[PATCH mptcp-next v6 0/5] BPF path manager, part 6

[PATCH mptcp-next v6 0/5] BPF path manager, part 6

[Geliang Tang]

From: Geliang Tang <tanggeliang@kylinos.cn>

v6:
 - squash accept_new_subflow patches into one.
 - change "pm->subflows == subflows_max - 1" to
	pm->subflows + 1 == subflows_max.
 - not to call accept_new_subflow under the PM lock.
 - add mptcp_pm_accept_subflow helpers.
 - drop READ_ONCE in mptcp_pm_worker.
 - clear all the pm status flags once in mptcp_pm_worker.

v5:
 - add comment "call from the subflow/msk context" for mptcp_sched_ops.
 - add new helper mptcp_pm_accept_new_subflow.
 - add "bool allow" parameter for mptcp_pm_accept_new_subflow, and drop
   .allow_new_subflow interface.
 - use a copy of pm->status in mptcp_pm_worker.
 - rename mptcp_pm_create_subflow_or_signal_addr with "__" prefix.
 - drop "!update_subflows" in mptcp_pm_subflow_check_next.
 - add_addr_received/rm_addr_received interfaces will be added in the
   next series.

v4:
 - address Matt's comments in v3. 
 - update pm locks in mptcp_pm_worker.
 - move the lock inside mptcp_pm_create_subflow_or_signal_addr.
 - move the lock inside mptcp_pm_nl_add_addr_received.
 - invoke add_addr_received interface from mptcp_pm_worker.
 - invoke rm_addr_received interface from mptcp_pm_rm_addr_or_subflow.
 - simply call mptcp_pm_close_subflow() in mptcp_pm_subflow_check_next.
 - https://patchwork.kernel.org/project/mptcp/cover/cover.1742804266.git.tanggeliang@kylinos.cn/

v3:
 - merge 'bugfixes for "BPF path manager, part 6, v2"' into this set.
 - https://patchwork.kernel.org/project/mptcp/cover/cover.1742521397.git.tanggeliang@kylinos.cn/

v2:
 - address Matt's comments in v1. 
 - add add_addr_received and rm_addr_received interfaces.
 - drop subflow_check_next interface.
 - add a "required" or "optional" comment for a group of interfaces in
   struct mptcp_pm_ops.

v1:
- https://patchwork.kernel.org/project/mptcp/cover/cover.1741685260.git.tanggeliang@kylinos.cn/

New interfaces for struct mptcp_pm_ops.

Geliang Tang (5):
  mptcp: pm: call pm worker handler without pm lock
  mptcp: pm: add accept_new_subflow() interface
  mptcp: pm: add established() interface
  mptcp: pm: add subflow_established() interface
  mptcp: pm: drop is_userspace in subflow_check_next

 include/net/mptcp.h      |   5 ++
 net/mptcp/pm.c           | 100 +++++++++++++++++++--------------------
 net/mptcp/pm_kernel.c    |  47 ++++++++++++------
 net/mptcp/pm_userspace.c |   9 +++-
 net/mptcp/protocol.h     |  27 ++++++++++-
 net/mptcp/subflow.c      |   6 +--
 6 files changed, 122 insertions(+), 72 deletions(-)

-- 
2.43.0

[PATCH mptcp-next v6 1/5] mptcp: pm: call pm worker handler without pm lock

[Geliang Tang]

From: Geliang Tang <tanggeliang@kylinos.cn>

Later functions that cannot hold the mptcp pm lock will be called from
the PM worker, so this patch modifies the order of holding the lock
at the beginning of this function and releasing the lock at the end.

The new order is to obtain a copy of pm->status and clear the flags of
pm->status while holding the mptcp pm lock, then read the copy after
releasing the lock.

For each PM status flag, call each handling function without locks.

Finally, hold the lock before calling __mptcp_pm_kernel_worker() and
release it afterwards.

Rename mptcp_pm_addr_send_ack() with "__" prefix to indicate that the
mptcp pm lock is not held.

Hold the mptcp pm lock inside mptcp_pm_rm_addr_recv().

Signed-off-by: Geliang Tang <tanggeliang@kylinos.cn>
---
 net/mptcp/pm.c           | 34 ++++++++++++++++++++++------------
 net/mptcp/pm_kernel.c    |  4 ++--
 net/mptcp/pm_userspace.c |  2 +-
 net/mptcp/protocol.h     |  2 +-
 4 files changed, 26 insertions(+), 16 deletions(-)

diff --git a/net/mptcp/pm.c b/net/mptcp/pm.c
index be1e27ee393e..4ed6a2ccb112 100644
--- a/net/mptcp/pm.c
+++ b/net/mptcp/pm.c
@@ -211,7 +211,7 @@ void mptcp_pm_send_ack(struct mptcp_sock *msk,
 	spin_lock_bh(&msk->pm.lock);
 }
 
-void mptcp_pm_addr_send_ack(struct mptcp_sock *msk)
+void __mptcp_pm_addr_send_ack(struct mptcp_sock *msk)
 {
 	struct mptcp_subflow_context *subflow, *alt = NULL;
 
@@ -238,6 +238,13 @@ void mptcp_pm_addr_send_ack(struct mptcp_sock *msk)
 		mptcp_pm_send_ack(msk, alt, false, false);
 }
 
+static void mptcp_pm_addr_send_ack(struct mptcp_sock *msk)
+{
+	spin_lock_bh(&msk->pm.lock);
+	__mptcp_pm_addr_send_ack(msk);
+	spin_unlock_bh(&msk->pm.lock);
+}
+
 int mptcp_pm_mp_prio_send_ack(struct mptcp_sock *msk,
 			      struct mptcp_addr_info *addr,
 			      struct mptcp_addr_info *rem,
@@ -437,7 +444,7 @@ int mptcp_pm_remove_addr(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_
 	msk->pm.rm_list_tx = *rm_list;
 	rm_addr |= BIT(MPTCP_RM_ADDR_SIGNAL);
 	WRITE_ONCE(msk->pm.addr_signal, rm_addr);
-	mptcp_pm_addr_send_ack(msk);
+	__mptcp_pm_addr_send_ack(msk);
 	return 0;
 }
 
@@ -715,7 +722,9 @@ static void mptcp_pm_rm_addr_or_subflow(struct mptcp_sock *msk,
 
 static void mptcp_pm_rm_addr_recv(struct mptcp_sock *msk)
 {
+	spin_lock_bh(&msk->pm.lock);
 	mptcp_pm_rm_addr_or_subflow(msk, &msk->pm.rm_list_rx, MPTCP_MIB_RMADDR);
+	spin_unlock_bh(&msk->pm.lock);
 }
 
 void mptcp_pm_rm_subflow(struct mptcp_sock *msk,
@@ -945,6 +954,8 @@ void mptcp_pm_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk)
 
 void mptcp_pm_worker(struct mptcp_sock *msk)
 {
+	u8 status, mask = BIT(MPTCP_PM_ADD_ADDR_SEND_ACK) |
+			  BIT(MPTCP_PM_RM_ADDR_RECEIVED);
 	struct mptcp_pm_data *pm = &msk->pm;
 
 	msk_owned_by_me(msk);
@@ -952,20 +963,19 @@ void mptcp_pm_worker(struct mptcp_sock *msk)
 	if (!(pm->status & MPTCP_PM_WORK_MASK))
 		return;
 
