[PATCH net-next v3 0/6] Implement reset reason mechanism to detect

[PATCH net-next v3 0/6] Implement reset reason mechanism to detect

[Jason Xing]

From: Jason Xing <kernelxing@tencent.com>

In production, there are so many cases about why the RST skb is sent but
we don't have a very convenient/fast method to detect the exact underlying
reasons.

RST is implemented in two kinds: passive kind (like tcp_v4_send_reset())
and active kind (like tcp_send_active_reset()). The former can be traced
carefully 1) in TCP, with the help of drop reasons, which is based on
Eric's idea[1], 2) in MPTCP, with the help of reset options defined in
RFC 8684. The latter is relatively independent, which should be
implemented on our own.

In this series, I focus on the fundamental implement mostly about how
the rstreason mechnism works and give the detailed passive part as an
example, not including the active reset part. In future, we can go
further and refine those NOT_SPECIFIED reasons.

Here are some examples when tracing:
<idle>-0       [002] ..s1.  1830.262425: tcp_send_reset: skbaddr=x
        skaddr=x src=x dest=x state=x reason=NOT_SPECIFIED
<idle>-0       [002] ..s1.  1830.262425: tcp_send_reset: skbaddr=x
        skaddr=x src=x dest=x state=x reason=NO_SOCKET

[1]
Link: https://lore.kernel.org/all/CANn89iJw8x-LqgsWOeJQQvgVg6DnL5aBRLi10QN2WBdr+X4k=w@mail.gmail.com/

v3
Link:
https://lore.kernel.org/all/20240404072047.11490-1-kerneljasonxing@gmail.com/
1. rebase (mptcp part) and address what Mat suggested.

v2
Link: https://lore.kernel.org/all/20240403185033.47ebc6a9@kernel.org/
1. rebase against the latest net-next tree

Jason Xing (6):
  net: introduce rstreason to detect why the RST is sent
  rstreason: prepare for passive reset
  rstreason: prepare for active reset
  tcp: support rstreason for passive reset
  mptcp: support rstreason for passive reset
  rstreason: make it work in trace world

 include/net/request_sock.h |  3 +-
 include/net/rstreason.h    | 93 ++++++++++++++++++++++++++++++++++++++
 include/net/tcp.h          |  2 +-
 include/trace/events/tcp.h | 37 +++++++++++++--
 net/dccp/ipv4.c            | 10 ++--
 net/dccp/ipv6.c            | 10 ++--
 net/dccp/minisocks.c       |  3 +-
 net/ipv4/tcp.c             | 15 ++++--
 net/ipv4/tcp_ipv4.c        | 14 +++---
 net/ipv4/tcp_minisocks.c   |  3 +-
 net/ipv4/tcp_output.c      |  4 +-
 net/ipv4/tcp_timer.c       |  9 ++--
 net/ipv6/tcp_ipv6.c        | 17 ++++---
 net/mptcp/protocol.c       |  4 +-
 net/mptcp/subflow.c        | 25 +++++++---
 15 files changed, 202 insertions(+), 47 deletions(-)
 create mode 100644 include/net/rstreason.h

-- 
2.37.3

[PATCH net-next v3 1/6] net: introduce rstreason to detect why the RST is sent

[Jason Xing]

From: Jason Xing <kernelxing@tencent.com>

Add a new standalone file for the easy future extension to support
both active reset and passive reset in the TCP/DCCP/MPTCP protocols.

This patch only does the preparations for reset reason mechanism,
nothing else changes.

The reset reasons are divided into three parts:
1) reuse drop reasons for passive reset in TCP
2) reuse MP_TCPRST option for MPTCP
3) our own reasons

I will implement the basic codes of active/passive reset reason in
those three protocols, which is not complete for this moment. But
it provides a new chance to let other people add more reasons into
it:)

Signed-off-by: Jason Xing <kernelxing@tencent.com>
---
 include/net/rstreason.h | 93 +++++++++++++++++++++++++++++++++++++++++
 1 file changed, 93 insertions(+)
 create mode 100644 include/net/rstreason.h

diff --git a/include/net/rstreason.h b/include/net/rstreason.h
new file mode 100644
index 000000000000..24d098a78a60
--- /dev/null
+++ b/include/net/rstreason.h
@@ -0,0 +1,93 @@
+/* SPDX-License-Identifier: GPL-2.0-or-later */
+
+#ifndef _LINUX_RSTREASON_H
+#define _LINUX_RSTREASON_H
+#include <net/dropreason-core.h>
+
+#define DEFINE_RST_REASON(FN, FNe)	\
+	FN(MPTCP_RST_EUNSPEC)		\
+	FN(MPTCP_RST_EMPTCP)		\
+	FN(MPTCP_RST_ERESOURCE)		\
+	FN(MPTCP_RST_EPROHIBIT)		\
+	FN(MPTCP_RST_EWQ2BIG)		\
+	FN(MPTCP_RST_EBADPERF)		\
+	FN(MPTCP_RST_EMIDDLEBOX)	\
+	FN(NOT_SPECIFIED)		\
+	FNe(MAX)
+
+#define RST_REASON_START (SKB_DROP_REASON_MAX + 1)
+
+/* There are three parts in order:
+ * 1) 0 - SKB_DROP_REASON_MAX: rely on drop reasons for passive reset in TCP
+ * 2) SKB_DROP_REASON_MAX + 1 - MPTCP_RST_EMIDDLEBOX: for MPTCP use
+ * 3) MPTCP_RST_EMIDDLEBOX - SK_RST_REASON_MAX: independent reset reason
+ */
+enum sk_rst_reason {
+	/* Leave this 'blank' part (0-SKB_DROP_REASON_MAX) for the reuse
+	 * of skb drop reason because rst reason relies on what drop reason
+	 * indicates exactly why it could happen.
+	 */
+
+	/* Copy from include/uapi/linux/mptcp.h.
+	 * These reset fields will not be changed since they adhere to
+	 * RFC 8684. So do not touch them. I'm going to list each definition
+	 * of them respectively.
+	 */
+	/* Unspecified error.
+	 * This is the default error; it implies that the subflow is no
+	 * longer available. The presence of this option shows that the
+	 * RST was generated by an MPTCP-aware device.
+	 */
+	SK_RST_REASON_MPTCP_RST_EUNSPEC = RST_REASON_START,
+	/* MPTCP-specific error.
+	 * An error has been detected in the processing of MPTCP options.
+	 * This is the usual reason code to return in the cases where a RST
+	 * is being sent to close a subflow because of an invalid response.
+	 */
+	SK_RST_REASON_MPTCP_RST_EMPTCP,
+	/* Lack of resources.
+	 * This code indicates that the sending host does not have enough
+	 * resources to support the terminated subflow.
+	 */
+	SK_RST_REASON_MPTCP_RST_ERESOURCE,
+	/* Administratively prohibited.
+	 * This code indicates that the requested subflow is prohibited by
+	 * the policies of the sending host.
+	 */
+	SK_RST_REASON_MPTCP_RST_EPROHIBIT,
+	/* Too much outstanding data.
+	 * This code indicates that there is an excessive amount of data
+	 * that needs to be transmitted over the terminated subflow while
+	 * having already been acknowledged over one or more other subflows.
+	 * This may occur if a path has been unavailable for a short period
+	 * and it is more efficient to reset and start again than it is to
+	 * retransmit the queued data.
+	 */
+	SK_RST_REASON_MPTCP_RST_EWQ2BIG,
+	/* Unacceptable performance.
+	 * This code indicates that the performance of this subflow was
+	 * too low compared to the other subflows of this Multipath TCP
+	 * connection.
+	 */
+	SK_RST_REASON_MPTCP_RST_EBADPERF,
+	/* Middlebox interference.
+	 * Middlebox interference has been detected over this subflow,
+	 * making MPTCP signaling invalid. For example, this may be sent
+	 * if the checksum does not validate.
+	 */
+	SK_RST_REASON_MPTCP_RST_EMIDDLEBOX,
+
+	/* For the real standalone socket reset reason, we start from here */
+	SK_RST_REASON_NOT_SPECIFIED,
+
+	/* Maximum of socket reset reasons.
+	 * It shouldn't be used as a real 'reason'.
+	 */
+	SK_RST_REASON_MAX,
+};
+
+static inline int convert_mptcp_reason(int reason)
+{
+	return reason += RST_REASON_START;
+}
+#endif
-- 
2.37.3

[PATCH net-next v3 2/6] rstreason: prepare for passive reset

[Jason Xing]

From: Jason Xing <kernelxing@tencent.com>

Adjust the parameter and support passing reason of reset which
is for now NOT_SPECIFIED. No functional changes.

