -0
index fa6bfd0d7d69..fbbaeb969e6a 100644
@@ -18,6 +18,7 @@
 	FN(TCP_TIMEWAIT_SOCKET)		\
 	FN(INVALID_SYN)			\
 	FN(TCP_ABORT_ON_CLOSE)		\
+	FN(TCP_ABORT_ON_LINGER)		\
 	FN(MPTCP_RST_EUNSPEC)		\
 	FN(MPTCP_RST_EMPTCP)		\
 	FN(MPTCP_RST_ERESOURCE)		\
@@ -90,6 +91,11 @@ enum sk_rst_reason {
 	 * corresponding to LINUX_MIB_TCPABORTONCLOSE
 	 */
 	SK_RST_REASON_TCP_ABORT_ON_CLOSE,
+	/**
+	 * @SK_RST_REASON_TCP_ABORT_ON_LINGER: abort on linger
+	 * corresponding to LINUX_MIB_TCPABORTONLINGER
+	 */
+	SK_RST_REASON_TCP_ABORT_ON_LINGER,
 
 	/* Copy from include/uapi/linux/mptcp.h.
 	 * These reset fields will not be changed since they adhere to
net/ipv4/tcp.c +1 -1
index 2e010add0317..5b0f1d1fc697 100644
@@ -2908,7 +2908,7 @@ void __tcp_close(struct sock *sk, long timeout)
 		if (READ_ONCE(tp->linger2) < 0) {
 			tcp_set_state(sk, TCP_CLOSE);
 			tcp_send_active_reset(sk, GFP_ATOMIC,
-					      SK_RST_REASON_NOT_SPECIFIED);
+					      SK_RST_REASON_TCP_ABORT_ON_LINGER);
 			__NET_INC_STATS(sock_net(sk),
 					LINUX_MIB_TCPABORTONLINGER);
 		} else {
Signature · 1 line
2.37.3
author
#3 lore raw

PATCH v3 3/7 tcp: rstreason: introduce SK_RST_REASON_TCP_ABORT_ON_MEMORY for active reset

In reply to Jason Xing

From: Jason Xing <[email protected]>

Introducing a new type TCP_ABORT_ON_MEMORY for tcp reset reason to handle out of memory case.

---
 include/net/rstreason.h | 6 ++++++
 net/ipv4/tcp.c          | 2 +-
 net/ipv4/tcp_timer.c    | 2 +-
 3 files changed, 8 insertions(+), 2 deletions(-)
include/net/rstreason.h +6 -0
index fbbaeb969e6a..eef658da8952 100644
@@ -19,6 +19,7 @@
 	FN(INVALID_SYN)			\
 	FN(TCP_ABORT_ON_CLOSE)		\
 	FN(TCP_ABORT_ON_LINGER)		\
+	FN(TCP_ABORT_ON_MEMORY)		\
 	FN(MPTCP_RST_EUNSPEC)		\
 	FN(MPTCP_RST_EMPTCP)		\
 	FN(MPTCP_RST_ERESOURCE)		\
@@ -96,6 +97,11 @@ enum sk_rst_reason {
 	 * corresponding to LINUX_MIB_TCPABORTONLINGER
 	 */
 	SK_RST_REASON_TCP_ABORT_ON_LINGER,
+	/**
+	 * @SK_RST_REASON_TCP_ABORT_ON_MEMORY: abort on memory
+	 * corresponding to LINUX_MIB_TCPABORTONMEMORY
+	 */
+	SK_RST_REASON_TCP_ABORT_ON_MEMORY,
 
 	/* Copy from include/uapi/linux/mptcp.h.
 	 * These reset fields will not be changed since they adhere to
net/ipv4/tcp.c +1 -1
index 5b0f1d1fc697..fd928c447ce8 100644
@@ -2927,7 +2927,7 @@ void __tcp_close(struct sock *sk, long timeout)
 		if (tcp_check_oom(sk, 0)) {
 			tcp_set_state(sk, TCP_CLOSE);
 			tcp_send_active_reset(sk, GFP_ATOMIC,
-					      SK_RST_REASON_NOT_SPECIFIED);
+					      SK_RST_REASON_TCP_ABORT_ON_MEMORY);
 			__NET_INC_STATS(sock_net(sk),
 					LINUX_MIB_TCPABORTONMEMORY);
 		} else if (!check_net(sock_net(sk))) {
net/ipv4/tcp_timer.c +1 -1
index 4d40615dc8fc..0fba4a4fb988 100644
@@ -125,7 +125,7 @@ static int tcp_out_of_resources(struct sock *sk, bool do_reset)
 			do_reset = true;
 		if (do_reset)
 			tcp_send_active_reset(sk, GFP_ATOMIC,
-					      SK_RST_REASON_NOT_SPECIFIED);
+					      SK_RST_REASON_TCP_ABORT_ON_MEMORY);
 		tcp_done(sk);
 		__NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPABORTONMEMORY);
 		return 1;
Signature · 1 line
2.37.3
author
#4 lore raw

PATCH v3 4/7 tcp: rstreason: introduce SK_RST_REASON_TCP_STATE for active reset

In reply to Jason Xing

From: Jason Xing <[email protected]>

Introducing a new type TCP_STATE to handle some reset conditions appearing in RFC 793 due to its socket state. Actually, we can look into RFC 9293 which has no discrepancy about this part.

