Skip to contentSkip to conversation

Squash-to "mptcp: pm: init and release mptcp_pm_ops"

PATCH Superseded Gang Yan 3 participants 4 messages Jul 1 – Jul 2

Subsystems (2)

v2 of 3 Superseded Latest is v3 →

lore
Share & tools
lore

Gang Yan

author
#0 lore raw

PATCH v2 1/1 Squash-to "mptcp: pm: init and release mptcp_pm_ops"

From: Gang Yan <[email protected]>

This patch removes the need to handle a NULL pm_ops in the fallback path, and moves the rcu_read_lock\unlock into mptcp_pm_ops_init.

Following sashiko's comments, it also moves the pr_debug() before bpf_module_put() in mptcp_pm_ops_release.

---
 net/mptcp/pm.c | 19 ++++++++++---------
 1 file changed, 10 insertions(+), 9 deletions(-)
net/mptcp/pm.c +10 -9
index 9dc7b41fb562..5fe4276cf895 100644
@@ -1137,12 +1137,14 @@ void mptcp_pm_worker(struct mptcp_sock *msk)
 	spin_unlock_bh(&msk->pm.lock);
 }
 
-static void mptcp_pm_ops_init(struct mptcp_sock *msk,
-			      struct mptcp_pm_ops *pm_ops)
+static void mptcp_pm_ops_init(struct mptcp_sock *msk, const char *pm_name)
 {
+	struct mptcp_pm_ops *pm_ops;
+
+	rcu_read_lock();
+	pm_ops = mptcp_pm_find(pm_name);
 	if (!pm_ops || !bpf_try_module_get(pm_ops, pm_ops->owner)) {
-		pr_warn_once("pm %s fails, fallback to default pm",
-			     pm_ops->name);
+		pr_warn_once("pm %s fails, fallback to default pm", pm_name);
 		pm_ops = &mptcp_pm_kernel;
 	}
 
@@ -1151,6 +1153,7 @@ static void mptcp_pm_ops_init(struct mptcp_sock *msk,
 		msk->pm.ops->init(msk);
 
 	pr_debug("pm %s initialized\n", pm_ops->name);
+	rcu_read_unlock();
 }
 
 static void mptcp_pm_ops_release(struct mptcp_sock *msk)
@@ -1161,9 +1164,9 @@ static void mptcp_pm_ops_release(struct mptcp_sock *msk)
 	if (pm_ops->release)
 		pm_ops->release(msk);
 
-	bpf_module_put(pm_ops, pm_ops->owner);
-
 	pr_debug("pm %s released\n", pm_ops->name);
+
+	bpf_module_put(pm_ops, pm_ops->owner);
 }
 
 void mptcp_pm_destroy(struct mptcp_sock *msk)
@@ -1184,9 +1187,7 @@ void mptcp_pm_data_reset(struct mptcp_sock *msk)
 	pm->rm_list_rx.nr = 0;
 	WRITE_ONCE(pm->pm_type, pm_type);
 
-	rcu_read_lock();
-	mptcp_pm_ops_init(msk, mptcp_pm_find(pm_name));
-	rcu_read_unlock();
+	mptcp_pm_ops_init(msk, pm_name);
 }
 
 void mptcp_pm_data_init(struct mptcp_sock *msk)
Signature · 1 line
2.43.0
#1 lore raw

Hi Gang,

Thank you for your modifications, that's great!

Our CI did some validations and here is its report:

- KVM Validation: normal (except selftest_mptcp_join): Unstable: 1 failed test(s): packetdrill_fastopen ⚠️
- KVM Validation: normal (only selftest_mptcp_join): Success! ✅
- KVM Validation: debug (except selftest_mptcp_join): Success! ✅
- KVM Validation: debug (only selftest_mptcp_join): 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/28503396809

Initiator: Patchew Applier
Commits: https://github.com/multipath-tcp/mptcp_net-next/commits/dac4cb248a7d
Patchwork: https://patchwork.kernel.org/project/mptcp/list/?series=1119502

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)

reviewer
#2 lore raw
In reply to Gang Yan

Hi Gang,

Thanks for this v2.

