Re: [PATCH net] ipv6: rpl: unclone the skb before modifying the packet

[Andrea Mayer]

On Thu, 01 Oct 2026 18:29:07 +0000
netdev-bot+sinfo@xxxxxxxxxx wrote:

> Hi!
> 
> This is an automated message. This series looks like a fix, but its
> commit messages seem to be missing some information:
> 
>  - Whether the issue was actually triggered, or is only theoretical
>    (e.g. found by code inspection). If it was triggered please include
>    the symptoms, like the stack trace or error messages.
> 
> [snip]

Hi,

Yes, I triggered it. I reproduced it in a VM.

One namespace sends pings to fc00:2::1 through the RPL segment
fc00::2, which belongs to a second namespace with rpl_seg_enabled=1.
Each ping arrives with destination address fc00::2 and a routing header
with Segments Left 1 that carries fc00:2::1. In the second namespace, a
Python program opens an AF_PACKET socket and reads the queued frames
only after the pings were processed. For each frame with a routing
header, it prints the destination address and Segments Left.

Without the fix, all these frames showed:

  daddr fc00:2::1 segments_left 0

These are the values written by ipv6_rpl_srh_rcv(), not the received
ones. With the fix, all the frames with a routing header showed the
received values:

  daddr fc00::2 segments_left 1

To reproduce it, I did not modify the kernel or use error injection.

This is how the packet reaches ipv6_rpl_srh_rcv(), with some calls
left out:

__netif_receive_skb_one_core
  __netif_receive_skb_core
    deliver_skb               [orig: users=2]
      packet_rcv
        skb_clone             clone queued to the AF_PACKET socket
        consume_skb(orig)     [orig: users=1, cloned=1]
  ipv6_rcv
    ip6_rcv_core              skb_share_check: no-op [orig: users=1]
    [...]
      ip6_protocol_deliver_rcu
        ipv6_rthdr_rcv
          ipv6_rpl_srh_rcv    writes into the data shared with the clone

The reproducer is a shell script and a Python program.
I can post it if it helps.

Thanks,
Andrea