[PATCH net] Revert "ipv6: preserve insertion order for same-scope addresses"

[PATCH net] Revert "ipv6: preserve insertion order for same-scope addresses"

[Fernando Fernandez Mancera]

Chris Adams reported that preserving insertion order for same-scope
addresses is causing SSH connections to be dropped after stopping a VM
while running NetworkManager.

NetworkManager is caching the networking information from the system and
later operating with it. It relies on the order of IPv6 addresses to
configure routes for outgoing traffic. Revert to the previous ordering.

This reverts commit cb3de96eea66f5e4a580086c6a1be46e765f97f4.

Fixes: cb3de96eea66 ("ipv6: preserve insertion order for same-scope addresses")
Reported-by: Chris Adams <linux@cmadams.net>
Closes: https://lore.kernel.org/netdev/20260521135310.GC977@cmadams.net/
Signed-off-by: Fernando Fernandez Mancera <fmancera@suse.de>
---
 net/ipv6/addrconf.c                  | 2 +-
 tools/testing/selftests/net/ioam6.sh | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/net/ipv6/addrconf.c b/net/ipv6/addrconf.c
index 5476b6536eb7..bb84a78b80f6 100644
--- a/net/ipv6/addrconf.c
+++ b/net/ipv6/addrconf.c
@@ -1013,7 +1013,7 @@ ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 	list_for_each(p, &idev->addr_list) {
 		struct inet6_ifaddr *ifa
 			= list_entry(p, struct inet6_ifaddr, if_list);
-		if (ifp_scope > ipv6_addr_src_scope(&ifa->addr))
+		if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 			break;
 	}
 
diff --git a/tools/testing/selftests/net/ioam6.sh b/tools/testing/selftests/net/ioam6.sh
index b2b99889942f..845c26dd01a9 100755
--- a/tools/testing/selftests/net/ioam6.sh
+++ b/tools/testing/selftests/net/ioam6.sh
@@ -273,8 +273,8 @@ setup()
   ip -netns $ioam_node_beta link set ioam-veth-betaR name veth1 &>/dev/null
   ip -netns $ioam_node_gamma link set ioam-veth-gamma name veth0 &>/dev/null
 
-  ip -netns $ioam_node_alpha addr add 2001:db8:1::2/64 dev veth0 &>/dev/null
   ip -netns $ioam_node_alpha addr add 2001:db8:1::50/64 dev veth0 &>/dev/null
+  ip -netns $ioam_node_alpha addr add 2001:db8:1::2/64 dev veth0 &>/dev/null
   ip -netns $ioam_node_alpha link set veth0 up &>/dev/null
   ip -netns $ioam_node_alpha link set lo up &>/dev/null
   ip -netns $ioam_node_alpha route add 2001:db8:2::/64 \
-- 
2.54.0

Re: [PATCH net] Revert "ipv6: preserve insertion order for same-scope addresses"

[Stefano Brivio]

On Thu, 28 May 2026 13:11:23 +0200
Fernando Fernandez Mancera <fmancera@suse.de> wrote:

> Chris Adams reported that preserving insertion order for same-scope
> addresses is causing SSH connections to be dropped after stopping a VM
> while running NetworkManager.
> 
> NetworkManager is caching the networking information from the system and
> later operating with it. It relies on the order of IPv6 addresses to
> configure routes for outgoing traffic.

I think this paragraph should make it clear that NetworkManager relies
on a _reversed_ order (first inserted is returned last), and that it's
the opposite compared to IPv4.

> Revert to the previous ordering.

See my concern from:

  https://lore.kernel.org/all/20260528131250.1352ab48@elisabeth/

(sorry, I sent that email moments after you sent this patch it seems). It
fixes the issue for NetworkManager assuming it's the only issue, but it
causes a regression for others.

If this revert is the only practical solution, so be it. I'm just wondering
if it really is.

> This reverts commit cb3de96eea66f5e4a580086c6a1be46e765f97f4.
> 
> Fixes: cb3de96eea66 ("ipv6: preserve insertion order for same-scope addresses")
> Reported-by: Chris Adams <linux@cmadams.net>
> Closes: https://lore.kernel.org/netdev/20260521135310.GC977@cmadams.net/
> Signed-off-by: Fernando Fernandez Mancera <fmancera@suse.de>
> ---
>  net/ipv6/addrconf.c                  | 2 +-
>  tools/testing/selftests/net/ioam6.sh | 2 +-
>  2 files changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/net/ipv6/addrconf.c b/net/ipv6/addrconf.c
> index 5476b6536eb7..bb84a78b80f6 100644
> --- a/net/ipv6/addrconf.c
> +++ b/net/ipv6/addrconf.c
> @@ -1013,7 +1013,7 @@ ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
>  	list_for_each(p, &idev->addr_list) {
>  		struct inet6_ifaddr *ifa
>  			= list_entry(p, struct inet6_ifaddr, if_list);
> -		if (ifp_scope > ipv6_addr_src_scope(&ifa->addr))
> +		if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
>  			break;
>  	}
>  
> diff --git a/tools/testing/selftests/net/ioam6.sh b/tools/testing/selftests/net/ioam6.sh
> index b2b99889942f..845c26dd01a9 100755
> --- a/tools/testing/selftests/net/ioam6.sh
> +++ b/tools/testing/selftests/net/ioam6.sh
> @@ -273,8 +273,8 @@ setup()
>    ip -netns $ioam_node_beta link set ioam-veth-betaR name veth1 &>/dev/null
>    ip -netns $ioam_node_gamma link set ioam-veth-gamma name veth0 &>/dev/null
>  
> -  ip -netns $ioam_node_alpha addr add 2001:db8:1::2/64 dev veth0 &>/dev/null
>    ip -netns $ioam_node_alpha addr add 2001:db8:1::50/64 dev veth0 &>/dev/null
> +  ip -netns $ioam_node_alpha addr add 2001:db8:1::2/64 dev veth0 &>/dev/null
>    ip -netns $ioam_node_alpha link set veth0 up &>/dev/null
>    ip -netns $ioam_node_alpha link set lo up &>/dev/null
>    ip -netns $ioam_node_alpha route add 2001:db8:2::/64 \

