[PATCH] neighbour: Move the netdev_put() out of RCU callback

[PATCH] neighbour: Move the netdev_put() out of RCU callback

[Zqiang]

For builts with CONFIG_NET_DEV_REFCNT_TRACKER=y kernels, netdev_put()
records the caller's stack trace via ref_tracker_free() for debugging
reference leaks. however, the netdev_put() is within pneigh_destroy()
RCU callback, this recorded stack is uninformative, it does not reveal
which code path initiated the deletion.

This commit therefore move netdev_put() from the RCU callback into the
caller's process context, and replace call_rcu() with kfree_rcu().

Fixes: d539d8fbd8fc ("neighbour: Free pneigh_entry after RCU grace period.")
Signed-off-by: Zqiang <qiang.zhang@linux.dev>
---
 net/core/neighbour.c | 14 ++++----------
 1 file changed, 4 insertions(+), 10 deletions(-)

diff --git a/net/core/neighbour.c b/net/core/neighbour.c
index 6ed8eb075146..4263689fc415 100644
--- a/net/core/neighbour.c
+++ b/net/core/neighbour.c
@@ -818,14 +818,6 @@ int pneigh_create(struct neigh_table *tbl,
 	return err;
 }
 
-static void pneigh_destroy(struct rcu_head *rcu)
-{
-	struct pneigh_entry *n = container_of(rcu, struct pneigh_entry, rcu);
-
-	netdev_put(n->dev, &n->dev_tracker);
-	kfree(n);
-}
-
 int pneigh_delete(struct neigh_table *tbl, const void *pkey,
 		  struct net_device *dev)
 {
@@ -849,7 +841,8 @@ int pneigh_delete(struct neigh_table *tbl, const void *pkey,
 			if (tbl->pdestructor)
 				tbl->pdestructor(n);
 
-			call_rcu(&n->rcu, pneigh_destroy);
+			netdev_put(n->dev, &n->dev_tracker);
+			kfree_rcu(n, rcu);
 			return 0;
 		}
 	}
@@ -891,7 +884,8 @@ static void pneigh_ifdown(struct neigh_table *tbl, struct net_device *dev,
 		if (tbl->pdestructor)
 			tbl->pdestructor(n);
 
-		call_rcu(&n->rcu, pneigh_destroy);
+		netdev_put(n->dev, &n->dev_tracker);
+		kfree_rcu(n, rcu);
 	}
 }
 
-- 
2.17.1

Re: [PATCH] neighbour: Move the netdev_put() out of RCU callback

[Kuniyuki Iwashima]

On Fri, Oct 9, 2026 at 2:19 AM Zqiang <qiang.zhang@linux.dev> wrote:
>
> For builts with CONFIG_NET_DEV_REFCNT_TRACKER=y kernels, netdev_put()
> records the caller's stack trace via ref_tracker_free() for debugging
> reference leaks. however, the netdev_put() is within pneigh_destroy()
> RCU callback, this recorded stack is uninformative, it does not reveal
> which code path initiated the deletion.

Have you seen pneigh double-free ?  What motivates this change ?


>
> This commit therefore move netdev_put() from the RCU callback into the
> caller's process context, and replace call_rcu() with kfree_rcu().
>
> Fixes: d539d8fbd8fc ("neighbour: Free pneigh_entry after RCU grace period.")

Please elaborate why your change is safe in the commit message
and remove the tag.


