Re: [PATCH net v2 2/2] net: ipv6: Clamp to IP6_MAX_MTU in ip6_dst_mtu_maybe_forward

["Alice Mikityanska" <alice.kernel@fastmail.im>]

On Thu, Aug 13, 2026, at 23:25, Willem de Bruijn wrote:
> Alice Mikityanska wrote:
>> From: Alice Mikityanska <alice@isovalent.com>
>> 
>> Commit 427faee167bc ("net: ipv6: introduce ip6_dst_mtu_maybe_forward")
>> dropped the IP6_MAX_MTU clamp that used to be present in ip6_mtu(). A
>> similar IPv4 commit ac6627a28dbf ("net: ipv4: Consolidate ipv4_mtu and
>> ip_dst_mtu_maybe_forward") preserves the IP_MAX_MTU clamp.
>> 
>> Restore the upper bound in the IPv6 flow to avoid potential 16-bit
>> overflows in forwarding paths.
>> 
>> Fixes: 427faee167bc ("net: ipv6: introduce ip6_dst_mtu_maybe_forward")
>> Signed-off-by: Alice Mikityanska <alice@isovalent.com>
>> Suggested-by: Willem de Bruijn <willemdebruijn.kernel@gmail.com>
>
> Reviewed-by: Willem de Bruijn <willemb@google.com>
>
>> ---
>>  include/net/ip6_route.h | 2 ++
>>  1 file changed, 2 insertions(+)
>> 
>> diff --git a/include/net/ip6_route.h b/include/net/ip6_route.h
>> index 09ffe0f13ce7..fb59a5885faa 100644
>> --- a/include/net/ip6_route.h
>> +++ b/include/net/ip6_route.h
>> @@ -382,6 +382,8 @@ static inline unsigned int ip6_dst_mtu_maybe_forward(const struct dst_entry *dst
>>  	rcu_read_unlock();
>>  
>>  out:
>> +	mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
>> +
>>  	return mtu - lwtunnel_headroom(dst->lwtstate, mtu);
>>  }
>
> It appears IPv4 only clamps device MTU, not route MTU:
>
>         mtu = dst_metric_raw(dst, RTAX_MTU);
>         if (!mtu)
>                 mtu = min(READ_ONCE(dst->dev->mtu), IP_MAX_MTU);

This is some old code from v5.14, it changed in commit ac6627a28dbf
("net: ipv4: Consolidate ipv4_mtu and ip_dst_mtu_maybe_forward"), and
IPv4 clamps MTU in both cases since then.

> I don't think that was necessarily intentional. Perhaps route MTU
> itself is already bounds checked on configuration. The device MTU
> min() was added after a syzbot report, in commit c780a049f9b.

This commit merely adds READ_ONCE to the existing min.

> Current IPv6 proposal is arguably more robust, covering both. There
> just remains a difference between IPv4 and IPv6 code paths.

So, looking at the fresh checkout, it seems that my patch covers the
difference, right?