-	spin_lock_bh(&msk->pm.lock);
+	spin_lock_bh(&pm->lock);
+	status = pm->status;
+	pm->status &= ~mask;
+	spin_unlock_bh(&pm->lock);
 
-	pr_debug("msk=%p status=%x\n", msk, pm->status);
-	if (pm->status & BIT(MPTCP_PM_ADD_ADDR_SEND_ACK)) {
-		pm->status &= ~BIT(MPTCP_PM_ADD_ADDR_SEND_ACK);
+	pr_debug("msk=%p status=%x\n", msk, status);
+	if (status & BIT(MPTCP_PM_ADD_ADDR_SEND_ACK))
 		mptcp_pm_addr_send_ack(msk);
-	}
-	if (pm->status & BIT(MPTCP_PM_RM_ADDR_RECEIVED)) {
-		pm->status &= ~BIT(MPTCP_PM_RM_ADDR_RECEIVED);
+	if (status & BIT(MPTCP_PM_RM_ADDR_RECEIVED))
 		mptcp_pm_rm_addr_recv(msk);
-	}
+	spin_lock_bh(&pm->lock);
 	__mptcp_pm_kernel_worker(msk);
-
-	spin_unlock_bh(&msk->pm.lock);
+	spin_unlock_bh(&pm->lock);
 }
 
 static void mptcp_pm_ops_init(struct mptcp_sock *msk,
diff --git a/net/mptcp/pm_kernel.c b/net/mptcp/pm_kernel.c
index 7ec81d5195d4..8935df61aa50 100644
--- a/net/mptcp/pm_kernel.c
+++ b/net/mptcp/pm_kernel.c
@@ -326,7 +326,7 @@ static void mptcp_pm_create_subflow_or_signal_addr(struct mptcp_sock *msk)
 			local.addr.id = 0;
 
 		mptcp_pm_announce_addr(msk, &local.addr, false);
-		mptcp_pm_addr_send_ack(msk);
+		__mptcp_pm_addr_send_ack(msk);
 
 		if (local.flags & MPTCP_PM_ADDR_FLAG_SUBFLOW)
 			signal_and_subflow = true;
@@ -460,7 +460,7 @@ static void mptcp_pm_nl_add_addr_received(struct mptcp_sock *msk)
 
 	remote = msk->pm.remote;
 	mptcp_pm_announce_addr(msk, &remote, true);
-	mptcp_pm_addr_send_ack(msk);
+	__mptcp_pm_addr_send_ack(msk);
 
 	if (lookup_subflow_by_daddr(&msk->conn_list, &remote))
 		return;
diff --git a/net/mptcp/pm_userspace.c b/net/mptcp/pm_userspace.c
index 7fc19b844384..d9b0a93e42f6 100644
--- a/net/mptcp/pm_userspace.c
+++ b/net/mptcp/pm_userspace.c
@@ -231,7 +231,7 @@ int mptcp_pm_nl_announce_doit(struct sk_buff *skb, struct genl_info *info)
 	if (mptcp_pm_alloc_anno_list(msk, &addr_val.addr)) {
 		msk->pm.add_addr_signaled++;
 		mptcp_pm_announce_addr(msk, &addr_val.addr, false);
-		mptcp_pm_addr_send_ack(msk);
+		__mptcp_pm_addr_send_ack(msk);
 	}
 
 	spin_unlock_bh(&msk->pm.lock);
diff --git a/net/mptcp/protocol.h b/net/mptcp/protocol.h
index 64aa091cb685..3d51898a2781 100644
--- a/net/mptcp/protocol.h
+++ b/net/mptcp/protocol.h
@@ -1022,7 +1022,7 @@ void mptcp_pm_add_addr_send_ack(struct mptcp_sock *msk);
 void mptcp_pm_send_ack(struct mptcp_sock *msk,
 		       struct mptcp_subflow_context *subflow,
 		       bool prio, bool backup);
-void mptcp_pm_addr_send_ack(struct mptcp_sock *msk);
+void __mptcp_pm_addr_send_ack(struct mptcp_sock *msk);
 void mptcp_pm_nl_rm_addr(struct mptcp_sock *msk, u8 rm_id);
 void mptcp_pm_rm_subflow(struct mptcp_sock *msk,
 			 const struct mptcp_rm_list *rm_list);
-- 
2.43.0

Re: [PATCH mptcp-next v6 1/5] mptcp: pm: call pm worker handler without pm lock

[Geliang Tang]