-- 
Stefano

Re: [PATCH net] Revert "ipv6: preserve insertion order for same-scope addresses"

[Fernando Fernandez Mancera]

On 5/28/26 1:35 PM, Stefano Brivio wrote:
> On Thu, 28 May 2026 13:11:23 +0200
> Fernando Fernandez Mancera <fmancera@suse.de> wrote:
> 
>> Chris Adams reported that preserving insertion order for same-scope
>> addresses is causing SSH connections to be dropped after stopping a VM
>> while running NetworkManager.
>>
>> NetworkManager is caching the networking information from the system and
>> later operating with it. It relies on the order of IPv6 addresses to
>> configure routes for outgoing traffic.
> 
> I think this paragraph should make it clear that NetworkManager relies
> on a _reversed_ order (first inserted is returned last), and that it's
> the opposite compared to IPv4.
> 

FWIW; I still think that NetworkManager relies on the ordering but not 
for the route.. AFAIU, the route should be for the prefix and should not 
change at all.

So I need to figure out what is changing and document it properly on the 
commit.

>> Revert to the previous ordering.
> 
> See my concern from:
> 
>    https://lore.kernel.org/all/20260528131250.1352ab48@elisabeth/
> 
> (sorry, I sent that email moments after you sent this patch it seems). It
> fixes the issue for NetworkManager assuming it's the only issue, but it
> causes a regression for others.
> 
> If this revert is the only practical solution, so be it. I'm just wondering
> if it really is.
> 
>> This reverts commit cb3de96eea66f5e4a580086c6a1be46e765f97f4.
>>
>> Fixes: cb3de96eea66 ("ipv6: preserve insertion order for same-scope addresses")
>> Reported-by: Chris Adams <linux@cmadams.net>
>> Closes: https://lore.kernel.org/netdev/20260521135310.GC977@cmadams.net/
>> Signed-off-by: Fernando Fernandez Mancera <fmancera@suse.de>
>> ---
>>   net/ipv6/addrconf.c                  | 2 +-
>>   tools/testing/selftests/net/ioam6.sh | 2 +-
>>   2 files changed, 2 insertions(+), 2 deletions(-)
>>
>> diff --git a/net/ipv6/addrconf.c b/net/ipv6/addrconf.c
>> index 5476b6536eb7..bb84a78b80f6 100644
>> --- a/net/ipv6/addrconf.c
>> +++ b/net/ipv6/addrconf.c
>> @@ -1013,7 +1013,7 @@ ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
>>   	list_for_each(p, &idev->addr_list) {
>>   		struct inet6_ifaddr *ifa
>>   			= list_entry(p, struct inet6_ifaddr, if_list);
>> -		if (ifp_scope > ipv6_addr_src_scope(&ifa->addr))
>> +		if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
>>   			break;
>>   	}
>>   
>> diff --git a/tools/testing/selftests/net/ioam6.sh b/tools/testing/selftests/net/ioam6.sh
>> index b2b99889942f..845c26dd01a9 100755
>> --- a/tools/testing/selftests/net/ioam6.sh
>> +++ b/tools/testing/selftests/net/ioam6.sh
>> @@ -273,8 +273,8 @@ setup()
>>     ip -netns $ioam_node_beta link set ioam-veth-betaR name veth1 &>/dev/null
>>     ip -netns $ioam_node_gamma link set ioam-veth-gamma name veth0 &>/dev/null
>>   
>> -  ip -netns $ioam_node_alpha addr add 2001:db8:1::2/64 dev veth0 &>/dev/null
>>     ip -netns $ioam_node_alpha addr add 2001:db8:1::50/64 dev veth0 &>/dev/null
>> +  ip -netns $ioam_node_alpha addr add 2001:db8:1::2/64 dev veth0 &>/dev/null
>>     ip -netns $ioam_node_alpha link set veth0 up &>/dev/null
>>     ip -netns $ioam_node_alpha link set lo up &>/dev/null
>>     ip -netns $ioam_node_alpha route add 2001:db8:2::/64 \
>