Re: [PATCH -next] net: neigh: use dev_kfree_skb_irq instead of kfree_skb()

[Yang Yingliang]

Hi,

On 2022/8/18 17:00, Denis V. Lunev wrote:
> On 18.08.2022 06:37, 'Yang Yingliang' via den wrote:
> > It is not allowed to call kfree_skb() from hardware interrupt
> > context or with interrupts being disabled. So replace kfree_skb()
> > with dev_kfree_skb_irq() under spin_lock_irqsave().
> >
> > Fixes: 66ba215cb513 ("neigh: fix possible DoS due to net iface 
> > start/stop loop")
> > Signed-off-by: Yang Yingliang <yangyingliang@xxxxxxxxxx>
> > ---
> >   net/core/neighbour.c | 2 +-
> >   1 file changed, 1 insertion(+), 1 deletion(-)
> >
> > diff --git a/net/core/neighbour.c b/net/core/neighbour.c
> > index 5b669eb80270..167826200f3e 100644
> > --- a/net/core/neighbour.c
> > +++ b/net/core/neighbour.c
> > @@ -328,7 +328,7 @@ static void pneigh_queue_purge(struct 
> > sk_buff_head *list, struct net *net)
> >               __skb_unlink(skb, list);
> >                 dev_put(dev);
> > -            kfree_skb(skb);
> > +            dev_kfree_skb_irq(skb);
> >           }
> >           skb = skb_next;
> >       }
>
> Technically this is pretty much correct, but would it be better to
> move all skb to purge into the new list and after that purge
> them at once?
>
> what about something like this?
>
> diff --git a/net/core/neighbour.c b/net/core/neighbour.c
> index aa16a8017c5e..f7e30daa46ae 100644
> --- a/net/core/neighbour.c
> +++ b/net/core/neighbour.c
> @@ -311,6 +311,9 @@ static void pneigh_queue_purge(struct sk_buff_head 
> *list, struct net *net)
>  {
>         unsigned long flags;
>         struct sk_buff *skb;
> +       struct sk_buff_head tmp;
> +
> +       skb_queue_head_init(&tmp);
>
>         spin_lock_irqsave(&list->lock, flags);
>         skb = skb_peek(list);
> @@ -318,12 +321,16 @@ static void pneigh_queue_purge(struct 
> sk_buff_head *list, struct net *net)
>                 struct sk_buff *skb_next = skb_peek_next(skb, list);
>                 if (net == NULL || net == dev_net(skb->dev)) {
>                         __skb_unlink(skb, list);
> -                       dev_put(skb->dev);
> -                       kfree_skb(skb);
> +                       __skb_queue_tail(&tmp, skb);
>                 }
>                 skb = skb_next;
>         } while (skb != NULL);
>         spin_unlock_irqrestore(&list->lock, flags);
> +
> +       while ((skb = __skb_dequeue(&tmp)) != NULL) {
> +               dev_put(skb->dev);
> +               kfree_skb(skb);
> +       }
>  }
It's better, I can send a v2 later.

Thanks,
Yang
> iris ~/src/linux-2.6 $
>
> .