Hi Matt,

On Fri, 2025-03-28 at 12:01 +0800, Geliang Tang wrote:
> From: Geliang Tang <tanggeliang@kylinos.cn>
> 
> Later functions that cannot hold the mptcp pm lock will be called
> from
> the PM worker, so this patch modifies the order of holding the lock
> at the beginning of this function and releasing the lock at the end.
> 
> The new order is to obtain a copy of pm->status and clear the flags
> of
> pm->status while holding the mptcp pm lock, then read the copy after
> releasing the lock.
> 
> For each PM status flag, call each handling function without locks.
> 
> Finally, hold the lock before calling __mptcp_pm_kernel_worker() and
> release it afterwards.
> 
> Rename mptcp_pm_addr_send_ack() with "__" prefix to indicate that the
> mptcp pm lock is not held.
> 
> Hold the mptcp pm lock inside mptcp_pm_rm_addr_recv().
> 
> Signed-off-by: Geliang Tang <tanggeliang@kylinos.cn>
> ---
>  net/mptcp/pm.c           | 34 ++++++++++++++++++++++------------
>  net/mptcp/pm_kernel.c    |  4 ++--
>  net/mptcp/pm_userspace.c |  2 +-
>  net/mptcp/protocol.h     |  2 +-
>  4 files changed, 26 insertions(+), 16 deletions(-)
> 
> diff --git a/net/mptcp/pm.c b/net/mptcp/pm.c
> index be1e27ee393e..4ed6a2ccb112 100644
> --- a/net/mptcp/pm.c
> +++ b/net/mptcp/pm.c
> @@ -211,7 +211,7 @@ void mptcp_pm_send_ack(struct mptcp_sock *msk,
>  	spin_lock_bh(&msk->pm.lock);
>  }
>  
> -void mptcp_pm_addr_send_ack(struct mptcp_sock *msk)
> +void __mptcp_pm_addr_send_ack(struct mptcp_sock *msk)
>  {
>  	struct mptcp_subflow_context *subflow, *alt = NULL;
>  
> @@ -238,6 +238,13 @@ void mptcp_pm_addr_send_ack(struct mptcp_sock
> *msk)
>  		mptcp_pm_send_ack(msk, alt, false, false);
>  }
>  
> +static void mptcp_pm_addr_send_ack(struct mptcp_sock *msk)
> +{
> +	spin_lock_bh(&msk->pm.lock);
> +	__mptcp_pm_addr_send_ack(msk);
> +	spin_unlock_bh(&msk->pm.lock);
> +}
> +
>  int mptcp_pm_mp_prio_send_ack(struct mptcp_sock *msk,
>  			      struct mptcp_addr_info *addr,
>  			      struct mptcp_addr_info *rem,
> @@ -437,7 +444,7 @@ int mptcp_pm_remove_addr(struct mptcp_sock *msk,
> const struct mptcp_rm_list *rm_
>  	msk->pm.rm_list_tx = *rm_list;
>  	rm_addr |= BIT(MPTCP_RM_ADDR_SIGNAL);
>  	WRITE_ONCE(msk->pm.addr_signal, rm_addr);
> -	mptcp_pm_addr_send_ack(msk);
> +	__mptcp_pm_addr_send_ack(msk);
>  	return 0;
>  }
>  
> @@ -715,7 +722,9 @@ static void mptcp_pm_rm_addr_or_subflow(struct
> mptcp_sock *msk,
>  
>  static void mptcp_pm_rm_addr_recv(struct mptcp_sock *msk)
>  {
> +	spin_lock_bh(&msk->pm.lock);
>  	mptcp_pm_rm_addr_or_subflow(msk, &msk->pm.rm_list_rx,
> MPTCP_MIB_RMADDR);
> +	spin_unlock_bh(&msk->pm.lock);
>  }
>  
>  void mptcp_pm_rm_subflow(struct mptcp_sock *msk,
> @@ -945,6 +954,8 @@ void mptcp_pm_subflow_chk_stale(const struct
> mptcp_sock *msk, struct sock *ssk)
>  
>  void mptcp_pm_worker(struct mptcp_sock *msk)
>  {
> +	u8 status, mask = BIT(MPTCP_PM_ADD_ADDR_SEND_ACK) |
> +			  BIT(MPTCP_PM_RM_ADDR_RECEIVED);
>  	struct mptcp_pm_data *pm = &msk->pm;
>  
>  	msk_owned_by_me(msk);
> @@ -952,20 +963,19 @@ void mptcp_pm_worker(struct mptcp_sock *msk)
>  	if (!(pm->status & MPTCP_PM_WORK_MASK))
>  		return;
>  
> -	spin_lock_bh(&msk->pm.lock);
> +	spin_lock_bh(&pm->lock);
> +	status = pm->status;
> +	pm->status &= ~mask;

I found that there is no need to define a 'mask' variable, we can
directly use MPTCP_PM_WORK_MASK here, I will update this in v7.

Thanks,
-Geliang

> +	spin_unlock_bh(&pm->lock);
>  
> -	pr_debug("msk=%p status=%x\n", msk, pm->status);
> -	if (pm->status & BIT(MPTCP_PM_ADD_ADDR_SEND_ACK)) {
> -		pm->status &= ~BIT(MPTCP_PM_ADD_ADDR_SEND_ACK);
> +	pr_debug("msk=%p status=%x\n", msk, status);
> +	if (status & BIT(MPTCP_PM_ADD_ADDR_SEND_ACK))
>  		mptcp_pm_addr_send_ack(msk);
> -	}
> -	if (pm->status & BIT(MPTCP_PM_RM_ADDR_RECEIVED)) {
> -		pm->status &= ~BIT(MPTCP_PM_RM_ADDR_RECEIVED);
> +	if (status & BIT(MPTCP_PM_RM_ADDR_RECEIVED))
>  		mptcp_pm_rm_addr_recv(msk);
> -	}
> +	spin_lock_bh(&pm->lock);
>  	__mptcp_pm_kernel_worker(msk);
> -
> -	spin_unlock_bh(&msk->pm.lock);
> +	spin_unlock_bh(&pm->lock);
>  }
>  
>  static void mptcp_pm_ops_init(struct mptcp_sock *msk,
> diff --git a/net/mptcp/pm_kernel.c b/net/mptcp/pm_kernel.c
> index 7ec81d5195d4..8935df61aa50 100644
> --- a/net/mptcp/pm_kernel.c
> +++ b/net/mptcp/pm_kernel.c
> @@ -326,7 +326,7 @@ static void
> mptcp_pm_create_subflow_or_signal_addr(struct mptcp_sock *msk)
>  			local.addr.id = 0;
>  
>  		mptcp_pm_announce_addr(msk, &local.addr, false);
> -		mptcp_pm_addr_send_ack(msk);
> +		__mptcp_pm_addr_send_ack(msk);
>  
>  		if (local.flags & MPTCP_PM_ADDR_FLAG_SUBFLOW)
>  			signal_and_subflow = true;
> @@ -460,7 +460,7 @@ static void mptcp_pm_nl_add_addr_received(struct
> mptcp_sock *msk)
>  
>  	remote = msk->pm.remote;
>  	mptcp_pm_announce_addr(msk, &remote, true);
> -	mptcp_pm_addr_send_ack(msk);
> +	__mptcp_pm_addr_send_ack(msk);
>  
>  	if (lookup_subflow_by_daddr(&msk->conn_list, &remote))
>  		return;
> diff --git a/net/mptcp/pm_userspace.c b/net/mptcp/pm_userspace.c
> index 7fc19b844384..d9b0a93e42f6 100644
> --- a/net/mptcp/pm_userspace.c
> +++ b/net/mptcp/pm_userspace.c
> @@ -231,7 +231,7 @@ int mptcp_pm_nl_announce_doit(struct sk_buff
> *skb, struct genl_info *info)
>  	if (mptcp_pm_alloc_anno_list(msk, &addr_val.addr)) {
>  		msk->pm.add_addr_signaled++;
>  		mptcp_pm_announce_addr(msk, &addr_val.addr, false);
> -		mptcp_pm_addr_send_ack(msk);
> +		__mptcp_pm_addr_send_ack(msk);
>  	}
>  
>  	spin_unlock_bh(&msk->pm.lock);
> diff --git a/net/mptcp/protocol.h b/net/mptcp/protocol.h
> index 64aa091cb685..3d51898a2781 100644
> --- a/net/mptcp/protocol.h
> +++ b/net/mptcp/protocol.h
> @@ -1022,7 +1022,7 @@ void mptcp_pm_add_addr_send_ack(struct
> mptcp_sock *msk);
>  void mptcp_pm_send_ack(struct mptcp_sock *msk,
>  		       struct mptcp_subflow_context *subflow,
>  		       bool prio, bool backup);
> -void mptcp_pm_addr_send_ack(struct mptcp_sock *msk);
> +void __mptcp_pm_addr_send_ack(struct mptcp_sock *msk);
>  void mptcp_pm_nl_rm_addr(struct mptcp_sock *msk, u8 rm_id);
>  void mptcp_pm_rm_subflow(struct mptcp_sock *msk,
>  			 const struct mptcp_rm_list *rm_list);