> Signed-off-by: Zqiang <qiang.zhang@linux.dev>

nit: full name

> ---
>  net/core/neighbour.c | 14 ++++----------
>  1 file changed, 4 insertions(+), 10 deletions(-)
>
> diff --git a/net/core/neighbour.c b/net/core/neighbour.c
> index 6ed8eb075146..4263689fc415 100644
> --- a/net/core/neighbour.c
> +++ b/net/core/neighbour.c
> @@ -818,14 +818,6 @@ int pneigh_create(struct neigh_table *tbl,
>         return err;
>  }
>
> -static void pneigh_destroy(struct rcu_head *rcu)
> -{
> -       struct pneigh_entry *n = container_of(rcu, struct pneigh_entry, rcu);
> -
> -       netdev_put(n->dev, &n->dev_tracker);
> -       kfree(n);
> -}
> -
>  int pneigh_delete(struct neigh_table *tbl, const void *pkey,
>                   struct net_device *dev)
>  {
> @@ -849,7 +841,8 @@ int pneigh_delete(struct neigh_table *tbl, const void *pkey,
>                         if (tbl->pdestructor)
>                                 tbl->pdestructor(n);
>
> -                       call_rcu(&n->rcu, pneigh_destroy);
> +                       netdev_put(n->dev, &n->dev_tracker);
> +                       kfree_rcu(n, rcu);
>                         return 0;
>                 }
>         }
> @@ -891,7 +884,8 @@ static void pneigh_ifdown(struct neigh_table *tbl, struct net_device *dev,
>                 if (tbl->pdestructor)
>                         tbl->pdestructor(n);
>
> -               call_rcu(&n->rcu, pneigh_destroy);
> +               netdev_put(n->dev, &n->dev_tracker);
> +               kfree_rcu(n, rcu);
>         }
>  }
>
> --
> 2.17.1
>

Re: [PATCH] neighbour: Move the netdev_put() out of RCU callback

[Zqiang]

> 
> On Fri, Oct 9, 2026 at 2:19 AM Zqiang <qiang.zhang@linux.dev> wrote:
> 
> > 
> > For builts with CONFIG_NET_DEV_REFCNT_TRACKER=y kernels, netdev_put()
> >  records the caller's stack trace via ref_tracker_free() for debugging
> >  reference leaks. however, the netdev_put() is within pneigh_destroy()
> >  RCU callback, this recorded stack is uninformative, it does not reveal
> >  which code path initiated the deletion.
> > 
> Have you seen pneigh double-free ? What motivates this change ?

Firstly, Apologies for delayed reply.



Actually, I didn't see any double-free happening,  just wanted it to
record a meaningful callstack which on release path.  After all, if double-free
really occurs, the callstack of the RCU callback execution path will not be helpful.


> 
> > 
> > This commit therefore move netdev_put() from the RCU callback into the
> >  caller's process context, and replace call_rcu() with kfree_rcu().
> > 
> >  Fixes: d539d8fbd8fc ("neighbour: Free pneigh_entry after RCU grace period.")
> > 
> Please elaborate why your change is safe in the commit message
> and remove the tag.

Does it refer to using kfree_rcu instead of call_rcu ensures safely ?

Thanks
Zqiang

> 
> > 
> > Signed-off-by: Zqiang <qiang.zhang@linux.dev>
> > 
> nit: full name
> 
> > 
> > ---
> >  net/core/neighbour.c | 14 ++++----------
> >  1 file changed, 4 insertions(+), 10 deletions(-)
> > 
> >  diff --git a/net/core/neighbour.c b/net/core/neighbour.c
> >  index 6ed8eb075146..4263689fc415 100644
> >  --- a/net/core/neighbour.c
> >  +++ b/net/core/neighbour.c
> >  @@ -818,14 +818,6 @@ int pneigh_create(struct neigh_table *tbl,
> >  return err;
> >  }
> > 
> >  -static void pneigh_destroy(struct rcu_head *rcu)
> >  -{
> >  - struct pneigh_entry *n = container_of(rcu, struct pneigh_entry, rcu);
> >  -
> >  - netdev_put(n->dev, &n->dev_tracker);
> >  - kfree(n);
> >  -}
> >  -
> >  int pneigh_delete(struct neigh_table *tbl, const void *pkey,
> >  struct net_device *dev)
> >  {
> >  @@ -849,7 +841,8 @@ int pneigh_delete(struct neigh_table *tbl, const void *pkey,
> >  if (tbl->pdestructor)
> >  tbl->pdestructor(n);
> > 
> >  - call_rcu(&n->rcu, pneigh_destroy);
> >  + netdev_put(n->dev, &n->dev_tracker);
> >  + kfree_rcu(n, rcu);
> >  return 0;
> >  }
> >  }
> >  @@ -891,7 +884,8 @@ static void pneigh_ifdown(struct neigh_table *tbl, struct net_device *dev,
> >  if (tbl->pdestructor)
> >  tbl->pdestructor(n);
> > 
> >  - call_rcu(&n->rcu, pneigh_destroy);
> >  + netdev_put(n->dev, &n->dev_tracker);
> >  + kfree_rcu(n, rcu);
> >  }
> >  }
> > 
> >  --
> >  2.17.1
> >
>