---
v3

1. remove one case from tcp_disconnect, which will be separately categorized as another reason in the later patch (Eric)

V2

1. use RFC 9293 instead of RFC 793 which is too old (Kuniyuki)

---
 include/net/rstreason.h |  6 ++++++
 net/ipv4/tcp.c          | 10 ++++++----
 net/ipv4/tcp_timer.c    |  2 +-
 3 files changed, 13 insertions(+), 5 deletions(-)
include/net/rstreason.h +6 -0
index eef658da8952..bbf20d0bbde7 100644
@@ -20,6 +20,7 @@
 	FN(TCP_ABORT_ON_CLOSE)		\
 	FN(TCP_ABORT_ON_LINGER)		\
 	FN(TCP_ABORT_ON_MEMORY)		\
+	FN(TCP_STATE)			\
 	FN(MPTCP_RST_EUNSPEC)		\
 	FN(MPTCP_RST_EMPTCP)		\
 	FN(MPTCP_RST_ERESOURCE)		\
@@ -102,6 +103,11 @@ enum sk_rst_reason {
 	 * corresponding to LINUX_MIB_TCPABORTONMEMORY
 	 */
 	SK_RST_REASON_TCP_ABORT_ON_MEMORY,
+	/**
+	 * @SK_RST_REASON_TCP_STATE: abort on tcp state
+	 * Please see RFC 9293 for all possible reset conditions
+	 */
+	SK_RST_REASON_TCP_STATE,
 
 	/* Copy from include/uapi/linux/mptcp.h.
 	 * These reset fields will not be changed since they adhere to
net/ipv4/tcp.c +6 -4
index fd928c447ce8..24777e48bcc8 100644
@@ -3025,9 +3025,11 @@ int tcp_disconnect(struct sock *sk, int flags)
 		inet_csk_listen_stop(sk);
 	} else if (unlikely(tp->repair)) {
 		WRITE_ONCE(sk->sk_err, ECONNABORTED);
-	} else if (tcp_need_reset(old_state) ||
-		   (tp->snd_nxt != tp->write_seq &&
-		    (1 << old_state) & (TCPF_CLOSING | TCPF_LAST_ACK))) {
+	} else if (tcp_need_reset(old_state)) {
+		tcp_send_active_reset(sk, gfp_any(), SK_RST_REASON_TCP_STATE);
+		WRITE_ONCE(sk->sk_err, ECONNRESET);
+	} else if (tp->snd_nxt != tp->write_seq &&
+		   (1 << old_state) & (TCPF_CLOSING | TCPF_LAST_ACK)) {
 		/* The last check adjusts for discrepancy of Linux wrt. RFC
 		 * states
 		 */
@@ -4649,7 +4651,7 @@ int tcp_abort(struct sock *sk, int err)
 	if (!sock_flag(sk, SOCK_DEAD)) {
 		if (tcp_need_reset(sk->sk_state))
 			tcp_send_active_reset(sk, GFP_ATOMIC,
-					      SK_RST_REASON_NOT_SPECIFIED);
+					      SK_RST_REASON_TCP_STATE);
 		tcp_done_with_error(sk, err);
 	}
 
net/ipv4/tcp_timer.c +1 -1
index 0fba4a4fb988..3910f6d8614e 100644
@@ -779,7 +779,7 @@ static void tcp_keepalive_timer (struct timer_list *t)
 				goto out;
 			}
 		}
-		tcp_send_active_reset(sk, GFP_ATOMIC, SK_RST_REASON_NOT_SPECIFIED);
+		tcp_send_active_reset(sk, GFP_ATOMIC, SK_RST_REASON_TCP_STATE);
 		goto death;
 	}
 
Signature · 1 line
2.37.3
author
#7 lore raw

PATCH v3 5/7 tcp: rstreason: introduce SK_RST_REASON_TCP_KEEPALIVE_TIMEOUT for active reset

In reply to Jason Xing

From: Jason Xing <[email protected]>

When we find keepalive timeout here, we should send an RST to the other side.