[PATCH mptcp-next v6 2/5] mptcp: pm: add accept_new_subflow() interface

[Geliang Tang]

From: Geliang Tang <tanggeliang@kylinos.cn>

The helper mptcp_pm_is_userspace() is used to distinguish userspace PM
operations from in-kernel PM in mptcp_can_accept_new_subflow(). It seems
reasonable to add a mandatory .accept_new_subflow interface for struct
mptcp_pm_ops.

The boolean type 'allow' parameter is used to control whether to allow new
subflow unconditionally.

A new helper mptcp_pm_accept_new_subflow() is added in pm.c to ensure that
pm->ops is only used in pm.c.

This interface of the in-kernel PM is implemented as a wrapper of
READ_ONCE(pm->accept_subflow), and this interface of the userspace PM is
implemented as a wrapper of mptcp_userspace_pm_active().

To reduce the dependency on mptcp_pm_is_userspace() to identify userspace
PM from in-kernel PM by using mptcp_pm_accept_new_subflow() helper in
mptcp_pm_allow_new_subflow().

Signed-off-by: Geliang Tang <tanggeliang@kylinos.cn>
---
 include/net/mptcp.h      |  1 +
 net/mptcp/pm.c           | 38 +++++++++++++-------------------------
 net/mptcp/pm_kernel.c    | 16 ++++++++++++++++
 net/mptcp/pm_userspace.c |  7 +++++++
 net/mptcp/protocol.h     | 25 +++++++++++++++++++++++++
 net/mptcp/subflow.c      |  6 ++----
 6 files changed, 64 insertions(+), 29 deletions(-)

diff --git a/include/net/mptcp.h b/include/net/mptcp.h
index 4cf59e83c1c5..549b8075cc86 100644
--- a/include/net/mptcp.h
+++ b/include/net/mptcp.h
@@ -123,6 +123,7 @@ struct mptcp_pm_ops {
 			    struct mptcp_pm_addr_entry *skc);
 	bool (*get_priority)(struct mptcp_sock *msk,
 			     struct mptcp_addr_info *skc);
+	bool (*accept_new_subflow)(struct mptcp_sock *msk, bool allow);
 
 	char			name[MPTCP_PM_NAME_MAX];
 	struct module		*owner;
diff --git a/net/mptcp/pm.c b/net/mptcp/pm.c
index 4ed6a2ccb112..a96f3495d220 100644
--- a/net/mptcp/pm.c
+++ b/net/mptcp/pm.c
@@ -460,38 +460,25 @@ void mptcp_pm_new_connection(struct mptcp_sock *msk, const struct sock *ssk, int
 	mptcp_event(MPTCP_EVENT_CREATED, msk, ssk, GFP_ATOMIC);
 }
 
+bool mptcp_pm_accept_new_subflow(struct mptcp_sock *msk, bool allow)
+{
+	return msk->pm.ops->accept_new_subflow(msk, allow);
+}
+
 bool mptcp_pm_allow_new_subflow(struct mptcp_sock *msk)
 {
 	struct mptcp_pm_data *pm = &msk->pm;
-	unsigned int subflows_max;
 	int ret = 0;
 
-	if (mptcp_pm_is_userspace(msk)) {
-		if (mptcp_userspace_pm_active(msk)) {
-			spin_lock_bh(&pm->lock);
-			pm->subflows++;
-			spin_unlock_bh(&pm->lock);
-			return true;
-		}
-		return false;
-	}
-
-	subflows_max = mptcp_pm_get_subflows_max(msk);
-
-	pr_debug("msk=%p subflows=%d max=%d allow=%d\n", msk, pm->subflows,
-		 subflows_max, READ_ONCE(pm->accept_subflow));
-
-	/* try to avoid acquiring the lock below */
-	if (!READ_ONCE(pm->accept_subflow))
+	if (!mptcp_pm_accept_new_subflow(msk, true))
 		return false;
 
-	spin_lock_bh(&pm->lock);
-	if (READ_ONCE(pm->accept_subflow)) {
-		ret = pm->subflows < subflows_max;
-		if (ret && ++pm->subflows == subflows_max)
-			WRITE_ONCE(pm->accept_subflow, false);
+	ret = mptcp_pm_accept_new_subflow(msk, false);
+	if (ret) {
+		spin_lock_bh(&pm->lock);
+		pm->subflows++;
+		spin_unlock_bh(&pm->lock);
 	}
-	spin_unlock_bh(&pm->lock);
 
 	return ret;
 }
@@ -1060,7 +1047,8 @@ struct mptcp_pm_ops *mptcp_pm_find(const char *name)
 
 int mptcp_pm_validate(struct mptcp_pm_ops *pm_ops)
 {
-	if (!pm_ops->get_local_id || !pm_ops->get_priority) {
+	if (!pm_ops->get_local_id || !pm_ops->get_priority ||
+	    !pm_ops->accept_new_subflow) {
 		pr_err("%s does not implement required ops\n", pm_ops->name);
 		return -EINVAL;
 	}
diff --git a/net/mptcp/pm_kernel.c b/net/mptcp/pm_kernel.c
index 8935df61aa50..eaf73c4cedd3 100644
--- a/net/mptcp/pm_kernel.c
+++ b/net/mptcp/pm_kernel.c
@@ -1399,6 +1399,21 @@ static struct pernet_operations mptcp_pm_pernet_ops = {
 	.size = sizeof(struct pm_nl_pernet),
 };
 