Signed-off-by: Jason Xing <kernelxing@tencent.com>
---
 include/net/request_sock.h |  3 ++-
 net/dccp/ipv4.c            | 10 ++++++----
 net/dccp/ipv6.c            | 10 ++++++----
 net/dccp/minisocks.c       |  3 ++-
 net/ipv4/tcp_ipv4.c        | 12 +++++++-----
 net/ipv4/tcp_minisocks.c   |  3 ++-
 net/ipv6/tcp_ipv6.c        | 15 +++++++++------
 net/mptcp/subflow.c        |  8 +++++---
 8 files changed, 39 insertions(+), 25 deletions(-)

diff --git a/include/net/request_sock.h b/include/net/request_sock.h
index 004e651e6067..93f9fee7e52f 100644
--- a/include/net/request_sock.h
+++ b/include/net/request_sock.h
@@ -34,7 +34,8 @@ struct request_sock_ops {
 	void		(*send_ack)(const struct sock *sk, struct sk_buff *skb,
 				    struct request_sock *req);
 	void		(*send_reset)(const struct sock *sk,
-				      struct sk_buff *skb);
+				      struct sk_buff *skb,
+				      int reason);
 	void		(*destructor)(struct request_sock *req);
 	void		(*syn_ack_timeout)(const struct request_sock *req);
 };
diff --git a/net/dccp/ipv4.c b/net/dccp/ipv4.c
index 9fc9cea4c251..11b8d14be3e2 100644
--- a/net/dccp/ipv4.c
+++ b/net/dccp/ipv4.c
@@ -24,6 +24,7 @@
 #include <net/xfrm.h>
 #include <net/secure_seq.h>
 #include <net/netns/generic.h>
+#include <net/rstreason.h>
 
 #include "ackvec.h"
 #include "ccid.h"
@@ -521,7 +522,8 @@ static int dccp_v4_send_response(const struct sock *sk, struct request_sock *req
 	return err;
 }
 
-static void dccp_v4_ctl_send_reset(const struct sock *sk, struct sk_buff *rxskb)
+static void dccp_v4_ctl_send_reset(const struct sock *sk, struct sk_buff *rxskb,
+				   int reason)
 {
 	int err;
 	const struct iphdr *rxiph;
@@ -706,7 +708,7 @@ int dccp_v4_do_rcv(struct sock *sk, struct sk_buff *skb)
 	return 0;
 
 reset:
-	dccp_v4_ctl_send_reset(sk, skb);
+	dccp_v4_ctl_send_reset(sk, skb, SK_RST_REASON_NOT_SPECIFIED);
 	kfree_skb(skb);
 	return 0;
 }
@@ -869,7 +871,7 @@ static int dccp_v4_rcv(struct sk_buff *skb)
 		if (nsk == sk) {
 			reqsk_put(req);
 		} else if (dccp_child_process(sk, nsk, skb)) {
-			dccp_v4_ctl_send_reset(sk, skb);
+			dccp_v4_ctl_send_reset(sk, skb, SK_RST_REASON_NOT_SPECIFIED);
 			goto discard_and_relse;
 		} else {
 			sock_put(sk);
@@ -909,7 +911,7 @@ static int dccp_v4_rcv(struct sk_buff *skb)
 	if (dh->dccph_type != DCCP_PKT_RESET) {
 		DCCP_SKB_CB(skb)->dccpd_reset_code =
 					DCCP_RESET_CODE_NO_CONNECTION;
-		dccp_v4_ctl_send_reset(sk, skb);
+		dccp_v4_ctl_send_reset(sk, skb, SK_RST_REASON_NOT_SPECIFIED);
 	}
 
 discard_it:
diff --git a/net/dccp/ipv6.c b/net/dccp/ipv6.c
index c8ca703dc331..232092dc3887 100644
--- a/net/dccp/ipv6.c
+++ b/net/dccp/ipv6.c
@@ -29,6 +29,7 @@
 #include <net/secure_seq.h>
 #include <net/netns/generic.h>
 #include <net/sock.h>
+#include <net/rstreason.h>
 
 #include "dccp.h"
 #include "ipv6.h"
@@ -256,7 +257,8 @@ static void dccp_v6_reqsk_destructor(struct request_sock *req)
 	kfree_skb(inet_rsk(req)->pktopts);
 }
 
-static void dccp_v6_ctl_send_reset(const struct sock *sk, struct sk_buff *rxskb)
+static void dccp_v6_ctl_send_reset(const struct sock *sk, struct sk_buff *rxskb,
+				   int reason)
 {
 	const struct ipv6hdr *rxip6h;
 	struct sk_buff *skb;
@@ -656,7 +658,7 @@ static int dccp_v6_do_rcv(struct sock *sk, struct sk_buff *skb)
 	return 0;
 
 reset:
-	dccp_v6_ctl_send_reset(sk, skb);
+	dccp_v6_ctl_send_reset(sk, skb, SK_RST_REASON_NOT_SPECIFIED);
 discard:
 	if (opt_skb != NULL)
 		__kfree_skb(opt_skb);
@@ -762,7 +764,7 @@ static int dccp_v6_rcv(struct sk_buff *skb)
 		if (nsk == sk) {
 			reqsk_put(req);
 		} else if (dccp_child_process(sk, nsk, skb)) {
-			dccp_v6_ctl_send_reset(sk, skb);
+			dccp_v6_ctl_send_reset(sk, skb, SK_RST_REASON_NOT_SPECIFIED);
 			goto discard_and_relse;
 		} else {
 			sock_put(sk);
@@ -801,7 +803,7 @@ static int dccp_v6_rcv(struct sk_buff *skb)
 	if (dh->dccph_type != DCCP_PKT_RESET) {
 		DCCP_SKB_CB(skb)->dccpd_reset_code =
 					DCCP_RESET_CODE_NO_CONNECTION;
-		dccp_v6_ctl_send_reset(sk, skb);
+		dccp_v6_ctl_send_reset(sk, skb, SK_RST_REASON_NOT_SPECIFIED);
 	}
 
 discard_it:
diff --git a/net/dccp/minisocks.c b/net/dccp/minisocks.c
index 64d805b27add..251a57cf5822 100644
--- a/net/dccp/minisocks.c
+++ b/net/dccp/minisocks.c
@@ -15,6 +15,7 @@
 #include <net/sock.h>
 #include <net/xfrm.h>
 #include <net/inet_timewait_sock.h>
+#include <net/rstreason.h>
 
 #include "ackvec.h"
 #include "ccid.h"
@@ -202,7 +203,7 @@ struct sock *dccp_check_req(struct sock *sk, struct sk_buff *skb,
 	DCCP_SKB_CB(skb)->dccpd_reset_code = DCCP_RESET_CODE_TOO_BUSY;
 drop:
 	if (dccp_hdr(skb)->dccph_type != DCCP_PKT_RESET)
-		req->rsk_ops->send_reset(sk, skb);
+		req->rsk_ops->send_reset(sk, skb, SK_RST_REASON_NOT_SPECIFIED);
 
 	inet_csk_reqsk_queue_drop(sk, req);
 out:
diff --git a/net/ipv4/tcp_ipv4.c b/net/ipv4/tcp_ipv4.c
index 52963c3bb8ca..21fa69445f7a 100644
--- a/net/ipv4/tcp_ipv4.c
+++ b/net/ipv4/tcp_ipv4.c
@@ -70,6 +70,7 @@
 #include <net/xfrm.h>
 #include <net/secure_seq.h>
 #include <net/busy_poll.h>
+#include <net/rstreason.h>
 