Re: [PATCH] neighbour: Move the netdev_put() out of RCU callback

[Kuniyuki Iwashima]

On Sat, 07 Nov 2026 11:47:45 +0000, Zqiang wrote:
> > 
> > On Fri, Oct 9, 2026 at 2:19 AM Zqiang <qiang.zhang@linux.dev> wrote:
> > 
> > > 
> > > For builts with CONFIG_NET_DEV_REFCNT_TRACKER=y kernels, netdev_put()
> > >  records the caller's stack trace via ref_tracker_free() for debugging
> > >  reference leaks. however, the netdev_put() is within pneigh_destroy()
> > >  RCU callback, this recorded stack is uninformative, it does not reveal
> > >  which code path initiated the deletion.
> > > 
> > Have you seen pneigh double-free ? What motivates this change ?
> 
> Firstly, Apologies for delayed reply.
> 
> 
> 
> Actually, I didn't see any double-free happening,  just wanted it to
> record a meaningful callstack which on release path.  After all, if double-free
> really occurs, the callstack of the RCU callback execution path will not be helpful.
> 
> 
> > 
> > > 
> > > This commit therefore move netdev_put() from the RCU callback into the
> > >  caller's process context, and replace call_rcu() with kfree_rcu().
> > > 
> > >  Fixes: d539d8fbd8fc ("neighbour: Free pneigh_entry after RCU grace period.")
> > > 
> > Please elaborate why your change is safe in the commit message
> > and remove the tag.
> 
> Does it refer to using kfree_rcu instead of call_rcu ensures safely ?

I mean why dropping netdev refcnt waiting RCU gp is safe.


> 
> Thanks
> Zqiang
> 
> > 
> > > 
> > > Signed-off-by: Zqiang <qiang.zhang@linux.dev>
> > > 
> > nit: full name
> > 
> > > 
> > > ---
> > >  net/core/neighbour.c | 14 ++++----------
> > >  1 file changed, 4 insertions(+), 10 deletions(-)
> > > 
> > >  diff --git a/net/core/neighbour.c b/net/core/neighbour.c
> > >  index 6ed8eb075146..4263689fc415 100644
> > >  --- a/net/core/neighbour.c
> > >  +++ b/net/core/neighbour.c
> > >  @@ -818,14 +818,6 @@ int pneigh_create(struct neigh_table *tbl,
> > >  return err;
> > >  }
> > > 
> > >  -static void pneigh_destroy(struct rcu_head *rcu)
> > >  -{
> > >  - struct pneigh_entry *n = container_of(rcu, struct pneigh_entry, rcu);
> > >  -
> > >  - netdev_put(n->dev, &n->dev_tracker);
> > >  - kfree(n);
> > >  -}
> > >  -
> > >  int pneigh_delete(struct neigh_table *tbl, const void *pkey,
> > >  struct net_device *dev)
> > >  {
> > >  @@ -849,7 +841,8 @@ int pneigh_delete(struct neigh_table *tbl, const void *pkey,
> > >  if (tbl->pdestructor)
> > >  tbl->pdestructor(n);
> > > 
> > >  - call_rcu(&n->rcu, pneigh_destroy);
> > >  + netdev_put(n->dev, &n->dev_tracker);
> > >  + kfree_rcu(n, rcu);
> > >  return 0;
> > >  }
> > >  }
> > >  @@ -891,7 +884,8 @@ static void pneigh_ifdown(struct neigh_table *tbl, struct net_device *dev,
> > >  if (tbl->pdestructor)
> > >  tbl->pdestructor(n);
> > > 
> > >  - call_rcu(&n->rcu, pneigh_destroy);
> > >  + netdev_put(n->dev, &n->dev_tracker);
> > >  + kfree_rcu(n, rcu);
> > >  }
> > >  }
> > > 
> > >  --
> > >  2.17.1
> > >
> >