+static bool mptcp_pm_kernel_accept_new_subflow(struct mptcp_sock *msk,
+					       bool allow)
+{
+	bool ret = false;
+
+	if (READ_ONCE(msk->pm.accept_subflow)) {
+		if (allow)
+			return true;
+
+		ret = mptcp_pm_accept_subflow(msk);
+	}
+
+	return ret;
+}
+
 static void mptcp_pm_kernel_init(struct mptcp_sock *msk)
 {
 	bool subflows_allowed = !!mptcp_pm_get_subflows_max(msk);
@@ -1422,6 +1437,7 @@ static void mptcp_pm_kernel_init(struct mptcp_sock *msk)
 struct mptcp_pm_ops mptcp_pm_kernel = {
 	.get_local_id		= mptcp_pm_kernel_get_local_id,
 	.get_priority		= mptcp_pm_kernel_get_priority,
+	.accept_new_subflow	= mptcp_pm_kernel_accept_new_subflow,
 	.init			= mptcp_pm_kernel_init,
 	.name			= "kernel",
 	.owner			= THIS_MODULE,
diff --git a/net/mptcp/pm_userspace.c b/net/mptcp/pm_userspace.c
index d9b0a93e42f6..31da3b9f4cc8 100644
--- a/net/mptcp/pm_userspace.c
+++ b/net/mptcp/pm_userspace.c
@@ -683,6 +683,12 @@ int mptcp_userspace_pm_get_addr(u8 id, struct mptcp_pm_addr_entry *addr,
 	return ret;
 }
 
+static bool mptcp_pm_userspace_accept_new_subflow(struct mptcp_sock *msk,
+						  bool allow)
+{
+	return mptcp_userspace_pm_active(msk);
+}
+
 static void mptcp_pm_userspace_release(struct mptcp_sock *msk)
 {
 	mptcp_userspace_pm_free_local_addr_list(msk);
@@ -691,6 +697,7 @@ static void mptcp_pm_userspace_release(struct mptcp_sock *msk)
 static struct mptcp_pm_ops mptcp_pm_userspace = {
 	.get_local_id		= mptcp_pm_userspace_get_local_id,
 	.get_priority		= mptcp_pm_userspace_get_priority,
+	.accept_new_subflow	= mptcp_pm_userspace_accept_new_subflow,
 	.release		= mptcp_pm_userspace_release,
 	.name			= "userspace",
 	.owner			= THIS_MODULE,
diff --git a/net/mptcp/protocol.h b/net/mptcp/protocol.h
index 3d51898a2781..134b46b53014 100644
--- a/net/mptcp/protocol.h
+++ b/net/mptcp/protocol.h
@@ -1008,6 +1008,7 @@ bool mptcp_pm_addr_families_match(const struct sock *sk,
 void mptcp_pm_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk);
 void mptcp_pm_new_connection(struct mptcp_sock *msk, const struct sock *ssk, int server_side);
 void mptcp_pm_fully_established(struct mptcp_sock *msk, const struct sock *ssk);
+bool mptcp_pm_accept_new_subflow(struct mptcp_sock *msk, bool allow);
 bool mptcp_pm_allow_new_subflow(struct mptcp_sock *msk);
 void mptcp_pm_connection_closed(struct mptcp_sock *msk);
 void mptcp_pm_subflow_established(struct mptcp_sock *msk);
@@ -1184,6 +1185,30 @@ static inline void mptcp_pm_close_subflow(struct mptcp_sock *msk)
 	spin_unlock_bh(&msk->pm.lock);
 }
 
+static inline bool __mptcp_pm_accept_subflow(struct mptcp_sock *msk)
+{
+	unsigned int subflows_max = mptcp_pm_get_subflows_max(msk);
+	struct mptcp_pm_data *pm = &msk->pm;
+	bool ret;
+
+	ret = pm->subflows < subflows_max;
+	if (ret && pm->subflows + 1 == subflows_max)
+		WRITE_ONCE(pm->accept_subflow, false);
+
+	return ret;
+}
+
+static inline bool mptcp_pm_accept_subflow(struct mptcp_sock *msk)
+{
+	bool ret;
+
+	spin_lock_bh(&msk->pm.lock);
+	ret = __mptcp_pm_accept_subflow(msk);
+	spin_unlock_bh(&msk->pm.lock);
+
+	return ret;
+}
+
 void mptcp_sockopt_sync_locked(struct mptcp_sock *msk, struct sock *ssk);
 
 static inline struct mptcp_ext *mptcp_get_ext(const struct sk_buff *skb)
diff --git a/net/mptcp/subflow.c b/net/mptcp/subflow.c
index 409bd415ef1d..b13a679bdf05 100644
--- a/net/mptcp/subflow.c
+++ b/net/mptcp/subflow.c
@@ -58,12 +58,10 @@ static void subflow_generate_hmac(u64 key1, u64 key2, u32 nonce1, u32 nonce2,
 	mptcp_crypto_hmac_sha(key1, key2, msg, 8, hmac);
 }
 
-static bool mptcp_can_accept_new_subflow(const struct mptcp_sock *msk)
+static bool mptcp_can_accept_new_subflow(struct mptcp_sock *msk)
 {
 	return mptcp_is_fully_established((void *)msk) &&
-		((mptcp_pm_is_userspace(msk) &&
-		  mptcp_userspace_pm_active(msk)) ||
-		 READ_ONCE(msk->pm.accept_subflow));
+		mptcp_pm_accept_new_subflow(msk, true);
 }
 
 /* validate received token and create truncated hmac and nonce for SYN-ACK */
-- 
2.43.0

[PATCH mptcp-next v6 3/5] mptcp: pm: add established() interface

[Geliang Tang]

From: Geliang Tang <tanggeliang@kylinos.cn>

This patch adds a .established interface for struct mptcp_pm_ops, and
calls pm->ops->established in from mptcp_pm_worker(). Then get rid of
the corresponding code from __mptcp_pm_kernel_worker().

Since mptcp_pm_addr_send_ack() is a sleepable kfunc, which is invoked
by mptcp_pm_create_subflow_or_signal_addr(), .established() interface
of BPF PM should be invoked by __bpf_prog_enter_sleepable(), which
can't be invoked under a lock. This patch unlocks the pm lock before
invoking this interface in mptcp_pm_worker(), while holding this lock
in mptcp_pm_kernel_established().

Add a helper to call mptcp_pm_create_subflow_or_signal_addr() while
holding mptcp pm lock, and rename the original one with "__" prefix.