 #include <linux/inet.h>
 #include <linux/ipv6.h>
@@ -723,7 +724,8 @@ static bool tcp_v4_ao_sign_reset(const struct sock *sk, struct sk_buff *skb,
  *	Exception: precedence violation. We do not implement it in any case.
  */
 
-static void tcp_v4_send_reset(const struct sock *sk, struct sk_buff *skb)
+static void tcp_v4_send_reset(const struct sock *sk, struct sk_buff *skb,
+			      int reason)
 {
 	const struct tcphdr *th = tcp_hdr(skb);
 	struct {
@@ -1933,7 +1935,7 @@ int tcp_v4_do_rcv(struct sock *sk, struct sk_buff *skb)
 	return 0;
 
 reset:
-	tcp_v4_send_reset(rsk, skb);
+	tcp_v4_send_reset(rsk, skb, SK_RST_REASON_NOT_SPECIFIED);
 discard:
 	kfree_skb_reason(skb, reason);
 	/* Be careful here. If this function gets more complicated and
@@ -2276,7 +2278,7 @@ int tcp_v4_rcv(struct sk_buff *skb)
 		} else {
 			drop_reason = tcp_child_process(sk, nsk, skb);
 			if (drop_reason) {
-				tcp_v4_send_reset(nsk, skb);
+				tcp_v4_send_reset(nsk, skb, SK_RST_REASON_NOT_SPECIFIED);
 				goto discard_and_relse;
 			}
 			sock_put(sk);
@@ -2354,7 +2356,7 @@ int tcp_v4_rcv(struct sk_buff *skb)
 bad_packet:
 		__TCP_INC_STATS(net, TCP_MIB_INERRS);
 	} else {
-		tcp_v4_send_reset(NULL, skb);
+		tcp_v4_send_reset(NULL, skb, SK_RST_REASON_NOT_SPECIFIED);
 	}
 
 discard_it:
@@ -2405,7 +2407,7 @@ int tcp_v4_rcv(struct sk_buff *skb)
 		tcp_v4_timewait_ack(sk, skb);
 		break;
 	case TCP_TW_RST:
-		tcp_v4_send_reset(sk, skb);
+		tcp_v4_send_reset(sk, skb, SK_RST_REASON_NOT_SPECIFIED);
 		inet_twsk_deschedule_put(inet_twsk(sk));
 		goto discard_it;
 	case TCP_TW_SUCCESS:;
diff --git a/net/ipv4/tcp_minisocks.c b/net/ipv4/tcp_minisocks.c
index 5b21a07ddf9a..2d7d81428c07 100644
--- a/net/ipv4/tcp_minisocks.c
+++ b/net/ipv4/tcp_minisocks.c
@@ -22,6 +22,7 @@
 #include <net/tcp.h>
 #include <net/xfrm.h>
 #include <net/busy_poll.h>
+#include <net/rstreason.h>
 
 static bool tcp_in_window(u32 seq, u32 end_seq, u32 s_win, u32 e_win)
 {
@@ -879,7 +880,7 @@ struct sock *tcp_check_req(struct sock *sk, struct sk_buff *skb,
 		 * avoid becoming vulnerable to outside attack aiming at
 		 * resetting legit local connections.
 		 */
-		req->rsk_ops->send_reset(sk, skb);
+		req->rsk_ops->send_reset(sk, skb, SK_RST_REASON_NOT_SPECIFIED);
 	} else if (fastopen) { /* received a valid RST pkt */
 		reqsk_fastopen_remove(sk, req, true);
 		tcp_reset(sk, skb);
diff --git a/net/ipv6/tcp_ipv6.c b/net/ipv6/tcp_ipv6.c
index cffebaec66f1..7e591521b193 100644
--- a/net/ipv6/tcp_ipv6.c
+++ b/net/ipv6/tcp_ipv6.c
@@ -60,6 +60,7 @@
 #include <net/secure_seq.h>
 #include <net/hotdata.h>
 #include <net/busy_poll.h>
+#include <net/rstreason.h>
 
 #include <linux/proc_fs.h>
 #include <linux/seq_file.h>
@@ -69,7 +70,8 @@
 
 #include <trace/events/tcp.h>
 
-static void	tcp_v6_send_reset(const struct sock *sk, struct sk_buff *skb);
+static void tcp_v6_send_reset(const struct sock *sk, struct sk_buff *skb,
+			      int reason);
 static void	tcp_v6_reqsk_send_ack(const struct sock *sk, struct sk_buff *skb,
 				      struct request_sock *req);
 
@@ -1006,7 +1008,8 @@ static void tcp_v6_send_response(const struct sock *sk, struct sk_buff *skb, u32
 	kfree_skb(buff);
 }
 
-static void tcp_v6_send_reset(const struct sock *sk, struct sk_buff *skb)
+static void tcp_v6_send_reset(const struct sock *sk, struct sk_buff *skb,
+			      int reason)
 {
 	const struct tcphdr *th = tcp_hdr(skb);
 	struct ipv6hdr *ipv6h = ipv6_hdr(skb);
@@ -1675,7 +1678,7 @@ int tcp_v6_do_rcv(struct sock *sk, struct sk_buff *skb)
 	return 0;
 
 reset:
-	tcp_v6_send_reset(sk, skb);
+	tcp_v6_send_reset(sk, skb, SK_RST_REASON_NOT_SPECIFIED);
 discard:
 	if (opt_skb)
 		__kfree_skb(opt_skb);
@@ -1861,7 +1864,7 @@ INDIRECT_CALLABLE_SCOPE int tcp_v6_rcv(struct sk_buff *skb)
 		} else {
 			drop_reason = tcp_child_process(sk, nsk, skb);
 			if (drop_reason) {
-				tcp_v6_send_reset(nsk, skb);
+				tcp_v6_send_reset(nsk, skb, SK_RST_REASON_NOT_SPECIFIED);
 				goto discard_and_relse;
 			}
 			sock_put(sk);
@@ -1937,7 +1940,7 @@ INDIRECT_CALLABLE_SCOPE int tcp_v6_rcv(struct sk_buff *skb)
 bad_packet:
 		__TCP_INC_STATS(net, TCP_MIB_INERRS);
 	} else {
-		tcp_v6_send_reset(NULL, skb);
+		tcp_v6_send_reset(NULL, skb, SK_RST_REASON_NOT_SPECIFIED);
 	}
 
 discard_it:
@@ -1992,7 +1995,7 @@ INDIRECT_CALLABLE_SCOPE int tcp_v6_rcv(struct sk_buff *skb)
 		tcp_v6_timewait_ack(sk, skb);
 		break;
 	case TCP_TW_RST:
-		tcp_v6_send_reset(sk, skb);
+		tcp_v6_send_reset(sk, skb, SK_RST_REASON_NOT_SPECIFIED);
 		inet_twsk_deschedule_put(inet_twsk(sk));
 		goto discard_it;
 	case TCP_TW_SUCCESS:
diff --git a/net/mptcp/subflow.c b/net/mptcp/subflow.c
index 162b218d9858..744c87b6d5a4 100644
--- a/net/mptcp/subflow.c
+++ b/net/mptcp/subflow.c
@@ -20,6 +20,8 @@
 #include <net/transp_v6.h>
 #endif
 #include <net/mptcp.h>
+#include <net/rstreason.h>
+
 #include "protocol.h"
 #include "mib.h"
 
@@ -307,7 +309,7 @@ static struct dst_entry *subflow_v4_route_req(const struct sock *sk,
 
 	dst_release(dst);
 	if (!req->syncookie)
-		tcp_request_sock_ops.send_reset(sk, skb);
+		tcp_request_sock_ops.send_reset(sk, skb, SK_RST_REASON_NOT_SPECIFIED);
 	return NULL;
 }
 
@@ -374,7 +376,7 @@ static struct dst_entry *subflow_v6_route_req(const struct sock *sk,
 
 	dst_release(dst);
 	if (!req->syncookie)
-		tcp6_request_sock_ops.send_reset(sk, skb);
+		tcp6_request_sock_ops.send_reset(sk, skb, SK_RST_REASON_NOT_SPECIFIED);
 	return NULL;
 }
 #endif
@@ -909,7 +911,7 @@ static struct sock *subflow_syn_recv_sock(const struct sock *sk,
 	tcp_rsk(req)->drop_req = true;
 	inet_csk_prepare_for_destroy_sock(child);
 	tcp_done(child);
-	req->rsk_ops->send_reset(sk, skb);
+	req->rsk_ops->send_reset(sk, skb, SK_RST_REASON_NOT_SPECIFIED);
 
 	/* The last child reference will be released by the caller */
 	return child;
-- 
2.37.3

[PATCH net-next v3 3/6] rstreason: prepare for active reset

[Jason Xing]

From: Jason Xing <kernelxing@tencent.com>

Like what we did to passive reset:
only passing possible reset reason in each active reset path.

