Skip to contentSkip to conversation

MP_FAIL echo and retrans

PATCH Superseded Geliang Tang 3 participants 6 messages Mar 9, 2022 – Mar 10, 2022

Subsystems (3)

v3 of 5 Superseded Latest is v5 →

lore
Share & tools
lore
#0 lore raw

v3 cover MP_FAIL echo and retrans

v3:
- use msk->sk_timer and a msk->flags bit.
- use READ_ONCE/WRITE_ONCE for subflow->mp_fail_response_expect.
- update selftest.

v2:
- don't clear mp_fail_response_expect flag.
- add a helper mp_fail_response_expect_subflow to get the subflow, instead of using msk->first.
- add locks as Mat suggested.
- add a selftest.

Geliang Tang (3):
mptcp: add MP_FAIL response support
mptcp: reset subflow when MP_FAIL doesn't respond
selftests: mptcp: check MP_FAIL response mibs

 net/mptcp/pm.c                                | 11 ++++--
 net/mptcp/protocol.c                          | 21 +++++++++++
 net/mptcp/protocol.h                          |  2 ++
 net/mptcp/subflow.c                           |  4 +++
 .../testing/selftests/net/mptcp/mptcp_join.sh | 36 ++++++++++++++++---
 5 files changed, 67 insertions(+), 7 deletions(-)
Signature · 1 line
2.34.1
#1 lore raw

PATCH v3 1/3 mptcp: add MP_FAIL response support

This patch added a new struct member mp_fail_response_expect in struct mptcp_subflow_context to support MP_FAIL response. In the single subflow with checksum error and contiguous data special case, a MP_FAIL sent in response to another MP_FAIL.

---
 net/mptcp/pm.c       | 9 +++++++--
 net/mptcp/protocol.h | 1 +
 net/mptcp/subflow.c  | 2 ++
 3 files changed, 10 insertions(+), 2 deletions(-)
net/mptcp/pm.c +7 -2
index d0d31d5c198a..c4622b07b7f5 100644
@@ -279,8 +279,13 @@ void mptcp_pm_mp_fail_received(struct sock *sk, u64 fail_seq)
 
 	pr_debug("fail_seq=%llu", fail_seq);
 
-	if (!mptcp_has_another_subflow(sk) && READ_ONCE(msk->allow_infinite_fallback))
-		subflow->send_infinite_map = 1;
+	if (!mptcp_has_another_subflow(sk) && READ_ONCE(msk->allow_infinite_fallback)) {
+		if (!READ_ONCE(subflow->mp_fail_response_expect)) {
+			subflow->send_mp_fail = 1;
+			MPTCP_INC_STATS(sock_net(sk), MPTCP_MIB_MPFAILTX);
+			subflow->send_infinite_map = 1;
+		}
+	}
 }
 
 /* path manager helpers */
net/mptcp/protocol.h +1 -0
index c8bada4537e2..83f0205f0d95 100644
@@ -456,6 +456,7 @@ struct mptcp_subflow_context {
 		stale : 1,	    /* unable to snd/rcv data, do not use for xmit */
 		local_id_valid : 1; /* local_id is correctly initialized */
 	enum mptcp_data_avail data_avail;
+	bool	mp_fail_response_expect;
 	u32	remote_nonce;
 	u64	thmac;
 	u32	local_nonce;
net/mptcp/subflow.c +2 -0
index 30ffb00661bb..ca2352ad20d4 100644
@@ -1217,6 +1217,8 @@ static bool subflow_check_data_avail(struct sock *ssk)
 				tcp_send_active_reset(ssk, GFP_ATOMIC);
 				while ((skb = skb_peek(&ssk->sk_receive_queue)))
 					sk_eat_skb(ssk, skb);
+			} else {
+				WRITE_ONCE(subflow->mp_fail_response_expect, true);
 			}
 			WRITE_ONCE(subflow->data_avail, MPTCP_SUBFLOW_NODATA);
 			return true;
Signature · 1 line
2.34.1
#2 lore raw

PATCH v3 2/3 mptcp: reset subflow when MP_FAIL doesn't respond

In reply to Geliang Tang

This patch added a new msk->flags bit MPTCP_FAIL_NO_RESPONSE, then reused sk_timer to trigger a check if we have not received a response from the peer after sending MP_FAIL. If the peer doesn't respond properly, reset the subflow.