Signed-off-by: Geliang Tang <tanggeliang@kylinos.cn>
---
 include/net/mptcp.h   |  3 +++
 net/mptcp/pm.c        |  8 ++++++--
 net/mptcp/pm_kernel.c | 22 +++++++++++++---------
 3 files changed, 22 insertions(+), 11 deletions(-)

diff --git a/include/net/mptcp.h b/include/net/mptcp.h
index 549b8075cc86..97a237c9bfe0 100644
--- a/include/net/mptcp.h
+++ b/include/net/mptcp.h
@@ -125,6 +125,9 @@ struct mptcp_pm_ops {
 			     struct mptcp_addr_info *skc);
 	bool (*accept_new_subflow)(struct mptcp_sock *msk, bool allow);
 
+	/* optional, call from the msk context */
+	void (*established)(struct mptcp_sock *msk);
+
 	char			name[MPTCP_PM_NAME_MAX];
 	struct module		*owner;
 	struct list_head	list;
diff --git a/net/mptcp/pm.c b/net/mptcp/pm.c
index a96f3495d220..e7b3600239e0 100644
--- a/net/mptcp/pm.c
+++ b/net/mptcp/pm.c
@@ -513,7 +513,8 @@ void mptcp_pm_fully_established(struct mptcp_sock *msk, const struct sock *ssk)
 	 * be sure to serve this event only once.
 	 */
 	if (READ_ONCE(pm->work_pending) &&
-	    !(pm->status & BIT(MPTCP_PM_ALREADY_ESTABLISHED)))
+	    !(pm->status & BIT(MPTCP_PM_ALREADY_ESTABLISHED)) &&
+	    pm->ops->established)
 		mptcp_pm_schedule_work(msk, MPTCP_PM_ESTABLISHED);
 
 	if ((pm->status & BIT(MPTCP_PM_ALREADY_ESTABLISHED)) == 0)
@@ -942,7 +943,8 @@ void mptcp_pm_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk)
 void mptcp_pm_worker(struct mptcp_sock *msk)
 {
 	u8 status, mask = BIT(MPTCP_PM_ADD_ADDR_SEND_ACK) |
-			  BIT(MPTCP_PM_RM_ADDR_RECEIVED);
+			  BIT(MPTCP_PM_RM_ADDR_RECEIVED) |
+			  BIT(MPTCP_PM_ESTABLISHED);
 	struct mptcp_pm_data *pm = &msk->pm;
 
 	msk_owned_by_me(msk);
@@ -960,6 +962,8 @@ void mptcp_pm_worker(struct mptcp_sock *msk)
 		mptcp_pm_addr_send_ack(msk);
 	if (status & BIT(MPTCP_PM_RM_ADDR_RECEIVED))
 		mptcp_pm_rm_addr_recv(msk);
+	if (status & BIT(MPTCP_PM_ESTABLISHED))
+		pm->ops->established(msk);
 	spin_lock_bh(&pm->lock);
 	__mptcp_pm_kernel_worker(msk);
 	spin_unlock_bh(&pm->lock);
diff --git a/net/mptcp/pm_kernel.c b/net/mptcp/pm_kernel.c
index eaf73c4cedd3..31794193442d 100644
--- a/net/mptcp/pm_kernel.c
+++ b/net/mptcp/pm_kernel.c
@@ -253,7 +253,7 @@ __lookup_addr(struct pm_nl_pernet *pernet, const struct mptcp_addr_info *info)
 	return NULL;
 }
 
-static void mptcp_pm_create_subflow_or_signal_addr(struct mptcp_sock *msk)
+static void __mptcp_pm_create_subflow_or_signal_addr(struct mptcp_sock *msk)
 {
 	struct sock *sk = (struct sock *)msk;
 	unsigned int add_addr_signal_max;
@@ -367,14 +367,21 @@ static void mptcp_pm_create_subflow_or_signal_addr(struct mptcp_sock *msk)
 	mptcp_pm_nl_check_work_pending(msk);
 }
 
-static void mptcp_pm_nl_fully_established(struct mptcp_sock *msk)
+static void mptcp_pm_create_subflow_or_signal_addr(struct mptcp_sock *msk)
+{
+	spin_lock_bh(&msk->pm.lock);
+	__mptcp_pm_create_subflow_or_signal_addr(msk);
+	spin_unlock_bh(&msk->pm.lock);
+}
+
+static void mptcp_pm_kernel_established(struct mptcp_sock *msk)
 {
 	mptcp_pm_create_subflow_or_signal_addr(msk);
 }
 
 static void mptcp_pm_nl_subflow_established(struct mptcp_sock *msk)
 {
-	mptcp_pm_create_subflow_or_signal_addr(msk);
+	__mptcp_pm_create_subflow_or_signal_addr(msk);
 }
 
 /* Fill all the local addresses into the array addrs[],
@@ -758,7 +765,7 @@ static int mptcp_nl_add_subflow_or_signal_addr(struct net *net,
 		spin_lock_bh(&msk->pm.lock);
 		if (mptcp_addresses_equal(addr, &mpc_addr, addr->port))
 			msk->mpc_endpoint_id = addr->id;
-		mptcp_pm_create_subflow_or_signal_addr(msk);
+		__mptcp_pm_create_subflow_or_signal_addr(msk);
 		spin_unlock_bh(&msk->pm.lock);
 		release_sock(sk);
 
@@ -1243,7 +1250,7 @@ static void mptcp_pm_nl_fullmesh(struct mptcp_sock *msk,
 	spin_lock_bh(&msk->pm.lock);
 	mptcp_pm_rm_subflow(msk, &list);
 	__mark_subflow_endp_available(msk, list.ids[0]);
-	mptcp_pm_create_subflow_or_signal_addr(msk);
+	__mptcp_pm_create_subflow_or_signal_addr(msk);
 	spin_unlock_bh(&msk->pm.lock);
 }
 
@@ -1348,10 +1355,6 @@ void __mptcp_pm_kernel_worker(struct mptcp_sock *msk)
 		pm->status &= ~BIT(MPTCP_PM_ADD_ADDR_RECEIVED);
 		mptcp_pm_nl_add_addr_received(msk);
 	}
-	if (pm->status & BIT(MPTCP_PM_ESTABLISHED)) {
-		pm->status &= ~BIT(MPTCP_PM_ESTABLISHED);
-		mptcp_pm_nl_fully_established(msk);
-	}
 	if (pm->status & BIT(MPTCP_PM_SUBFLOW_ESTABLISHED)) {
 		pm->status &= ~BIT(MPTCP_PM_SUBFLOW_ESTABLISHED);
 		mptcp_pm_nl_subflow_established(msk);
@@ -1438,6 +1441,7 @@ struct mptcp_pm_ops mptcp_pm_kernel = {
 	.get_local_id		= mptcp_pm_kernel_get_local_id,
 	.get_priority		= mptcp_pm_kernel_get_priority,
 	.accept_new_subflow	= mptcp_pm_kernel_accept_new_subflow,
+	.established		= mptcp_pm_kernel_established,
 	.init			= mptcp_pm_kernel_init,
 	.name			= "kernel",
 	.owner			= THIS_MODULE,
-- 
2.43.0

[PATCH mptcp-next v6 4/5] mptcp: pm: add subflow_established() interface

[Geliang Tang]

From: Geliang Tang <tanggeliang@kylinos.cn>

This patch adds a .subflow_established interface for struct mptcp_pm_ops,
and calls pm->ops->subflow_established in from mptcp_pm_worker(). Then
get rid of the corresponding code from __mptcp_pm_kernel_worker().