No functional changes.

Signed-off-by: Jason Xing <kernelxing@tencent.com>
---
 include/net/tcp.h     |  2 +-
 net/ipv4/tcp.c        | 15 ++++++++++-----
 net/ipv4/tcp_output.c |  2 +-
 net/ipv4/tcp_timer.c  |  9 ++++++---
 net/mptcp/protocol.c  |  4 +++-
 net/mptcp/subflow.c   |  5 +++--
 6 files changed, 24 insertions(+), 13 deletions(-)

diff --git a/include/net/tcp.h b/include/net/tcp.h
index 9ab5b37e9d53..67ab4dbf7805 100644
--- a/include/net/tcp.h
+++ b/include/net/tcp.h
@@ -667,7 +667,7 @@ int tcp_fragment(struct sock *sk, enum tcp_queue tcp_queue,
 void tcp_send_probe0(struct sock *);
 int tcp_write_wakeup(struct sock *, int mib);
 void tcp_send_fin(struct sock *sk);
-void tcp_send_active_reset(struct sock *sk, gfp_t priority);
+void tcp_send_active_reset(struct sock *sk, gfp_t priority, int reason);
 int tcp_send_synack(struct sock *);
 void tcp_push_one(struct sock *, unsigned int mss_now);
 void __tcp_send_ack(struct sock *sk, u32 rcv_nxt);
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index 664c8ecb076b..d1610d4deb8f 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -275,6 +275,7 @@
 #include <net/xfrm.h>
 #include <net/ip.h>
 #include <net/sock.h>
+#include <net/rstreason.h>
 
 #include <linux/uaccess.h>
 #include <asm/ioctls.h>
@@ -2805,7 +2806,8 @@ void __tcp_close(struct sock *sk, long timeout)
 		/* Unread data was tossed, zap the connection. */
 		NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPABORTONCLOSE);
 		tcp_set_state(sk, TCP_CLOSE);
-		tcp_send_active_reset(sk, sk->sk_allocation);
+		tcp_send_active_reset(sk, sk->sk_allocation,
+				      SK_RST_REASON_NOT_SPECIFIED);
 	} else if (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime) {
 		/* Check zero linger _after_ checking for unread data. */
 		sk->sk_prot->disconnect(sk, 0);
@@ -2879,7 +2881,8 @@ void __tcp_close(struct sock *sk, long timeout)
 		struct tcp_sock *tp = tcp_sk(sk);
 		if (READ_ONCE(tp->linger2) < 0) {
 			tcp_set_state(sk, TCP_CLOSE);
-			tcp_send_active_reset(sk, GFP_ATOMIC);
+			tcp_send_active_reset(sk, GFP_ATOMIC,
+					      SK_RST_REASON_NOT_SPECIFIED);
 			__NET_INC_STATS(sock_net(sk),
 					LINUX_MIB_TCPABORTONLINGER);
 		} else {
@@ -2897,7 +2900,8 @@ void __tcp_close(struct sock *sk, long timeout)
 	if (sk->sk_state != TCP_CLOSE) {
 		if (tcp_check_oom(sk, 0)) {
 			tcp_set_state(sk, TCP_CLOSE);
-			tcp_send_active_reset(sk, GFP_ATOMIC);
+			tcp_send_active_reset(sk, GFP_ATOMIC,
+					      SK_RST_REASON_NOT_SPECIFIED);
 			__NET_INC_STATS(sock_net(sk),
 					LINUX_MIB_TCPABORTONMEMORY);
 		} else if (!check_net(sock_net(sk))) {
@@ -3001,7 +3005,7 @@ 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());
+		tcp_send_active_reset(sk, gfp_any(), SK_RST_REASON_NOT_SPECIFIED);
 		WRITE_ONCE(sk->sk_err, ECONNRESET);
 	} else if (old_state == TCP_SYN_SENT)
 		WRITE_ONCE(sk->sk_err, ECONNRESET);
@@ -4557,7 +4561,8 @@ int tcp_abort(struct sock *sk, int err)
 		smp_wmb();
 		sk_error_report(sk);
 		if (tcp_need_reset(sk->sk_state))
-			tcp_send_active_reset(sk, GFP_ATOMIC);
+			tcp_send_active_reset(sk, GFP_ATOMIC,
+					      SK_RST_REASON_NOT_SPECIFIED);
 		tcp_done(sk);
 	}
 
diff --git a/net/ipv4/tcp_output.c b/net/ipv4/tcp_output.c
index 9282fafc0e61..6d807b5c1b9c 100644
--- a/net/ipv4/tcp_output.c
+++ b/net/ipv4/tcp_output.c
@@ -3585,7 +3585,7 @@ void tcp_send_fin(struct sock *sk)
  * was unread data in the receive queue.  This behavior is recommended
  * by RFC 2525, section 2.17.  -DaveM
  */
-void tcp_send_active_reset(struct sock *sk, gfp_t priority)
+void tcp_send_active_reset(struct sock *sk, gfp_t priority, int reason)
 {
 	struct sk_buff *skb;
 
diff --git a/net/ipv4/tcp_timer.c b/net/ipv4/tcp_timer.c
index 976db57b95d4..83fe7f62f7f1 100644
--- a/net/ipv4/tcp_timer.c
+++ b/net/ipv4/tcp_timer.c
@@ -22,6 +22,7 @@
 #include <linux/module.h>
 #include <linux/gfp.h>
 #include <net/tcp.h>
+#include <net/rstreason.h>
 
 static u32 tcp_clamp_rto_to_user_timeout(const struct sock *sk)
 {
@@ -127,7 +128,8 @@ static int tcp_out_of_resources(struct sock *sk, bool do_reset)
 		    (!tp->snd_wnd && !tp->packets_out))
 			do_reset = true;
 		if (do_reset)
-			tcp_send_active_reset(sk, GFP_ATOMIC);
+			tcp_send_active_reset(sk, GFP_ATOMIC,
+					      SK_RST_REASON_NOT_SPECIFIED);
 		tcp_done(sk);
 		__NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPABORTONMEMORY);
 		return 1;
@@ -768,7 +770,7 @@ static void tcp_keepalive_timer (struct timer_list *t)
 				goto out;
 			}
 		}
-		tcp_send_active_reset(sk, GFP_ATOMIC);
+		tcp_send_active_reset(sk, GFP_ATOMIC, SK_RST_REASON_NOT_SPECIFIED);
 		goto death;
 	}
 