---
 net/mptcp/pm.c       |  2 ++
 net/mptcp/protocol.c | 21 +++++++++++++++++++++
 net/mptcp/protocol.h |  1 +
 net/mptcp/subflow.c  |  2 ++
 4 files changed, 26 insertions(+)
net/mptcp/pm.c +2 -0
index c4622b07b7f5..c41aca97ba36 100644
@@ -284,6 +284,8 @@ void mptcp_pm_mp_fail_received(struct sock *sk, u64 fail_seq)
 			subflow->send_mp_fail = 1;
 			MPTCP_INC_STATS(sock_net(sk), MPTCP_MIB_MPFAILTX);
 			subflow->send_infinite_map = 1;
+		} else {
+			clear_bit(MPTCP_FAIL_NO_RESPONSE, &msk->flags);
 		}
 	}
 }
net/mptcp/protocol.c +21 -0
index 3cb975227d12..6ff7e6eb44ee 100644
@@ -2494,6 +2494,24 @@ static void __mptcp_retrans(struct sock *sk)
 		mptcp_reset_timer(sk);
 }
 
+static void mptcp_mp_fail_no_response(struct mptcp_sock *msk)
+{
+	struct mptcp_subflow_context *subflow;
+
+	mptcp_for_each_subflow(msk, subflow) {
+		if (READ_ONCE(subflow->mp_fail_response_expect)) {
+			struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
+			bool slow;
+
+			pr_debug("MP_FAIL doesn't respond, reset the subflow");
+			slow = lock_sock_fast(ssk);
+			mptcp_subflow_reset(ssk);
+			unlock_sock_fast(ssk, slow);
+			break;
+		}
+	}
+}
+
 static void mptcp_worker(struct work_struct *work)
 {
 	struct mptcp_sock *msk = container_of(work, struct mptcp_sock, work);
@@ -2534,6 +2552,9 @@ static void mptcp_worker(struct work_struct *work)
 	if (test_and_clear_bit(MPTCP_WORK_RTX, &msk->flags))
 		__mptcp_retrans(sk);
 
+	if (test_and_clear_bit(MPTCP_FAIL_NO_RESPONSE, &msk->flags))
+		mptcp_mp_fail_no_response(msk);
+
 unlock:
 	release_sock(sk);
 	sock_put(sk);
net/mptcp/protocol.h +1 -0
index 83f0205f0d95..bf58d3c886f5 100644
@@ -116,6 +116,7 @@
 #define MPTCP_WORK_EOF		3
 #define MPTCP_FALLBACK_DONE	4
 #define MPTCP_WORK_CLOSE_SUBFLOW 5
+#define MPTCP_FAIL_NO_RESPONSE	6
 
 /* MPTCP socket release cb flags */
 #define MPTCP_PUSH_PENDING	1
net/mptcp/subflow.c +2 -0
index ca2352ad20d4..5a7aa6e37ca6 100644
@@ -1009,6 +1009,7 @@ static enum mapping_status get_mapping_status(struct sock *ssk,
 		pr_debug("infinite mapping received");
 		MPTCP_INC_STATS(sock_net(ssk), MPTCP_MIB_INFINITEMAPRX);
 		subflow->map_data_len = 0;
+		clear_bit(MPTCP_FAIL_NO_RESPONSE, &msk->flags);
 		return MAPPING_INVALID;
 	}
 
@@ -1219,6 +1220,7 @@ static bool subflow_check_data_avail(struct sock *ssk)
 					sk_eat_skb(ssk, skb);
 			} else {
 				WRITE_ONCE(subflow->mp_fail_response_expect, true);
+				set_bit(MPTCP_FAIL_NO_RESPONSE, &msk->flags);
 			}
 			WRITE_ONCE(subflow->data_avail, MPTCP_SUBFLOW_NODATA);
 			return true;
Signature · 1 line
2.34.1
#4 lore raw

PATCH v3 3/3 selftests: mptcp: check MP_FAIL response mibs

In reply to Geliang Tang

This patch extended chk_fail_nr to check the MP_FAIL response mibs.

Added a new argument invert for chk_fail_nr to allow it can check the MP_FAIL Tx and Rx mibs from the opposite direction.

When the infinite map was received before the MP_FAIL response, the response will be lost. A '-' can be added into fail_tx or fail_rx to represent that MP_FAIL response Tx or Rx can be lost when doing the checks.