Since mptcp_pm_addr_send_ack() is a sleepable kfunc, which is invoked
by mptcp_pm_create_subflow_or_signal_addr(), .subflow_established()
interface of BPF PM should be invoked by __bpf_prog_enter_sleepable(),
which can't be invoked under a lock. This patch unlocks the pm lock
before invoking this interface in mptcp_pm_worker(), while holding this
lock in mptcp_pm_kernel_subflow_established().

Signed-off-by: Geliang Tang <tanggeliang@kylinos.cn>
---
 include/net/mptcp.h   |  1 +
 net/mptcp/pm.c        | 12 ++++++++----
 net/mptcp/pm_kernel.c |  9 +++------
 3 files changed, 12 insertions(+), 10 deletions(-)

diff --git a/include/net/mptcp.h b/include/net/mptcp.h
index 97a237c9bfe0..1eae68c9e662 100644
--- a/include/net/mptcp.h
+++ b/include/net/mptcp.h
@@ -127,6 +127,7 @@ struct mptcp_pm_ops {
 
 	/* optional, call from the msk context */
 	void (*established)(struct mptcp_sock *msk);
+	void (*subflow_established)(struct mptcp_sock *msk);
 
 	char			name[MPTCP_PM_NAME_MAX];
 	struct module		*owner;
diff --git a/net/mptcp/pm.c b/net/mptcp/pm.c
index e7b3600239e0..673313b9b25f 100644
--- a/net/mptcp/pm.c
+++ b/net/mptcp/pm.c
@@ -541,7 +541,7 @@ void mptcp_pm_subflow_established(struct mptcp_sock *msk)
 
 	pr_debug("msk=%p\n", msk);
 
-	if (!READ_ONCE(pm->work_pending))
+	if (!READ_ONCE(pm->work_pending) || !pm->ops->subflow_established)
 		return;
 
 	spin_lock_bh(&pm->lock);
@@ -568,7 +568,8 @@ void mptcp_pm_subflow_check_next(struct mptcp_sock *msk,
 		return;
 	}
 
-	if (!READ_ONCE(pm->work_pending) && !update_subflows)
+	if (!pm->ops->subflow_established ||
+	    (!READ_ONCE(pm->work_pending) && !update_subflows))
 		return;
 
 	spin_lock_bh(&pm->lock);
@@ -626,7 +627,7 @@ void mptcp_pm_add_addr_echoed(struct mptcp_sock *msk,
 
 	pr_debug("msk=%p\n", msk);
 
-	if (!READ_ONCE(pm->work_pending))
+	if (!READ_ONCE(pm->work_pending) || !pm->ops->subflow_established)
 		return;
 
 	spin_lock_bh(&pm->lock);
@@ -944,7 +945,8 @@ void mptcp_pm_worker(struct mptcp_sock *msk)
 {
 	u8 status, mask = BIT(MPTCP_PM_ADD_ADDR_SEND_ACK) |
 			  BIT(MPTCP_PM_RM_ADDR_RECEIVED) |
-			  BIT(MPTCP_PM_ESTABLISHED);
+			  BIT(MPTCP_PM_ESTABLISHED) |
+			  BIT(MPTCP_PM_SUBFLOW_ESTABLISHED);
 	struct mptcp_pm_data *pm = &msk->pm;
 
 	msk_owned_by_me(msk);
@@ -964,6 +966,8 @@ void mptcp_pm_worker(struct mptcp_sock *msk)
 		mptcp_pm_rm_addr_recv(msk);
 	if (status & BIT(MPTCP_PM_ESTABLISHED))
 		pm->ops->established(msk);
+	if (status & BIT(MPTCP_PM_SUBFLOW_ESTABLISHED))
+		pm->ops->subflow_established(msk);
 	spin_lock_bh(&pm->lock);
 	__mptcp_pm_kernel_worker(msk);
 	spin_unlock_bh(&pm->lock);
diff --git a/net/mptcp/pm_kernel.c b/net/mptcp/pm_kernel.c
index 31794193442d..000f4f173ad5 100644
--- a/net/mptcp/pm_kernel.c
+++ b/net/mptcp/pm_kernel.c
@@ -379,9 +379,9 @@ static void mptcp_pm_kernel_established(struct mptcp_sock *msk)
 	mptcp_pm_create_subflow_or_signal_addr(msk);
 }
 
-static void mptcp_pm_nl_subflow_established(struct mptcp_sock *msk)
+static void mptcp_pm_kernel_subflow_established(struct mptcp_sock *msk)
 {
-	__mptcp_pm_create_subflow_or_signal_addr(msk);
+	mptcp_pm_create_subflow_or_signal_addr(msk);
 }
 
 /* Fill all the local addresses into the array addrs[],
@@ -1355,10 +1355,6 @@ void __mptcp_pm_kernel_worker(struct mptcp_sock *msk)
 		pm->status &= ~BIT(MPTCP_PM_ADD_ADDR_RECEIVED);
 		mptcp_pm_nl_add_addr_received(msk);
 	}
-	if (pm->status & BIT(MPTCP_PM_SUBFLOW_ESTABLISHED)) {
-		pm->status &= ~BIT(MPTCP_PM_SUBFLOW_ESTABLISHED);
-		mptcp_pm_nl_subflow_established(msk);
-	}
 }
 
 static int __net_init pm_nl_init_net(struct net *net)
@@ -1442,6 +1438,7 @@ struct mptcp_pm_ops mptcp_pm_kernel = {
 	.get_priority		= mptcp_pm_kernel_get_priority,
 	.accept_new_subflow	= mptcp_pm_kernel_accept_new_subflow,
 	.established		= mptcp_pm_kernel_established,
+	.subflow_established	= mptcp_pm_kernel_subflow_established,
 	.init			= mptcp_pm_kernel_init,
 	.name			= "kernel",
 	.owner			= THIS_MODULE,
-- 
2.43.0

[PATCH mptcp-next v6 5/5] mptcp: pm: drop is_userspace in subflow_check_next

[Geliang Tang]

From: Geliang Tang <tanggeliang@kylinos.cn>

This patch moves mptcp_pm_close_subflow() forward to let it be used by both
the userspace PM and the in-kernel PM in mptcp_pm_subflow_check_next().
Then mptcp_pm_is_userspace() here can be dropped.