@@ -795,7 +797,8 @@ static void tcp_keepalive_timer (struct timer_list *t)
 		    icsk->icsk_probes_out > 0) ||
 		    (user_timeout == 0 &&
 		    icsk->icsk_probes_out >= keepalive_probes(tp))) {
-			tcp_send_active_reset(sk, GFP_ATOMIC);
+			tcp_send_active_reset(sk, GFP_ATOMIC,
+					      SK_RST_REASON_NOT_SPECIFIED);
 			tcp_write_err(sk);
 			goto out;
 		}
diff --git a/net/mptcp/protocol.c b/net/mptcp/protocol.c
index 995b53cd021c..212a1db3b2e3 100644
--- a/net/mptcp/protocol.c
+++ b/net/mptcp/protocol.c
@@ -21,6 +21,7 @@
 #endif
 #include <net/mptcp.h>
 #include <net/xfrm.h>
+#include <net/rstreason.h>
 #include <asm/ioctls.h>
 #include "protocol.h"
 #include "mib.h"
@@ -2565,7 +2566,8 @@ static void mptcp_check_fastclose(struct mptcp_sock *msk)
 
 		slow = lock_sock_fast(tcp_sk);
 		if (tcp_sk->sk_state != TCP_CLOSE) {
-			tcp_send_active_reset(tcp_sk, GFP_ATOMIC);
+			tcp_send_active_reset(tcp_sk, GFP_ATOMIC,
+					      SK_RST_REASON_NOT_SPECIFIED);
 			tcp_set_state(tcp_sk, TCP_CLOSE);
 		}
 		unlock_sock_fast(tcp_sk, slow);
diff --git a/net/mptcp/subflow.c b/net/mptcp/subflow.c
index 744c87b6d5a4..ba0a252c113f 100644
--- a/net/mptcp/subflow.c
+++ b/net/mptcp/subflow.c
@@ -412,7 +412,7 @@ void mptcp_subflow_reset(struct sock *ssk)
 	/* must hold: tcp_done() could drop last reference on parent */
 	sock_hold(sk);
 
-	tcp_send_active_reset(ssk, GFP_ATOMIC);
+	tcp_send_active_reset(ssk, GFP_ATOMIC, SK_RST_REASON_NOT_SPECIFIED);
 	tcp_done(ssk);
 	if (!test_and_set_bit(MPTCP_WORK_CLOSE_SUBFLOW, &mptcp_sk(sk)->flags))
 		mptcp_schedule_work(sk);
@@ -1348,7 +1348,8 @@ static bool subflow_check_data_avail(struct sock *ssk)
 			tcp_set_state(ssk, TCP_CLOSE);
 			while ((skb = skb_peek(&ssk->sk_receive_queue)))
 				sk_eat_skb(ssk, skb);
-			tcp_send_active_reset(ssk, GFP_ATOMIC);
+			tcp_send_active_reset(ssk, GFP_ATOMIC,
+					      SK_RST_REASON_NOT_SPECIFIED);
 			WRITE_ONCE(subflow->data_avail, false);
 			return false;
 		}
-- 
2.37.3

[PATCH net-next v3 4/6] tcp: support rstreason for passive reset

[Jason Xing]

From: Jason Xing <kernelxing@tencent.com>

Reuse the dropreason logic to show the exact reason of tcp reset,
so we don't need to implement those duplicated reset reasons.
This patch replaces all the prior NOT_SPECIFIED reasons.

Signed-off-by: Jason Xing <kernelxing@tencent.com>
---
 net/ipv4/tcp_ipv4.c | 8 ++++----
 net/ipv6/tcp_ipv6.c | 8 ++++----
 2 files changed, 8 insertions(+), 8 deletions(-)

diff --git a/net/ipv4/tcp_ipv4.c b/net/ipv4/tcp_ipv4.c
index 21fa69445f7a..03c5af9decbf 100644
--- a/net/ipv4/tcp_ipv4.c
+++ b/net/ipv4/tcp_ipv4.c
@@ -1935,7 +1935,7 @@ int tcp_v4_do_rcv(struct sock *sk, struct sk_buff *skb)
 	return 0;
 
 reset:
-	tcp_v4_send_reset(rsk, skb, SK_RST_REASON_NOT_SPECIFIED);
+	tcp_v4_send_reset(rsk, skb, reason);
 discard:
 	kfree_skb_reason(skb, reason);
 	/* Be careful here. If this function gets more complicated and
@@ -2278,7 +2278,7 @@ int tcp_v4_rcv(struct sk_buff *skb)
 		} else {
 			drop_reason = tcp_child_process(sk, nsk, skb);
 			if (drop_reason) {
-				tcp_v4_send_reset(nsk, skb, SK_RST_REASON_NOT_SPECIFIED);
+				tcp_v4_send_reset(nsk, skb, drop_reason);
 				goto discard_and_relse;
 			}
 			sock_put(sk);
@@ -2356,7 +2356,7 @@ int tcp_v4_rcv(struct sk_buff *skb)
 bad_packet:
 		__TCP_INC_STATS(net, TCP_MIB_INERRS);
 	} else {
-		tcp_v4_send_reset(NULL, skb, SK_RST_REASON_NOT_SPECIFIED);
+		tcp_v4_send_reset(NULL, skb, drop_reason);
 	}
 
 discard_it:
@@ -2407,7 +2407,7 @@ int tcp_v4_rcv(struct sk_buff *skb)
 		tcp_v4_timewait_ack(sk, skb);
 		break;
 	case TCP_TW_RST:
-		tcp_v4_send_reset(sk, skb, SK_RST_REASON_NOT_SPECIFIED);
+		tcp_v4_send_reset(sk, skb, drop_reason);
 		inet_twsk_deschedule_put(inet_twsk(sk));
 		goto discard_it;
 	case TCP_TW_SUCCESS:;
diff --git a/net/ipv6/tcp_ipv6.c b/net/ipv6/tcp_ipv6.c
index 7e591521b193..6889ea70c760 100644
--- a/net/ipv6/tcp_ipv6.c
+++ b/net/ipv6/tcp_ipv6.c
@@ -1678,7 +1678,7 @@ int tcp_v6_do_rcv(struct sock *sk, struct sk_buff *skb)
 	return 0;
 
 reset:
-	tcp_v6_send_reset(sk, skb, SK_RST_REASON_NOT_SPECIFIED);
+	tcp_v6_send_reset(sk, skb, reason);
 discard:
 	if (opt_skb)
 		__kfree_skb(opt_skb);
@@ -1864,7 +1864,7 @@ INDIRECT_CALLABLE_SCOPE int tcp_v6_rcv(struct sk_buff *skb)
 		} else {
 			drop_reason = tcp_child_process(sk, nsk, skb);
 			if (drop_reason) {
-				tcp_v6_send_reset(nsk, skb, SK_RST_REASON_NOT_SPECIFIED);
+				tcp_v6_send_reset(nsk, skb, drop_reason);
 				goto discard_and_relse;
 			}
 			sock_put(sk);
@@ -1940,7 +1940,7 @@ INDIRECT_CALLABLE_SCOPE int tcp_v6_rcv(struct sk_buff *skb)
 bad_packet:
 		__TCP_INC_STATS(net, TCP_MIB_INERRS);
 	} else {
-		tcp_v6_send_reset(NULL, skb, SK_RST_REASON_NOT_SPECIFIED);
+		tcp_v6_send_reset(NULL, skb, drop_reason);
 	}
 
 discard_it:
@@ -1995,7 +1995,7 @@ INDIRECT_CALLABLE_SCOPE int tcp_v6_rcv(struct sk_buff *skb)
 		tcp_v6_timewait_ack(sk, skb);
 		break;
 	case TCP_TW_RST:
-		tcp_v6_send_reset(sk, skb, SK_RST_REASON_NOT_SPECIFIED);
+		tcp_v6_send_reset(sk, skb, drop_reason);
 		inet_twsk_deschedule_put(inet_twsk(sk));
 		goto discard_it;
 	case TCP_TW_SUCCESS:
-- 
2.37.3

[PATCH net-next v3 5/6] mptcp: support rstreason for passive reset

[Jason Xing]

From: Jason Xing <kernelxing@tencent.com>

It relys on what reset options in the skb are as rfc8684 says. Reusing
this logic can save us much energy. This patch replaces most of the prior
NOT_SPECIFIED reasons.