---
 .../testing/selftests/net/mptcp/mptcp_join.sh | 36 ++++++++++++++++---
 1 file changed, 31 insertions(+), 5 deletions(-)
tools/testing/selftests/net/mptcp/mptcp_join.sh +31 -5
index 959e46122a84..294396225f31 100755
@@ -1055,13 +1055,35 @@ chk_fail_nr()
 {
 	local fail_tx=$1
 	local fail_rx=$2
+	local ns_invert=${3:-""}
 	local count
 	local dump_stats
+	local ns_tx=$ns1
+	local ns_rx=$ns2
+	local extra_msg=""
+	local allow_tx_lost=0
+	local allow_rx_lost=0
+
+	if [[ $ns_invert = "invert" ]]; then
+		ns_tx=$ns2
+		ns_rx=$ns1
+		extra_msg="   invert"
+	fi
+
+	if [[ "${fail_tx}" = "-"* ]]; then
+		allow_tx_lost=1
+		fail_tx=${fail_tx:1}
+	fi
+	if [[ "${fail_rx}" = "-"* ]]; then
+		allow_rx_lost=1
+		fail_rx=${fail_rx:1}
+	fi
 
 	printf "%-${nr_blank}s %s" " " "ftx"
-	count=$(ip netns exec $ns1 nstat -as | grep MPTcpExtMPFailTx | awk '{print $2}')
+	count=$(ip netns exec $ns_tx nstat -as | grep MPTcpExtMPFailTx | awk '{print $2}')
 	[ -z "$count" ] && count=0
-	if [ "$count" != "$fail_tx" ]; then
+	if { [ "$count" != "$fail_tx" ] && [ $allow_tx_lost -eq 0 ]; } ||
+	   { [ "$count" -gt "$fail_tx" ] && [ $allow_tx_lost -eq 1 ]; }; then
 		echo "[fail] got $count MP_FAIL[s] TX expected $fail_tx"
 		fail_test
 		dump_stats=1
@@ -1070,17 +1092,20 @@ chk_fail_nr()
 	fi
 
 	echo -n " - failrx"
-	count=$(ip netns exec $ns2 nstat -as | grep MPTcpExtMPFailRx | awk '{print $2}')
+	count=$(ip netns exec $ns_rx nstat -as | grep MPTcpExtMPFailRx | awk '{print $2}')
 	[ -z "$count" ] && count=0
-	if [ "$count" != "$fail_rx" ]; then
+	if { [ "$count" != "$fail_rx" ] && [ $allow_rx_lost -eq 0 ]; } ||
+	   { [ "$count" -gt "$fail_rx" ] && [ $allow_rx_lost -eq 1 ]; }; then
 		echo "[fail] got $count MP_FAIL[s] RX expected $fail_rx"
 		fail_test
 		dump_stats=1
 	else
-		echo "[ ok ]"
+		echo -n "[ ok ]"
 	fi
 
 	[ "${dump_stats}" = 1 ] && dump_stats
+
+	echo "$extra_msg"
 }
 
 chk_fclose_nr()
@@ -2672,6 +2697,7 @@ fail_tests()
 	if reset_with_fail "Infinite map" 1; then
 		run_tests $ns1 $ns2 10.0.1.1 128
 		chk_join_nr 0 0 0 +1 +0 1 0 1 "$(pedit_action_pkts)"
+		chk_fail_nr 1 -1 invert
 	fi
 }
 
Signature · 1 line
2.34.1
#5 lore raw

Re: selftests: mptcp: check MP_FAIL response mibs: Tests Results

Hi Geliang,

Thank you for your modifications, that's great!

Our CI did some validations and here is its report:

- KVM Validation: normal:
- Success! ✅:
- Task: https://cirrus-ci.com/task/5689563053031424
- Summary: https://api.cirrus-ci.com/v1/artifact/task/5689563053031424/summary/summary.txt

- KVM Validation: debug:
- Unstable: 2 failed test(s): selftest_diag selftest_mptcp_join 🔴:
- Task: https://cirrus-ci.com/task/5646932549828608
- Summary: https://api.cirrus-ci.com/v1/artifact/task/5646932549828608/summary/summary.txt

Initiator: Patchew Applier
Commits: https://github.com/multipath-tcp/mptcp_net-next/commits/f5598ef0d924

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 (Tessares)

#3 lore raw

Re: [PATCH mptcp-next v3 2/3] mptcp: reset subflow when MP_FAIL doesn't respond

In reply to Geliang Tang
Quoted 5 lines · Geliang Tang

This patch added a new msk->flags bit MPTCP_FAIL_NO_RESPONSE, then reused sk_timer to trigger a check if we have not received a response from the peer after sending MP_FAIL. If the peer doesn't respond properly, reset the subflow.

I don't see the sk_timer use?