Signed-off-by: Geliang Tang <tanggeliang@kylinos.cn>
---
 net/mptcp/pm.c | 14 +++-----------
 1 file changed, 3 insertions(+), 11 deletions(-)

diff --git a/net/mptcp/pm.c b/net/mptcp/pm.c
index 673313b9b25f..d576b03a64e1 100644
--- a/net/mptcp/pm.c
+++ b/net/mptcp/pm.c
@@ -559,22 +559,14 @@ void mptcp_pm_subflow_check_next(struct mptcp_sock *msk,
 	bool update_subflows;
 
 	update_subflows = subflow->request_join || subflow->mp_join;
-	if (mptcp_pm_is_userspace(msk)) {
-		if (update_subflows) {
-			spin_lock_bh(&pm->lock);
-			pm->subflows--;
-			spin_unlock_bh(&pm->lock);
-		}
-		return;
-	}
+	if (update_subflows)
+		mptcp_pm_close_subflow(msk);
 
 	if (!pm->ops->subflow_established ||
-	    (!READ_ONCE(pm->work_pending) && !update_subflows))
+	    !READ_ONCE(pm->work_pending))
 		return;
 
 	spin_lock_bh(&pm->lock);
-	if (update_subflows)
-		__mptcp_pm_close_subflow(msk);
 
 	/* Even if this subflow is not really established, tell the PM to try
 	 * to pick the next ones, if possible.
-- 
2.43.0

Re: [PATCH mptcp-next v6 0/5] BPF path manager, part 6

[MPTCP CI]

Hi Geliang,

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: Success! ✅
- KVM Validation: btf-normal (only bpftest_all): Success! ✅
- KVM Validation: btf-debug (only bpftest_all): Success! ✅
- Task: https://github.com/multipath-tcp/mptcp_net-next/actions/runs/14121889205

Initiator: Patchew Applier
Commits: https://github.com/multipath-tcp/mptcp_net-next/commits/c19b176f03f3
Patchwork: https://patchwork.kernel.org/project/mptcp/list/?series=947940


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 mptcp-next v6 0/5] BPF path manager, part 6

[Matthieu Baerts]

Hi Geliang,

On 28/03/2025 05:01, Geliang Tang wrote:
> From: Geliang Tang <tanggeliang@kylinos.cn>
> 
> v6:
>  - squash accept_new_subflow patches into one.
>  - change "pm->subflows == subflows_max - 1" to
> 	pm->subflows + 1 == subflows_max.
>  - not to call accept_new_subflow under the PM lock.
>  - add mptcp_pm_accept_subflow helpers.
>  - drop READ_ONCE in mptcp_pm_worker.
>  - clear all the pm status flags once in mptcp_pm_worker.

Thank you for the new version.

FYI, I'm still thinking about how to handle PM lock should be handled. I
will come back to you later. I might suggest a new version with some
modifications if that's OK, that might be easier.

Cheers,
Matt
-- 
Sponsored by the NGI0 Core fund.

Re: [PATCH mptcp-next v6 0/5] BPF path manager, part 6

[Geliang Tang]

Hi Matt,

On Tue, 2025-04-08 at 20:13 +0200, Matthieu Baerts wrote:
> Hi Geliang,
> 
> On 28/03/2025 05:01, Geliang Tang wrote:
> > From: Geliang Tang <tanggeliang@kylinos.cn>
> > 
> > v6:
> >  - squash accept_new_subflow patches into one.
> >  - change "pm->subflows == subflows_max - 1" to
> > 	pm->subflows + 1 == subflows_max.
> >  - not to call accept_new_subflow under the PM lock.
> >  - add mptcp_pm_accept_subflow helpers.
> >  - drop READ_ONCE in mptcp_pm_worker.
> >  - clear all the pm status flags once in mptcp_pm_worker.
> 
> Thank you for the new version.
> 
> FYI, I'm still thinking about how to handle PM lock should be
> handled. I
> will come back to you later. I might suggest a new version with some
> modifications if that's OK, that might be easier.

Thanks for helping me think about this part around the PM lock. Look
forward to your feedback.

-Geliang

> 
> Cheers,
> Matt

Re: [PATCH mptcp-next v6 0/5] BPF path manager, part 6

[Geliang Tang]

Hi Matt,

On Tue, 2025-04-15 at 11:16 +0800, Geliang Tang wrote:
> Hi Matt,
> 
> On Tue, 2025-04-08 at 20:13 +0200, Matthieu Baerts wrote:
> > Hi Geliang,
> > 
> > On 28/03/2025 05:01, Geliang Tang wrote:
> > > From: Geliang Tang <tanggeliang@kylinos.cn>
> > > 
> > > v6:
> > >  - squash accept_new_subflow patches into one.
> > >  - change "pm->subflows == subflows_max - 1" to
> > > 	pm->subflows + 1 == subflows_max.
> > >  - not to call accept_new_subflow under the PM lock.
> > >  - add mptcp_pm_accept_subflow helpers.
> > >  - drop READ_ONCE in mptcp_pm_worker.
> > >  - clear all the pm status flags once in mptcp_pm_worker.
> > 
> > Thank you for the new version.
> > 
> > FYI, I'm still thinking about how to handle PM lock should be
> > handled. I
> > will come back to you later. I might suggest a new version with
> > some
> > modifications if that's OK, that might be easier.
> 
> Thanks for helping me think about this part around the PM lock. Look
> forward to your feedback.

If you have time to take over this part around the PM lock, that would
be even better and I would be very grateful.

Thanks,
-Geliang

> 
> -Geliang
> 
> > 
> > Cheers,
> > Matt
> 
> 

Re: [PATCH mptcp-next v6 0/5] BPF path manager, part 6

[Matthieu Baerts]

Hi Geliang,

On 16/04/2025 09:29, Geliang Tang wrote:
> Hi Matt,
> 
> On Tue, 2025-04-15 at 11:16 +0800, Geliang Tang wrote:
>> Hi Matt,
>>
>> On Tue, 2025-04-08 at 20:13 +0200, Matthieu Baerts wrote:
>>> Hi Geliang,
>>>
>>> On 28/03/2025 05:01, Geliang Tang wrote:
>>>> From: Geliang Tang <tanggeliang@kylinos.cn>
>>>>
>>>> v6:
>>>>  - squash accept_new_subflow patches into one.
>>>>  - change "pm->subflows == subflows_max - 1" to
>>>> 	pm->subflows + 1 == subflows_max.
>>>>  - not to call accept_new_subflow under the PM lock.
>>>>  - add mptcp_pm_accept_subflow helpers.
>>>>  - drop READ_ONCE in mptcp_pm_worker.
>>>>  - clear all the pm status flags once in mptcp_pm_worker.
>>>
>>> Thank you for the new version.
>>>
>>> FYI, I'm still thinking about how to handle PM lock should be
>>> handled. I
>>> will come back to you later. I might suggest a new version with
>>> some
>>> modifications if that's OK, that might be easier.
>>
>> Thanks for helping me think about this part around the PM lock. Look
>> forward to your feedback.
> 
> If you have time to take over this part around the PM lock, that would
> be even better and I would be very grateful.

Sure, no problem, I will have a look. But sadly, for the moment, it is
hard for me to dedicate a long time on some tasks. It might take a bit
of time for me to look at it, but I think that's fine, we are not in a rush.

Cheers,
Matt
-- 
Sponsored by the NGI0 Core fund.