Re: Followup needed for netfilter: nft_fib_ipv6: walk fib6_siblings under RCU?

[Jiayuan Chen <jiayuan.chen@linux.dev>]

On 5/26/26 1:53 AM, Florian Westphal wrote:
> Hello,
>
> Sashiko claims we need this change:
>
> diff --git a/net/ipv6/netfilter/nft_fib_ipv6.c b/net/ipv6/netfilter/nft_fib_ipv6.c
> --- a/net/ipv6/netfilter/nft_fib_ipv6.c
> +++ b/net/ipv6/netfilter/nft_fib_ipv6.c
> @@ -191,6 +191,9 @@ static bool nft_fib6_info_nh_uses_dev(struct fib6_info *rt,
>   
>                  if (nft_fib6_info_nh_dev_match(nh_dev, dev))
>                          return true;
> +
> +               if (!READ_ONCE(rt->fib6_nsiblings))
> +                       return false;
>          }
>   
>          return false;
>
>
> Could ypu please check and send a followup if its correct?
> AFAICS the nodes can be relinked to a*different* list and
> then this turns into busyloop without the extra test.
>
> https://sashiko.dev/#/patchset/20260520023411.391233-1-jiayuan.chen%40linux.dev
>
> Thank you.


Hi,

Thanks, you're right. net/ipv6/route.c already uses the same
pattern, and Eric applied similar fixes:

f8d8ce1b515a ("ipv6: fix possible infinite loop in fib6_info_uses_dev()")
54e6fe9dd3b0 ("ipv6: prevent infinite loop in rt6_nlmsg_size()")

I'll send a follow-up patch.

Thanks,
Jiayuan