Signed-off-by: Jason Xing <kernelxing@tencent.com>
---
 net/mptcp/subflow.c | 18 +++++++++++++-----
 1 file changed, 13 insertions(+), 5 deletions(-)

diff --git a/net/mptcp/subflow.c b/net/mptcp/subflow.c
index ba0a252c113f..4f2be72d5b02 100644
--- a/net/mptcp/subflow.c
+++ b/net/mptcp/subflow.c
@@ -308,8 +308,11 @@ static struct dst_entry *subflow_v4_route_req(const struct sock *sk,
 		return dst;
 
 	dst_release(dst);
-	if (!req->syncookie)
-		tcp_request_sock_ops.send_reset(sk, skb, SK_RST_REASON_NOT_SPECIFIED);
+	if (!req->syncookie) {
+		struct mptcp_ext *mpext = mptcp_get_ext(skb);
+
+		tcp_request_sock_ops.send_reset(sk, skb, mpext->reset_reason);
+	}
 	return NULL;
 }
 
@@ -375,8 +378,11 @@ static struct dst_entry *subflow_v6_route_req(const struct sock *sk,
 		return dst;
 
 	dst_release(dst);
-	if (!req->syncookie)
-		tcp6_request_sock_ops.send_reset(sk, skb, SK_RST_REASON_NOT_SPECIFIED);
+	if (!req->syncookie) {
+		struct mptcp_ext *mpext = mptcp_get_ext(skb);
+
+		tcp6_request_sock_ops.send_reset(sk, skb, mpext->reset_reason);
+	}
 	return NULL;
 }
 #endif
@@ -783,6 +789,7 @@ static struct sock *subflow_syn_recv_sock(const struct sock *sk,
 	bool fallback, fallback_is_fatal;
 	struct mptcp_sock *owner;
 	struct sock *child;
+	int reason;
 
 	pr_debug("listener=%p, req=%p, conn=%p", listener, req, listener->conn);
 
@@ -911,7 +918,8 @@ static struct sock *subflow_syn_recv_sock(const struct sock *sk,
 	tcp_rsk(req)->drop_req = true;
 	inet_csk_prepare_for_destroy_sock(child);
 	tcp_done(child);
-	req->rsk_ops->send_reset(sk, skb, SK_RST_REASON_NOT_SPECIFIED);
+	reason = mptcp_get_ext(skb)->reset_reason;
+	req->rsk_ops->send_reset(sk, skb, convert_mptcp_reason(reason));
 
 	/* The last child reference will be released by the caller */
 	return child;
-- 
2.37.3

[PATCH net-next v3 6/6] rstreason: make it work in trace world

[Jason Xing]

From: Jason Xing <kernelxing@tencent.com>

At last, we should let it work by introducing this reset reason in
trace world.

One of the possible expected outputs is:
... tcp_send_reset: skbaddr=xxx skaddr=xxx src=xxx dest=xxx
state=TCP_ESTABLISHED reason=NOT_SPECIFIED

Signed-off-by: Jason Xing <kernelxing@tencent.com>
---
 include/trace/events/tcp.h | 37 +++++++++++++++++++++++++++++++++----
 net/ipv4/tcp_ipv4.c        |  2 +-
 net/ipv4/tcp_output.c      |  2 +-
 net/ipv6/tcp_ipv6.c        |  2 +-
 4 files changed, 36 insertions(+), 7 deletions(-)

diff --git a/include/trace/events/tcp.h b/include/trace/events/tcp.h
index 5c04a61a11c2..9bed9e63c9c5 100644
--- a/include/trace/events/tcp.h
+++ b/include/trace/events/tcp.h
@@ -11,6 +11,7 @@
 #include <net/ipv6.h>
 #include <net/tcp.h>
 #include <linux/sock_diag.h>
+#include <net/rstreason.h>
 
 /*
  * tcp event with arguments sk and skb
@@ -74,20 +75,38 @@ DEFINE_EVENT(tcp_event_sk_skb, tcp_retransmit_skb,
 	TP_ARGS(sk, skb)
 );
 
+#undef FN1
+#define FN1(reason)	TRACE_DEFINE_ENUM(SK_RST_REASON_##reason);
+#undef FN2
+#define FN2(reason)	TRACE_DEFINE_ENUM(SKB_DROP_REASON_##reason);
+DEFINE_RST_REASON(FN1, FN1)
+
+#undef FN1
+#undef FNe1
+#define FN1(reason)	{ SK_RST_REASON_##reason, #reason },
+#define FNe1(reason)	{ SK_RST_REASON_##reason, #reason }
+
+#undef FN2
+#undef FNe2
+#define FN2(reason)	{ SKB_DROP_REASON_##reason, #reason },
+#define FNe2(reason)	{ SKB_DROP_REASON_##reason, #reason }
 /*
  * skb of trace_tcp_send_reset is the skb that caused RST. In case of
  * active reset, skb should be NULL
  */
 TRACE_EVENT(tcp_send_reset,
 
-	TP_PROTO(const struct sock *sk, const struct sk_buff *skb),
+	TP_PROTO(const struct sock *sk,
+		 const struct sk_buff *skb,
+		 const int reason),
 
-	TP_ARGS(sk, skb),
+	TP_ARGS(sk, skb, reason),
 
 	TP_STRUCT__entry(
 		__field(const void *, skbaddr)
 		__field(const void *, skaddr)
 		__field(int, state)
+		__field(int, reason)
 		__array(__u8, saddr, sizeof(struct sockaddr_in6))
 		__array(__u8, daddr, sizeof(struct sockaddr_in6))
 	),
@@ -113,14 +132,24 @@ TRACE_EVENT(tcp_send_reset,
 			 */
 			TP_STORE_ADDR_PORTS_SKB(skb, th, entry->daddr, entry->saddr);
 		}
+		__entry->reason = reason;
 	),
 
-	TP_printk("skbaddr=%p skaddr=%p src=%pISpc dest=%pISpc state=%s",
+	TP_printk("skbaddr=%p skaddr=%p src=%pISpc dest=%pISpc state=%s reason=%s",
 		  __entry->skbaddr, __entry->skaddr,
 		  __entry->saddr, __entry->daddr,
-		  __entry->state ? show_tcp_state_name(__entry->state) : "UNKNOWN")
+		  __entry->state ? show_tcp_state_name(__entry->state) : "UNKNOWN",
+		  __entry->reason < RST_REASON_START ?
+			__print_symbolic(__entry->reason, DEFINE_DROP_REASON(FN2, FNe2)) :
+			__print_symbolic(__entry->reason, DEFINE_RST_REASON(FN1, FNe1)))
 );
 
+#undef FN1
+#undef FNe1
+
+#undef FN2
+#undef FNe2
+
 /*
  * tcp event with arguments sk
  *
diff --git a/net/ipv4/tcp_ipv4.c b/net/ipv4/tcp_ipv4.c
index 03c5af9decbf..4889fccbf754 100644
--- a/net/ipv4/tcp_ipv4.c
+++ b/net/ipv4/tcp_ipv4.c
@@ -871,7 +871,7 @@ static void tcp_v4_send_reset(const struct sock *sk, struct sk_buff *skb,
 	if (sk)
 		arg.bound_dev_if = sk->sk_bound_dev_if;
 
-	trace_tcp_send_reset(sk, skb);
+	trace_tcp_send_reset(sk, skb, reason);
 
 	BUILD_BUG_ON(offsetof(struct sock, sk_bound_dev_if) !=
 		     offsetof(struct inet_timewait_sock, tw_bound_dev_if));
diff --git a/net/ipv4/tcp_output.c b/net/ipv4/tcp_output.c
index 6d807b5c1b9c..710922f7d4d6 100644
--- a/net/ipv4/tcp_output.c
+++ b/net/ipv4/tcp_output.c
@@ -3610,7 +3610,7 @@ void tcp_send_active_reset(struct sock *sk, gfp_t priority, int reason)
 	/* 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);
+	trace_tcp_send_reset(sk, NULL, SK_RST_REASON_NOT_SPECIFIED);
 }
 
 /* Send a crossed SYN-ACK during socket establishment.
diff --git a/net/ipv6/tcp_ipv6.c b/net/ipv6/tcp_ipv6.c
index 6889ea70c760..3c995eff6e52 100644
--- a/net/ipv6/tcp_ipv6.c
+++ b/net/ipv6/tcp_ipv6.c
@@ -1131,7 +1131,7 @@ static void tcp_v6_send_reset(const struct sock *sk, struct sk_buff *skb,
 			label = ip6_flowlabel(ipv6h);
 	}
 
-	trace_tcp_send_reset(sk, skb);
+	trace_tcp_send_reset(sk, skb, reason);
 
 	tcp_v6_send_response(sk, skb, seq, ack_seq, 0, 0, 0, oif, 1,
 			     ipv6_get_dsfield(ipv6h), label, priority, txhash,
-- 
2.37.3

Re: [PATCH net-next v3 0/6] Implement reset reason mechanism to detect

[MPTCP CI]

Hi Jason,

Thank you for your modifications, that's great!