Re: [PATCH] neighbour: Move the netdev_put() out of RCU callback

[Kuniyuki Iwashima]

On Sat, 07 Nov 2026 19:34:07 +0000, Kuniyuki Iwashima wrote:
> On Sat, 07 Nov 2026 11:47:45 +0000, Zqiang wrote:
> > > 
> > > On Fri, Oct 9, 2026 at 2:19 AM Zqiang <qiang.zhang@linux.dev> wrote:
> > > 
> > > > 
> > > > For builts with CONFIG_NET_DEV_REFCNT_TRACKER=y kernels, netdev_put()
> > > >  records the caller's stack trace via ref_tracker_free() for debugging
> > > >  reference leaks. however, the netdev_put() is within pneigh_destroy()
> > > >  RCU callback, this recorded stack is uninformative, it does not reveal
> > > >  which code path initiated the deletion.
> > > > 
> > > Have you seen pneigh double-free ? What motivates this change ?
> > 
> > Firstly, Apologies for delayed reply.
> > 
> > 
> > 
> > Actually, I didn't see any double-free happening,  just wanted it to
> > record a meaningful callstack which on release path.  After all, if double-free
> > really occurs, the callstack of the RCU callback execution path will not be helpful.
> > 
> > 
> > > 
> > > > 
> > > > This commit therefore move netdev_put() from the RCU callback into the
> > > >  caller's process context, and replace call_rcu() with kfree_rcu().
> > > > 
> > > >  Fixes: d539d8fbd8fc ("neighbour: Free pneigh_entry after RCU grace period.")
> > > > 
> > > Please elaborate why your change is safe in the commit message
> > > and remove the tag.
> > 
> > Does it refer to using kfree_rcu instead of call_rcu ensures safely ?
> 
> I mean why dropping netdev refcnt waiting RCU gp is safe.

*without waiting


> 
> 
> > 
> > Thanks
> > Zqiang
> > 
> > > 
> > > > 
> > > > Signed-off-by: Zqiang <qiang.zhang@linux.dev>
> > > > 
> > > nit: full name
> > > 
> > > > 
> > > > ---
> > > >  net/core/neighbour.c | 14 ++++----------
> > > >  1 file changed, 4 insertions(+), 10 deletions(-)
> > > > 
> > > >  diff --git a/net/core/neighbour.c b/net/core/neighbour.c
> > > >  index 6ed8eb075146..4263689fc415 100644
> > > >  --- a/net/core/neighbour.c
> > > >  +++ b/net/core/neighbour.c
> > > >  @@ -818,14 +818,6 @@ int pneigh_create(struct neigh_table *tbl,
> > > >  return err;
> > > >  }
> > > > 
> > > >  -static void pneigh_destroy(struct rcu_head *rcu)
> > > >  -{
> > > >  - struct pneigh_entry *n = container_of(rcu, struct pneigh_entry, rcu);
> > > >  -
> > > >  - netdev_put(n->dev, &n->dev_tracker);
> > > >  - kfree(n);
> > > >  -}
> > > >  -
> > > >  int pneigh_delete(struct neigh_table *tbl, const void *pkey,
> > > >  struct net_device *dev)
> > > >  {
> > > >  @@ -849,7 +841,8 @@ int pneigh_delete(struct neigh_table *tbl, const void *pkey,
> > > >  if (tbl->pdestructor)
> > > >  tbl->pdestructor(n);
> > > > 
> > > >  - call_rcu(&n->rcu, pneigh_destroy);
> > > >  + netdev_put(n->dev, &n->dev_tracker);
> > > >  + kfree_rcu(n, rcu);
> > > >  return 0;
> > > >  }
> > > >  }
> > > >  @@ -891,7 +884,8 @@ static void pneigh_ifdown(struct neigh_table *tbl, struct net_device *dev,
> > > >  if (tbl->pdestructor)
> > > >  tbl->pdestructor(n);
> > > > 
> > > >  - call_rcu(&n->rcu, pneigh_destroy);
> > > >  + netdev_put(n->dev, &n->dev_tracker);
> > > >  + kfree_rcu(n, rcu);
> > > >  }
> > > >  }
> > > > 
> > > >  --
> > > >  2.17.1
> > > >
> > >