---
v3

1. chose a better reason name (Eric)

v2

1. correct the comment and changelog

---
 include/net/rstreason.h | 7 +++++++
 net/ipv4/tcp_timer.c    | 2 +-
 2 files changed, 8 insertions(+), 1 deletion(-)
include/net/rstreason.h +7 -0
index bbf20d0bbde7..9c0c46df0e73 100644
@@ -21,6 +21,7 @@
 	FN(TCP_ABORT_ON_LINGER)		\
 	FN(TCP_ABORT_ON_MEMORY)		\
 	FN(TCP_STATE)			\
+	FN(TCP_KEEPALIVE_TIMEOUT)	\
 	FN(MPTCP_RST_EUNSPEC)		\
 	FN(MPTCP_RST_EMPTCP)		\
 	FN(MPTCP_RST_ERESOURCE)		\
@@ -108,6 +109,12 @@ enum sk_rst_reason {
 	 * Please see RFC 9293 for all possible reset conditions
 	 */
 	SK_RST_REASON_TCP_STATE,
+	/**
+	 * @SK_RST_REASON_TCP_KEEPALIVE_TIMEOUT: time to timeout
+	 * When we have already run out of all the chances, which means
+	 * keepalive timeout, we have to reset the connection
+	 */
+	SK_RST_REASON_TCP_KEEPALIVE_TIMEOUT,
 
 	/* Copy from include/uapi/linux/mptcp.h.
 	 * These reset fields will not be changed since they adhere to
net/ipv4/tcp_timer.c +1 -1
index 3910f6d8614e..86169127e4d1 100644
@@ -807,7 +807,7 @@ static void tcp_keepalive_timer (struct timer_list *t)
 		    (user_timeout == 0 &&
 		    icsk->icsk_probes_out >= keepalive_probes(tp))) {
 			tcp_send_active_reset(sk, GFP_ATOMIC,
-					      SK_RST_REASON_NOT_SPECIFIED);
+					      SK_RST_REASON_TCP_KEEPALIVE_TIMEOUT);
 			tcp_write_err(sk);
 			goto out;
 		}
Signature · 1 line
2.37.3
author
#10 lore raw

PATCH v3 6/7 tcp: rstreason: introduce SK_RST_REASON_TCP_DISCONNECT_WITH_DATA for active reset

In reply to Jason Xing

From: Jason Xing <[email protected]>

When user tries to disconnect a socket and there are more data written into tcp write queue, we have to send an RST.

--
v3

1. This case is different from previous patch, so we need to write it into a new patch. (Eric)

---
 include/net/rstreason.h | 8 ++++++++
 net/ipv4/tcp.c          | 3 ++-
 2 files changed, 10 insertions(+), 1 deletion(-)
include/net/rstreason.h +8 -0
index 9c0c46df0e73..69cb2e52b7da 100644
@@ -22,6 +22,7 @@
 	FN(TCP_ABORT_ON_MEMORY)		\
 	FN(TCP_STATE)			\
 	FN(TCP_KEEPALIVE_TIMEOUT)	\
+	FN(TCP_DISCONNECT_WITH_DATA)	\
 	FN(MPTCP_RST_EUNSPEC)		\
 	FN(MPTCP_RST_EMPTCP)		\
 	FN(MPTCP_RST_ERESOURCE)		\
@@ -115,6 +116,13 @@ enum sk_rst_reason {
 	 * keepalive timeout, we have to reset the connection
 	 */
 	SK_RST_REASON_TCP_KEEPALIVE_TIMEOUT,
+	/**
+	 * @SK_RST_REASON_TCP_DISCONNECT_WITH_DATA: disconnect when write
+	 * queue is not empty
+	 * It means user has written data into the write queue when doing
+	 * disconnecting, so we have to send an RST.
+	 */
+	SK_RST_REASON_TCP_DISCONNECT_WITH_DATA,
 
 	/* Copy from include/uapi/linux/mptcp.h.
 	 * These reset fields will not be changed since they adhere to
net/ipv4/tcp.c +2 -1
index 24777e48bcc8..8514257f4ecd 100644
@@ -3033,7 +3033,8 @@ int tcp_disconnect(struct sock *sk, int flags)
 		/* The last check adjusts for discrepancy of Linux wrt. RFC
 		 * states
 		 */
-		tcp_send_active_reset(sk, gfp_any(), SK_RST_REASON_NOT_SPECIFIED);
+		tcp_send_active_reset(sk, gfp_any(),
+				      SK_RST_REASON_TCP_DISCONNECT_WITH_DATA);
 		WRITE_ONCE(sk->sk_err, ECONNRESET);
 	} else if (old_state == TCP_SYN_SENT)
 		WRITE_ONCE(sk->sk_err, ECONNRESET);
Signature · 1 line
2.37.3
author
#12 lore raw

PATCH v3 7/7 tcp: rstreason: let it work finally in tcp_send_active_reset()

In reply to Jason Xing

From: Jason Xing <[email protected]>

Now it's time to let it work by using the 'reason' parameter in the trace world :)