Our CI did some validations and here is its report:

- KVM Validation: normal: Success! ✅
- KVM Validation: debug: Unstable: 1 failed test(s): packetdrill_sockopts 🔴
- KVM Validation: btf (only bpftest_all): Success! ✅
- Task: https://github.com/multipath-tcp/mptcp_net-next/actions/runs/8614129316

Initiator: Patchew Applier
Commits: https://github.com/multipath-tcp/mptcp_net-next/commits/a1b3417a8588
Patchwork: https://patchwork.kernel.org/project/mptcp/list/?series=842746


If there are some issues, you can reproduce them using the same environment as
the one used by the CI thanks to a docker image, e.g.:

    $ cd [kernel source code]
    $ docker run -v "${PWD}:${PWD}:rw" -w "${PWD}" --privileged --rm -it \
        --pull always mptcp/mptcp-upstream-virtme-docker:latest \
        auto-normal

For more details:

    https://github.com/multipath-tcp/mptcp-upstream-virtme-docker


Please note that despite all the efforts that have been already done to have a
stable tests suite when executed on a public CI like here, it is possible some
reported issues are not due to your modifications. Still, do not hesitate to
help us improve that ;-)

Cheers,
MPTCP GH Action bot
Bot operated by Matthieu Baerts (NGI0 Core)

Re: [PATCH net-next v3 6/6] rstreason: make it work in trace world

[Steven Rostedt]

On Tue,  9 Apr 2024 18:09:34 +0800
Jason Xing <kerneljasonxing@gmail.com> wrote:

>  /*
>   * tcp event with arguments sk and skb
> @@ -74,20 +75,38 @@ DEFINE_EVENT(tcp_event_sk_skb, tcp_retransmit_skb,
>  	TP_ARGS(sk, skb)
>  );
>  
> +#undef FN1
> +#define FN1(reason)	TRACE_DEFINE_ENUM(SK_RST_REASON_##reason);
> +#undef FN2
> +#define FN2(reason)	TRACE_DEFINE_ENUM(SKB_DROP_REASON_##reason);
> +DEFINE_RST_REASON(FN1, FN1)

Interesting. I've never seen the passing of the internal macros to the main
macro before. I see that you are using it for handling both the
SK_RST_REASON and the SK_DROP_REASON.

> +
> +#undef FN1
> +#undef FNe1
> +#define FN1(reason)	{ SK_RST_REASON_##reason, #reason },
> +#define FNe1(reason)	{ SK_RST_REASON_##reason, #reason }
> +
> +#undef FN2
> +#undef FNe2
> +#define FN2(reason)	{ SKB_DROP_REASON_##reason, #reason },
> +#define FNe2(reason)	{ SKB_DROP_REASON_##reason, #reason }

Anyway, from a tracing point of view, as it looks like it would work
(I haven't tested it).

Reviewed-by: Steven Rostedt (Google) <rostedt@goodmis.org>

-- Steve

Re: [PATCH net-next v3 6/6] rstreason: make it work in trace world

[Jason Xing]

Hi Steven,

On Tue, Apr 9, 2024 at 11:36 PM Steven Rostedt <rostedt@goodmis.org> wrote:
>
> On Tue,  9 Apr 2024 18:09:34 +0800
> Jason Xing <kerneljasonxing@gmail.com> wrote:
>
> >  /*
> >   * tcp event with arguments sk and skb
> > @@ -74,20 +75,38 @@ DEFINE_EVENT(tcp_event_sk_skb, tcp_retransmit_skb,
> >       TP_ARGS(sk, skb)
> >  );
> >
> > +#undef FN1
> > +#define FN1(reason)  TRACE_DEFINE_ENUM(SK_RST_REASON_##reason);
> > +#undef FN2
> > +#define FN2(reason)  TRACE_DEFINE_ENUM(SKB_DROP_REASON_##reason);
> > +DEFINE_RST_REASON(FN1, FN1)
>
> Interesting. I've never seen the passing of the internal macros to the main
> macro before. I see that you are using it for handling both the
> SK_RST_REASON and the SK_DROP_REASON.

Yes, I want to cover two kinds of reasons and then strip them of
prefixes which can be reported to userspace.

>
> > +
> > +#undef FN1
> > +#undef FNe1
> > +#define FN1(reason)  { SK_RST_REASON_##reason, #reason },
> > +#define FNe1(reason) { SK_RST_REASON_##reason, #reason }
> > +
> > +#undef FN2
> > +#undef FNe2
> > +#define FN2(reason)  { SKB_DROP_REASON_##reason, #reason },
> > +#define FNe2(reason) { SKB_DROP_REASON_##reason, #reason }
>
> Anyway, from a tracing point of view, as it looks like it would work
> (I haven't tested it).

Sure, it works. One simple test if you're interested:
1) Apply this patchset locally
2) add 'trace_tcp_send_reset(sk, skb, [one reason])' in the receive
path, say, somewhere in the tcp_v4_rcv()

The possible result can be seen in the cover letter. I list here:
<idle>-0       [002] ..s1.  1830.262425: tcp_send_reset: skbaddr=x
        skaddr=x src=x dest=x state=x reason=NOT_SPECIFIED
<idle>-0       [002] ..s1.  1830.262425: tcp_send_reset: skbaddr=x
        skaddr=x src=x dest=x state=x reason=NO_SOCKET

>
> Reviewed-by: Steven Rostedt (Google) <rostedt@goodmis.org>

Thanks!

>
> -- Steve

Re: [PATCH net-next v3 2/6] rstreason: prepare for passive reset

[Antoine Tenart]

Quoting Jason Xing (2024-04-09 12:09:30)
>         void            (*send_reset)(const struct sock *sk,
> -                                     struct sk_buff *skb);
> +                                     struct sk_buff *skb,
> +                                     int reason);

I get that 'int' is used instead of 'enum sk_rst_reason' to allow
passing drop reasons too without casting, but that makes understanding
what should be 'reason' harder. Eg. when looking at the code or when
using BTF (to then install debugging probes with BPF) this is not
obvious.

A similar approach could be done as the one used for drop reasons: enum
skb_drop_reason is used for parameters (eg. kfree_skb_reason) but other
valid values (subsystem drop reasons) can be used too if casted (to
u32). We could use 'enum sk_rst_reason' and cast the other values. WDYT?

Thanks,
Antoine

Re: [PATCH net-next v3 2/6] rstreason: prepare for passive reset

[Jason Xing]

Hi Antoine,

On Wed, Apr 10, 2024 at 8:14 PM Antoine Tenart <atenart@kernel.org> wrote:
>
> Quoting Jason Xing (2024-04-09 12:09:30)
> >         void            (*send_reset)(const struct sock *sk,
> > -                                     struct sk_buff *skb);
> > +                                     struct sk_buff *skb,
> > +                                     int reason);
>
> I get that 'int' is used instead of 'enum sk_rst_reason' to allow
> passing drop reasons too without casting, but that makes understanding

Yes!

> what should be 'reason' harder. Eg. when looking at the code or when
> using BTF (to then install debugging probes with BPF) this is not
> obvious.

Only one number if we want to extract the reason with BPF, right? I
haven't tried it.

