Re: [PATCH net 2/2] net: Warn when overriding referenced dst entry

[Jakub Kicinski <kuba@kernel.org>]

On Tue, 22 Jul 2025 14:02:56 -0700 Stanislav Fomichev wrote:
> Add WARN_ON_ONCE with CONFIG_NET_DEBUG to warn on overriding referenced
> dst entries. This should get a better signal than tracing leaked
> objects from kmemleak.

Looks like we're tripping the severe nastiness in icmp.c:

		/* Ugh! */
		orefdst = skb_in->_skb_refdst; /* save old refdst */
		skb_dst_set(skb_in, NULL);
		err = ip_route_input(skb_in, fl4_dec.daddr, fl4_dec.saddr,
				     dscp, rt2->dst.dev) ? -EINVAL : 0;

		dst_release(&rt2->dst);
		rt2 = skb_rtable(skb_in);
		skb_in->_skb_refdst = orefdst; /* restore old refdst */

There's more of these around. I think we need a new helper for
"saving" and "restoring" the refdst? Either way, I reckon the
patch with the check belongs in net-next.

> diff --git a/include/linux/skbuff.h b/include/linux/skbuff.h
> index 5520524c93bf..c89931ac01e5 100644
> --- a/include/linux/skbuff.h
> +++ b/include/linux/skbuff.h
> @@ -1159,6 +1159,12 @@ static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
>  	return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
>  }
>  
> +static inline void skb_dst_check(struct sk_buff *skb)

skb_dst_check_unset() ? Right now we assume the caller will override.
Or will there be more conditions added?

> +{
> +	DEBUG_NET_WARN_ON_ONCE(skb_dst(skb) &&
> +			       !(skb->_skb_refdst & SKB_DST_NOREF));

Why not 

	DEBUG_NET_WARN_ON_ONCE(skb->_skb_refdst & SKB_DST_PTRMASK);

?
-- 
pw-bot: cr