Re: [PATCH net 1/1] ipv4: reject RTAX_MTU values below IPV4_MIN_MTU

在 2026/8/12 20:08, Ido Schimmel 写道:
> On Sat, Aug 08, 2026 at 04:01:15PM +0800, Ren Wei wrote:
>> From: Yong Wang <[email protected]>
>>
>> ip_metrics_convert() caps RTAX_MTU at the IPv4 maximum, but it still
>> accepts undersized non-zero values from userspace.
>>
>> A route installed with "mtu lock 20" can later reach the IPv4
>> forwarding fragmentation path. With a normal 20-byte IPv4 header,
>> ip_do_fragment() reduces the payload MTU to zero. ip_frag_next() then
>> keeps producing zero-length payload fragments, so the fragmentation
>> state never makes forward progress and the kernel loops until the
>> softlockup detector fires.
>>
>> Reject non-zero RTAX_MTU values smaller than IPV4_MIN_MTU while keeping
>> the existing "0 means use default MTU" behavior intact.
>>
>> This fixes the bug at the route metric input point and avoids adding
>> redundant checks in the fragmentation path.
> 
> Sashiko is correct that this is also reproducible without setting an MTU
> lock. See [1].
> 
> Better to fix it in ip_do_fragment(), in a similar fashion to IPv6.
> Something like [2].
> 
> Sashiko review:
> 
> https://netdev-ai.bots.linux.dev/sashiko/#/patchset/ccd14fb1411b8b9c466065582e43f6a6c0743842.1786094799.git.edragain%40163.com
> 
> And please note:
> 
> "Patch authors are expected to proactively look into the AI-generated
> reviews and handle such feedback as any other kind of review: either
> debate it or address it. In both cases a reply on the mailing list is
> expected."
> 
> https://docs.kernel.org/next/process/maintainer-netdev.html#review-timelines
> 
> [1]
> #!/bin/bash
> 
> sysctl -w net.ipv4.ip_forward=1
> 
> ip link add name dummy1 up mtu 20 type dummy
> ip address add 192.0.2.1/24 dev dummy1
> 
> ip link add veth0 type veth peer name veth1
> ip addr add 198.51.100.1/24 dev veth0
> ip link set veth0 up
> 
> ip netns add ns1
> ip link set veth1 netns ns1
> ip -n ns1 address add 198.51.100.2/24 dev veth1
> ip -n ns1 link set veth1 up
> ip -n ns1 route add default via 198.51.100.1
> ip netns exec ns1 ping -M dont -s 1000 -c 1 192.0.2.2
> 
> [2]
> diff --git a/net/ipv4/ip_output.c b/net/ipv4/ip_output.c
> index e6dd1e5b8c32..e6bbae103e4f 100644
> --- a/net/ipv4/ip_output.c
> +++ b/net/ipv4/ip_output.c
> @@ -790,6 +790,12 @@ int ip_do_fragment(struct net *net, struct sock *sk, struct sk_buff *skb,
>  	 */
>  
>  	hlen = iph->ihl * 4;
> +
> +	if (mtu < hlen + 8) {
> +		err = -EMSGSIZE;
> +		goto fail;
> +	}
> +
>  	mtu = mtu - hlen;	/* Size of data space */
>  	IPCB(skb)->flags |= IPSKB_FRAG_COMPLETE;
>  	ll_rs = LL_RESERVED_SPACE(rt->dst.dev);
Thanks for the review.

I checked the Sashiko feedback and reproduced the issue without route
MTU lock as well, using only a 20-byte MTU on the egress dummy device.

So the metrics side validation is not sufficient. I moved the fix to
ip_do_fragment(), adding a direct mtu < hlen + 8 check there.
I also verified that the old kernel still reproduces the
softlockup in the no-mtu-lock case, while the updated kernel no longer
gets stuck.

I will resend with that version.