>
> A similar approach could be done as the one used for drop reasons: enum
> skb_drop_reason is used for parameters (eg. kfree_skb_reason) but other
> valid values (subsystem drop reasons) can be used too if casted (to
> u32). We could use 'enum sk_rst_reason' and cast the other values. WDYT?

I have been haunted by this 'issue' for a long time...

Are you suggesting doing so as below for readability:
1) replace the reason parameter in all the related functions (like
.send_reset(), tcp_v4_send_reset(), etc) by using 'enum sk_rst_reason'
type?
2) in patch [4/6], when it needs to pass the specific reason in those
functions, we can cast it to 'enum sk_rst_reason'?

One modification I just made based on this patchset if I understand correctly:
diff --git a/net/ipv4/tcp_ipv4.c b/net/ipv4/tcp_ipv4.c
index 4889fccbf754..e0419b8496b5 100644
--- a/net/ipv4/tcp_ipv4.c
+++ b/net/ipv4/tcp_ipv4.c
@@ -725,7 +725,7 @@ static bool tcp_v4_ao_sign_reset(const struct sock
*sk, struct sk_buff *skb,
  */

 static void tcp_v4_send_reset(const struct sock *sk, struct sk_buff *skb,
-                             int reason)
+                             enum sk_rst_reason reason)
 {
        const struct tcphdr *th = tcp_hdr(skb);
        struct {
@@ -1935,7 +1935,7 @@ int tcp_v4_do_rcv(struct sock *sk, struct sk_buff *skb)
        return 0;

 reset:
-       tcp_v4_send_reset(rsk, skb, reason);
+       tcp_v4_send_reset(rsk, skb, (enum sk_rst_reason)reason);
 discard:
        kfree_skb_reason(skb, reason);
        /* Be careful here. If this function gets more complicated and

It, indeed, looks better (clean and clear) :)

So I also ought to adjust the trace_tcp_send_reset part...

Thanks,
Jason

Re: [PATCH net-next v3 2/6] rstreason: prepare for passive reset

[Antoine Tenart]

Quoting Jason Xing (2024-04-10 14:54:51)
> Hi Antoine,
> 
> On Wed, Apr 10, 2024 at 8:14 PM Antoine Tenart <atenart@kernel.org> wrote:
> >
> > Quoting Jason Xing (2024-04-09 12:09:30)
> > >         void            (*send_reset)(const struct sock *sk,
> > > -                                     struct sk_buff *skb);
> > > +                                     struct sk_buff *skb,
> > > +                                     int reason);
> 
> > what should be 'reason' harder. Eg. when looking at the code or when
> > using BTF (to then install debugging probes with BPF) this is not
> > obvious.
> 
> Only one number if we want to extract the reason with BPF, right? I
> haven't tried it.

Yes, we can get 'reason'. Knowing the type helps.

> > A similar approach could be done as the one used for drop reasons: enum
> > skb_drop_reason is used for parameters (eg. kfree_skb_reason) but other
> > valid values (subsystem drop reasons) can be used too if casted (to
> > u32). We could use 'enum sk_rst_reason' and cast the other values. WDYT?
> 
> I have been haunted by this 'issue' for a long time...
> 
> Are you suggesting doing so as below for readability:
> 1) replace the reason parameter in all the related functions (like
> .send_reset(), tcp_v4_send_reset(), etc) by using 'enum sk_rst_reason'
> type?
> 2) in patch [4/6], when it needs to pass the specific reason in those
> functions, we can cast it to 'enum sk_rst_reason'?
> 
> One modification I just made based on this patchset if I understand correctly:
> diff --git a/net/ipv4/tcp_ipv4.c b/net/ipv4/tcp_ipv4.c
> index 4889fccbf754..e0419b8496b5 100644
> --- a/net/ipv4/tcp_ipv4.c
> +++ b/net/ipv4/tcp_ipv4.c
> @@ -725,7 +725,7 @@ static bool tcp_v4_ao_sign_reset(const struct sock
> *sk, struct sk_buff *skb,
>   */
> 
>  static void tcp_v4_send_reset(const struct sock *sk, struct sk_buff *skb,
> -                             int reason)
> +                             enum sk_rst_reason reason)
>  {
>         const struct tcphdr *th = tcp_hdr(skb);
>         struct {
> @@ -1935,7 +1935,7 @@ int tcp_v4_do_rcv(struct sock *sk, struct sk_buff *skb)
>         return 0;
> 
>  reset:
> -       tcp_v4_send_reset(rsk, skb, reason);
> +       tcp_v4_send_reset(rsk, skb, (enum sk_rst_reason)reason);
>  discard:
>         kfree_skb_reason(skb, reason);
>         /* Be careful here. If this function gets more complicated and
> 

That's right. I think (u32) can also be used for the cast to make the
compiler happy in 2), but the above makes sense.

Thanks!
Antoine

Re: [PATCH net-next v3 2/6] rstreason: prepare for passive reset

[Jason Xing]

On Wed, Apr 10, 2024 at 9:21 PM Antoine Tenart <atenart@kernel.org> wrote:
>
> Quoting Jason Xing (2024-04-10 14:54:51)
> > Hi Antoine,
> >
> > On Wed, Apr 10, 2024 at 8:14 PM Antoine Tenart <atenart@kernel.org> wrote:
> > >
> > > Quoting Jason Xing (2024-04-09 12:09:30)
> > > >         void            (*send_reset)(const struct sock *sk,
> > > > -                                     struct sk_buff *skb);
> > > > +                                     struct sk_buff *skb,
> > > > +                                     int reason);
> >
> > > what should be 'reason' harder. Eg. when looking at the code or when
> > > using BTF (to then install debugging probes with BPF) this is not
> > > obvious.
> >
> > Only one number if we want to extract the reason with BPF, right? I
> > haven't tried it.
>
> Yes, we can get 'reason'. Knowing the type helps.
>
> > > A similar approach could be done as the one used for drop reasons: enum
> > > skb_drop_reason is used for parameters (eg. kfree_skb_reason) but other
> > > valid values (subsystem drop reasons) can be used too if casted (to
> > > u32). We could use 'enum sk_rst_reason' and cast the other values. WDYT?
> >
> > I have been haunted by this 'issue' for a long time...
> >
> > Are you suggesting doing so as below for readability:
> > 1) replace the reason parameter in all the related functions (like
> > .send_reset(), tcp_v4_send_reset(), etc) by using 'enum sk_rst_reason'
> > type?
> > 2) in patch [4/6], when it needs to pass the specific reason in those
> > functions, we can cast it to 'enum sk_rst_reason'?
> >
> > One modification I just made based on this patchset if I understand correctly:
> > diff --git a/net/ipv4/tcp_ipv4.c b/net/ipv4/tcp_ipv4.c
> > index 4889fccbf754..e0419b8496b5 100644
> > --- a/net/ipv4/tcp_ipv4.c
> > +++ b/net/ipv4/tcp_ipv4.c
> > @@ -725,7 +725,7 @@ static bool tcp_v4_ao_sign_reset(const struct sock
> > *sk, struct sk_buff *skb,
> >   */
> >
> >  static void tcp_v4_send_reset(const struct sock *sk, struct sk_buff *skb,
> > -                             int reason)
> > +                             enum sk_rst_reason reason)
> >  {
> >         const struct tcphdr *th = tcp_hdr(skb);
> >         struct {
> > @@ -1935,7 +1935,7 @@ int tcp_v4_do_rcv(struct sock *sk, struct sk_buff *skb)
> >         return 0;
> >
> >  reset:
> > -       tcp_v4_send_reset(rsk, skb, reason);
> > +       tcp_v4_send_reset(rsk, skb, (enum sk_rst_reason)reason);
> >  discard:
> >         kfree_skb_reason(skb, reason);
> >         /* Be careful here. If this function gets more complicated and
> >
>
> That's right. I think (u32) can also be used for the cast to make the
> compiler happy in 2), but the above makes sense.

Got it :) Will update soon.

Thanks,
Jason