Quoted 41 lines · Gang Yan

From: Gang Yan <[email protected]>

This patch removes the need to handle a NULL pm_ops in the fallback path, and moves the rcu_read_lock\unlock into mptcp_pm_ops_init.

Following sashiko's comments, it also moves the pr_debug() before bpf_module_put() in mptcp_pm_ops_release.

---
 net/mptcp/pm.c | 19 ++++++++++---------

 1 file changed, 10 insertions(+), 9 deletions(-)
net/mptcp/pm.c +0 -0
index 9dc7b41fb562..5fe4276cf895 100644
@@ -1137,12 +1137,14 @@ void mptcp_pm_worker(struct mptcp_sock *msk)
  	spin_unlock_bh(&msk->pm.lock);
  }

net/mptcp/pm.c +1 -2
-static void mptcp_pm_ops_init(struct mptcp_sock *msk,
-			      struct mptcp_pm_ops *pm_ops)
+static void mptcp_pm_ops_init(struct mptcp_sock *msk, const char

*pm_name)
 {

net/mptcp/pm.c +4 -0
+	struct mptcp_pm_ops *pm_ops;
+
+	rcu_read_lock();
+	pm_ops = mptcp_pm_find(pm_name);

  if (!pm_ops || !bpf_try_module_get(pm_ops, pm_ops->owner)) {

net/mptcp/pm.c +1 -2
-		pr_warn_once("pm %s fails, fallback to default pm",
-			     pm_ops->name);
+		pr_warn_once("pm %s fails, fallback to default pm",

pm_name);
  pm_ops = &mptcp_pm_kernel;
  }

net/mptcp/pm.c +0 -0
@@ -1151,6 +1153,7 @@ static void mptcp_pm_ops_init(struct mptcp_sock
 *msk,
  		msk->pm.ops->init(msk);
Quoted 2 lines


  pr_debug("pm %s initialized\n", pm_ops->name);

nit: Here we can also print "pm_name" to keep it consistent with pr_warn_once().

net/mptcp/pm.c static void mptcp_pm_ops_init(struct mptcp_sock *msk, · line 1156

Quoted 1 line
net/mptcp/pm.c +1 -0
+	rcu_read_unlock();

Then this unlock line can be moved ahead of the previous pr_debug() line.

Quoted 21 lines

 }

 static void mptcp_pm_ops_release(struct mptcp_sock *msk)

net/mptcp/pm.c +0 -0
@@ -1161,9 +1164,9 @@ static void mptcp_pm_ops_release(struct
 mptcp_sock *msk)
  	if (pm_ops->release)
  		pm_ops->release(msk);

net/mptcp/pm.c +0 -2
-	bpf_module_put(pm_ops, pm_ops->owner);
-

  pr_debug("pm %s released\n", pm_ops->name);

net/mptcp/pm.c +2 -0
+
+	bpf_module_put(pm_ops, pm_ops->owner);

 }

 void mptcp_pm_destroy(struct mptcp_sock *msk)

net/mptcp/pm.c +0 -0
@@ -1184,9 +1187,7 @@ void mptcp_pm_data_reset(struct mptcp_sock
 *msk)
  	pm->rm_list_rx.nr = 0;
  	WRITE_ONCE(pm->pm_type, pm_type);

net/mptcp/pm.c void mptcp_pm_data_reset(struct mptcp_sock *msk) · line 1189

Quoted 3 lines
net/mptcp/pm.c +0 -3
-	rcu_read_lock();
-	mptcp_pm_ops_init(msk, mptcp_pm_find(pm_name));
-	rcu_read_unlock();

Or we leave rcu_read_lock() and rcu_read_unlock() where they are, without moving them into mptcp_pm_ops_init().

Thanks,
-Geliang

Quoted 4 lines
net/mptcp/pm.c +1 -0
+	mptcp_pm_ops_init(msk, pm_name);

 }

 void mptcp_pm_data_init(struct mptcp_sock *msk)

#3 lore raw

July 2, 2026 at 10:15 AM, "Geliang Tang" <[email protected] mailto:[email protected]?to=%22Geliang%20Tang%22%20%3Cgeliang%40kernel.org%3E > wrote:

Hi Geliang,

Thanks for your review.