Quoted 13 lines

---

net/mptcp/pm.c       |  2 ++
net/mptcp/protocol.c | 21 +++++++++++++++++++++
net/mptcp/protocol.h |  1 +
net/mptcp/subflow.c  |  2 ++
4 files changed, 26 insertions(+)
net/mptcp/pm.c +0 -0
index c4622b07b7f5..c41aca97ba36 100644

net/mptcp/pm.c void mptcp_pm_mp_fail_received(struct sock *sk, u64 fail_seq) · line 288

Quoted 5 lines
net/mptcp/pm.c +2 -0
@@ -284,6 +284,8 @@ void mptcp_pm_mp_fail_received(struct sock *sk, u64 fail_seq)
 			subflow->send_mp_fail = 1;
 			MPTCP_INC_STATS(sock_net(sk), MPTCP_MIB_MPFAILTX);
 			subflow->send_infinite_map = 1;
+		} else {
+			clear_bit(MPTCP_FAIL_NO_RESPONSE, &msk->flags);

Here I think you'd clear the sk_timer (if socket not closed).

Quoted 33 lines
net/mptcp/pm.c +0 -0
 		}
 	}
 }
net/mptcp/protocol.c +18 -0
index 3cb975227d12..6ff7e6eb44ee 100644
@@ -2494,6 +2494,24 @@ static void __mptcp_retrans(struct sock *sk)
 		mptcp_reset_timer(sk);
 }
 
+static void mptcp_mp_fail_no_response(struct mptcp_sock *msk)
+{
+	struct mptcp_subflow_context *subflow;
+
+	mptcp_for_each_subflow(msk, subflow) {
+		if (READ_ONCE(subflow->mp_fail_response_expect)) {
+			struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
+			bool slow;
+
+			pr_debug("MP_FAIL doesn't respond, reset the subflow");
+			slow = lock_sock_fast(ssk);
+			mptcp_subflow_reset(ssk);
+			unlock_sock_fast(ssk, slow);
+			break;
+		}
+	}
+}
+
 static void mptcp_worker(struct work_struct *work)
 {
 	struct mptcp_sock *msk = container_of(work, struct mptcp_sock, work);

net/mptcp/protocol.c static void mptcp_worker(struct work_struct *work) · line 2557

Quoted 6 lines
net/mptcp/protocol.c +3 -0
@@ -2534,6 +2552,9 @@ static void mptcp_worker(struct work_struct *work)
 	if (test_and_clear_bit(MPTCP_WORK_RTX, &msk->flags))
 		__mptcp_retrans(sk);
 
+	if (test_and_clear_bit(MPTCP_FAIL_NO_RESPONSE, &msk->flags))
+		mptcp_mp_fail_no_response(msk);
+

Without the other suggested changes, this would reset the subflow if the worker runs for any reason (and at any time) after MP_FAIL is sent.

Quoted 20 lines

unlock:

	release_sock(sk);
	sock_put(sk);
net/mptcp/protocol.h +1 -0
index 83f0205f0d95..bf58d3c886f5 100644
@@ -116,6 +116,7 @@
 #define MPTCP_WORK_EOF		3
 #define MPTCP_FALLBACK_DONE	4
 #define MPTCP_WORK_CLOSE_SUBFLOW 5
+#define MPTCP_FAIL_NO_RESPONSE	6
 
 /* MPTCP socket release cb flags */
 #define MPTCP_PUSH_PENDING	1
net/mptcp/subflow.c +0 -0
index ca2352ad20d4..5a7aa6e37ca6 100644

net/mptcp/subflow.c static enum mapping_status get_mapping_status(struct sock *ssk, · line 1012

Quoted 4 lines
net/mptcp/subflow.c +1 -0
@@ -1009,6 +1009,7 @@ static enum mapping_status get_mapping_status(struct sock *ssk,
 		pr_debug("infinite mapping received");
 		MPTCP_INC_STATS(sock_net(ssk), MPTCP_MIB_INFINITEMAPRX);
 		subflow->map_data_len = 0;
+		clear_bit(MPTCP_FAIL_NO_RESPONSE, &msk->flags);

Another spot to clear the timer rather than this flag.

net/mptcp/subflow.c static bool subflow_check_data_avail(struct sock *ssk) · line 1223

Quoted 8 lines
net/mptcp/subflow.c +1 -0
 		return MAPPING_INVALID;
 	}
 
@@ -1219,6 +1220,7 @@ static bool subflow_check_data_avail(struct sock *ssk)
 					sk_eat_skb(ssk, skb);
 			} else {
 				WRITE_ONCE(subflow->mp_fail_response_expect, true);
+				set_bit(MPTCP_FAIL_NO_RESPONSE, &msk->flags);

Here, start the timer. In the timer handler, set MPTCP_FAIL_NO_RESPONSE flag (with required locking) if it was an MP_FAIL timeout.

Quoted 8 lines
net/mptcp/subflow.c +0 -0
 			}
 			WRITE_ONCE(subflow->data_avail, MPTCP_SUBFLOW_NODATA);
 			return true;
Signature · 1 line
2.34.1

--
Mat Martineau
Intel