---
 net/ipv4/tcp_output.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
net/ipv4/tcp_output.c +1 -1
index 16c48df8df4c..cdd0def14427 100644
@@ -3649,7 +3649,7 @@ void tcp_send_active_reset(struct sock *sk, gfp_t priority,
 	/* skb of trace_tcp_send_reset() keeps the skb that caused RST,
 	 * skb here is different to the troublesome skb, so use NULL
 	 */
-	trace_tcp_send_reset(sk, NULL, SK_RST_REASON_NOT_SPECIFIED);
+	trace_tcp_send_reset(sk, NULL, reason);
 }
 
 /* Send a crossed SYN-ACK during socket establishment.
Signature · 1 line
2.37.3
Reviewed-by on v3 7/7
#13 lore raw

Re: [PATCH net-next v3 7/7] tcp: rstreason: let it work finally in tcp_send_active_reset()

In reply to Jason Xing
Quoted 16 lines · Jason Xing

From: Jason Xing <[email protected]>

Now it's time to let it work by using the 'reason' parameter in the trace world :)

---
 net/ipv4/tcp_output.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
net/ipv4/tcp_output.c +0 -0
index 16c48df8df4c..cdd0def14427 100644

net/ipv4/tcp_output.c void tcp_send_active_reset(struct sock *sk, gfp_t priority, · line 3652

Quoted 5 lines
net/ipv4/tcp_output.c +1 -1
@@ -3649,7 +3649,7 @@ void tcp_send_active_reset(struct sock *sk, gfp_t priority,
        /* skb of trace_tcp_send_reset() keeps the skb that caused RST,
         * skb here is different to the troublesome skb, so use NULL
         */
-       trace_tcp_send_reset(sk, NULL, SK_RST_REASON_NOT_SPECIFIED);
+       trace_tcp_send_reset(sk, NULL, reason);
Reviewed-by on v3 5/7
#8 lore raw

Re: [PATCH net-next v3 5/7] tcp: rstreason: introduce SK_RST_REASON_TCP_KEEPALIVE_TIMEOUT for active reset

In reply to Jason Xing
Quoted 7 lines · Jason Xing

From: Jason Xing <[email protected]>

When we find keepalive timeout here, we should send an RST to the other side.

Note that this changelog does not really match the code.

We were sending an RST already.

Precise changelogs are needed to avoid extra work by stable teams, that can 'catch' things based on some keywords, not only Fixes: tags.

Reviewed-by on v3 4/7
#5 lore raw

Re: [PATCH net-next v3 4/7] tcp: rstreason: introduce SK_RST_REASON_TCP_STATE for active reset

In reply to Jason Xing
Quoted 8 lines · Jason Xing

From: Jason Xing <[email protected]>

Introducing a new type TCP_STATE to handle some reset conditions appearing in RFC 793 due to its socket state. Actually, we can look into RFC 9293 which has no discrepancy about this part.

I really think this SK_RST_REASON_TCP_STATE is weak.

'Please see RFC 9293' does not help, this RFC has more than 5000 lines in it :/

author
#9 lore raw

Re: [PATCH net-next v3 5/7] tcp: rstreason: introduce SK_RST_REASON_TCP_KEEPALIVE_TIMEOUT for active reset

In reply to Eric Dumazet

Hello Eric,

Quoted 8 lines · Eric Dumazet
Quoted 6 lines · Jason Xing

From: Jason Xing <[email protected]>

When we find keepalive timeout here, we should send an RST to the other side.

Quoted 8 lines
Quoted 1 line

Note that this changelog does not really match the code.

We were sending an RST already.

Precise changelogs are needed to avoid extra work by stable teams, that can 'catch' things based on some keywords, not only Fixes: tags.

Thanks for reminding me. I will revise the changelog.

Quoted 2 lines

Thanks!

author
#6 lore raw

Re: [PATCH net-next v3 4/7] tcp: rstreason: introduce SK_RST_REASON_TCP_STATE for active reset

In reply to Eric Dumazet

Hello Eric,

Quoted 9 lines · Eric Dumazet
Quoted 7 lines · Jason Xing

From: Jason Xing <[email protected]>

Introducing a new type TCP_STATE to handle some reset conditions appearing in RFC 793 due to its socket state. Actually, we can look into RFC 9293 which has no discrepancy about this part.

Quoted 5 lines
Quoted 1 line

I really think this SK_RST_REASON_TCP_STATE is weak.

'Please see RFC 9293' does not help, this RFC has more than 5000 lines in it :/

Yes, there are various reasons and conditions written in RFC 9293 nearly all over the place. I'm unable to conclude and get an union name. Sorry about that. If I figure out a better name in the future, I'll let you know.

Quoted 2 lines

Thanks for all your help.

Thanks,
Jason