Quoted 52 lines

Hi Gang,

Thanks for this v2.

Quoted 36 lines · Gang Yan

From: Gang Yan <[email protected]>

This patch removes the need to handle a NULL pm_ops in the fallback path, and moves the rcu_read_lock\unlock into mptcp_pm_ops_init.

Following sashiko's comments, it also moves the pr_debug() before bpf_module_put() in mptcp_pm_ops_release.

---

  net/mptcp/pm.c | 19 ++++++++++---------
  1 file changed, 10 insertions(+), 9 deletions(-)

diff --git a/net/mptcp/pm.c b/net/mptcp/pm.c
index 9dc7b41fb562..5fe4276cf895 100644
--- a/net/mptcp/pm.c
+++ b/net/mptcp/pm.c
@@ -1137,12 +1137,14 @@ void mptcp_pm_worker(struct mptcp_sock *msk) spin_unlock_bh(&msk->pm.lock);
 }

-static void mptcp_pm_ops_init(struct mptcp_sock *msk,
-       struct mptcp_pm_ops *pm_ops)
+static void mptcp_pm_ops_init(struct mptcp_sock *msk, const char *pm_name)
 {
+ struct mptcp_pm_ops *pm_ops;
+
+ rcu_read_lock();
+ pm_ops = mptcp_pm_find(pm_name);
  if (!pm_ops || !bpf_try_module_get(pm_ops, pm_ops->owner)) {
- pr_warn_once("pm %s fails, fallback to default pm",
-      pm_ops->name);
+ pr_warn_once("pm %s fails, fallback to default pm", pm_name);

Quoted 9 lines

  pm_ops = &mptcp_pm_kernel;
  }

@@ -1151,6 +1153,7 @@ static void mptcp_pm_ops_init(struct mptcp_sock *msk,
  msk->pm.ops->init(msk);

  pr_debug("pm %s initialized\n", pm_ops->name);

Quoted 2 lines

nit: Here we can also print "pm_name" to keep it consistent with pr_warn_once().

Here is the print of the name of initialized pm, for example, if we fallback to kernel_PM it will print like:

'''
pm bpf_pm fails, fallback to default pm
pm kernel initialized
'''

So I think it should use 'pm_ops->name' here.
WDYT?

Quoted 34 lines
Quoted 4 lines

+ rcu_read_unlock();

Then this unlock line can be moved ahead of the previous pr_debug() line.

Quoted 17 lines

}

 static void mptcp_pm_ops_release(struct mptcp_sock *msk) @@ -1161,9 +1164,9 @@ static void mptcp_pm_ops_release(struct mptcp_sock *msk)
  if (pm_ops->release)
  pm_ops->release(msk);

- bpf_module_put(pm_ops, pm_ops->owner);
-
  pr_debug("pm %s released\n", pm_ops->name);
+
+ bpf_module_put(pm_ops, pm_ops->owner);
 }

 void mptcp_pm_destroy(struct mptcp_sock *msk)

Quoted 9 lines

@@ -1184,9 +1187,7 @@ void mptcp_pm_data_reset(struct mptcp_sock *msk)
  pm->rm_list_rx.nr = 0;
  WRITE_ONCE(pm->pm_type, pm_type);

- rcu_read_lock();
- mptcp_pm_ops_init(msk, mptcp_pm_find(pm_name));
- rcu_read_unlock();

Quoted 3 lines

Or we leave rcu_read_lock() and rcu_read_unlock() where they are, without moving them into mptcp_pm_ops_init().

LGTM, I'll apply the reset suggestions when sending v3.

Quoted 4 lines

Thanks,
-Geliang

Quoted 1 line

Thanks
Gang

Quoted 6 lines
Quoted 5 lines

+ mptcp_pm_ops_init(msk, pm_name);
 }

 void mptcp_pm_data_init(struct mptcp_sock *msk)