2023-08-04 09:30:29

by Yang Yang

Subject: [PATCH] net: add net device refcount tracker to pneigh_queue

From: xu xin <[email protected]>

Add a static net device refcount tracker to neighbour.c to track
dev's refcnt when pneigh enqueuing and dequeuing.

Signed-off-by: xu xin <[email protected]>
Reviewed-by: Yang Yang <[email protected]>
Cc: Kuang Mingfu <[email protected]>
---
net/core/neighbour.c | 7 ++++---
1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/net/core/neighbour.c b/net/core/neighbour.c
index ddd0f32de20e..56e87130fa95 100644
--- a/net/core/neighbour.c
+++ b/net/core/neighbour.c
@@ -61,6 +61,7 @@ static int pneigh_ifdown_and_unlock(struct neigh_table *tbl,
static const struct seq_operations neigh_stat_seq_ops;
#endif

+static netdevice_tracker pneigh_queue_dev_tracker;
/*
Neighbour hash table buckets are protected with rwlock tbl->lock.

@@ -364,7 +365,7 @@ static void pneigh_queue_purge(struct sk_buff_head *list, struct net *net,
spin_unlock_irqrestore(&list->lock, flags);

while ((skb = __skb_dequeue(&tmp))) {
- dev_put(skb->dev);
+ netdev_put(skb->dev, &pneigh_queue_dev_tracker);
kfree_skb(skb);
}
}
@@ -1633,7 +1634,7 @@ static void neigh_proxy_process(struct timer_list *t)
kfree_skb(skb);
}

- dev_put(dev);
+ netdev_put(skb->dev, &pneigh_queue_dev_tracker);
} else if (!sched_next || tdif < sched_next)
sched_next = tdif;
}
@@ -1673,7 +1674,7 @@ void pneigh_enqueue(struct neigh_table *tbl, struct neigh_parms *p,
sched_next = tbl->proxy_timer.expires;
}
skb_dst_drop(skb);
- dev_hold(skb->dev);
+ netdev_hold(skb->dev, &pneigh_queue_dev_tracker, GFP_KERNEL);
__skb_queue_tail(&tbl->proxy_queue, skb);
p->qlen++;
mod_timer(&tbl->proxy_timer, sched_next);
--
2.15.2

2023-08-04 12:45:29

by Eric Dumazet

Subject: Re: [PATCH] net: add net device refcount tracker to pneigh_queue

On Fri, Aug 4, 2023 at 10:58 AM <[email protected]> wrote:
>
> From: xu xin <[email protected]>
>
> Add a static net device refcount tracker to neighbour.c to track
> dev's refcnt when pneigh enqueuing and dequeuing.
>
> Signed-off-by: xu xin <[email protected]>
> Reviewed-by: Yang Yang <[email protected]>
> Cc: Kuang Mingfu <[email protected]>
> ---


Can you explain how this patch was tested ?

Thanks.

2023-08-05 07:39:58

by xu

Subject: [PATCH] net: add net device refcount tracker to pneigh_queue

>> Add a static net device refcount tracker to neighbour.c to track
>> dev's refcnt when pneigh enqueuing and dequeuing.
>>
>> Signed-off-by: xu xin <[email protected]>
>> Reviewed-by: Yang Yang <[email protected]>
>> Cc: Kuang Mingfu <[email protected]>
>> ---
>
>
>Can you explain how this patch was tested ?
>
>Thanks.

Sorry, I misunderstand the usage of netdev_hold and its netdevice_tracker.

I thought this tracker was simple enough, so I just added an additional netdev_hold
in the same postion and test if there is a warning "unregister_netdevice: waiting for ...".

But as you said, we need a separate netdevice_tracker per netdev_hold(), and specially `per dev`?

Actually, I'm still a little confused.

For example, here, can I move this netdevice_tracker into struct sk_buff ?

>diff --git a/net/core/neighbour.c b/net/core/neighbour.c
>index ddd0f32de20e..56e87130fa95 100644
>--- a/net/core/neighbour.c
>+++ b/net/core/neighbour.c
>@@ -61,6 +61,7 @@ static int pneigh_ifdown_and_unlock(struct neigh_table *tbl,
> static const struct seq_operations neigh_stat_seq_ops;
> #endif
>
>+static netdevice_tracker pneigh_queue_dev_tracker;
> /*
> Neighbour hash table buckets are protected with rwlock tbl->lock.
>
>@@ -364,7 +365,7 @@ static void pneigh_queue_purge(struct sk_buff_head *list, struct net *net,
> spin_unlock_irqrestore(&list->lock, flags);
>
> while ((skb = __skb_dequeue(&tmp))) {
>- dev_put(skb->dev);
>+ netdev_put(skb->dev, &pneigh_queue_dev_tracker);
> kfree_skb(skb);
> }
> }
>@@ -1633,7 +1634,7 @@ static void neigh_proxy_process(struct timer_list *t)
> kfree_skb(skb);
> }
>
>- dev_put(dev);
>+ netdev_put(skb->dev, &pneigh_queue_dev_tracker);
> } else if (!sched_next || tdif < sched_next)
> sched_next = tdif;
> }
>@@ -1673,7 +1674,7 @@ void pneigh_enqueue(struct neigh_table *tbl, struct neigh_parms *p,
> sched_next = tbl->proxy_timer.expires;
> }
> skb_dst_drop(skb);
>- dev_hold(skb->dev);
>+ netdev_hold(skb->dev, &pneigh_queue_dev_tracker, GFP_KERNEL);
> __skb_queue_tail(&tbl->proxy_queue, skb);
> p->qlen++;
> mod_timer(&tbl->proxy_timer, sched_next);