Re: [PATCH net-next v2] ipv6: preserve insertion order for same-scope addresses

[David Ahern <dsahern@kernel.org>]

On 1/3/26 8:23 PM, Yumei Huang wrote:
> IPv6 addresses with the same scope are returned in reverse insertion
> order, unlike IPv4. For example, when adding a -> b -> c, the list is
> reported as c -> b -> a, while IPv4 preserves the original order.
> 
> This behavior causes:
> 
> a. When using `ip -6 a save` and `ip -6 a restore`, addresses are restored
>    in the opposite order from which they were saved. See example below
>    showing addresses added as 1::1, 1::2, 1::3 but displayed and saved
>    in reverse order.
> 
>    # ip -6 a a 1::1 dev x
>    # ip -6 a a 1::2 dev x
>    # ip -6 a a 1::3 dev x
>    # ip -6 a s dev x
>    2: x: <BROADCAST,MULTICAST> mtu 1500 qdisc noop state DOWN group default qlen 1000
>        inet6 1::3/128 scope global tentative
>        valid_lft forever preferred_lft forever
>        inet6 1::2/128 scope global tentative
>        valid_lft forever preferred_lft forever
>        inet6 1::1/128 scope global tentative
>        valid_lft forever preferred_lft forever
>    # ip -6 a save > dump
>    # ip -6 a d 1::1 dev x
>    # ip -6 a d 1::2 dev x
>    # ip -6 a d 1::3 dev x
>    # ip a d ::1 dev lo
>    # ip a restore < dump
>    # ip -6 a s dev x
>    2: x: <BROADCAST,MULTICAST> mtu 1500 qdisc noop state DOWN group default qlen 1000
>        inet6 1::1/128 scope global tentative
>        valid_lft forever preferred_lft forever
>        inet6 1::2/128 scope global tentative
>        valid_lft forever preferred_lft forever
>        inet6 1::3/128 scope global tentative
>        valid_lft forever preferred_lft forever
>    # ip a showdump < dump
>     if1:
>         inet6 ::1/128 scope host proto kernel_lo
>         valid_lft forever preferred_lft forever
>     if2:
>         inet6 1::3/128 scope global tentative
>         valid_lft forever preferred_lft forever
>     if2:
>         inet6 1::2/128 scope global tentative
>         valid_lft forever preferred_lft forever
>     if2:
>         inet6 1::1/128 scope global tentative
>         valid_lft forever preferred_lft forever
> 
> b. Addresses in pasta to appear in reversed order compared to host
>    addresses.
> 
> The ipv6 addresses were added in reverse order by commit e55ffac60117
> ("[IPV6]: order addresses by scope"), then it was changed by commit
> 502a2ffd7376 ("ipv6: convert idev_list to list macros"), and restored by
> commit b54c9b98bbfb ("ipv6: Preserve pervious behavior in
> ipv6_link_dev_addr()."). However, this reverse ordering within the same
> scope causes inconsistency with IPv4 and the issues described above.
> 
> This patch aligns IPv6 address ordering with IPv4 for consistency
> by changing the comparison from >= to > when inserting addresses
> into the address list. Also updates the ioam6 selftest to reflect
> the new address ordering behavior. Combine these two changes into
> one patch for bisectability.
> 
> Fixes: e55ffac60117 ("[IPV6]: order addresses by scope")
> Link: https://bugs.passt.top/show_bug.cgi?id=175
> Suggested-by: Stefano Brivio <sbrivio@redhat.com>
> Signed-off-by: Yumei Huang <yuhuang@redhat.com>
> ---
>  net/ipv6/addrconf.c                  | 2 +-
>  tools/testing/selftests/net/ioam6.sh | 2 +-
>  2 files changed, 2 insertions(+), 2 deletions(-)
> 
>

I am ok with the change, but I will not be surprised if someone
complains about a change in behavior.

Reviewed-by: David Ahern <dsahern@kernel.org>