On Thu, May 4, 2017 at 7:04 AM, David Ahern <dsah...@gmail.com> wrote:
> On 5/3/17 11:07 PM, Cong Wang wrote:
>> For each netns (except init_net), we initialize its null entry
>> in 3 places:
>>
>> 1) The template itself, as we use kmemdup()
>> 2) Code around dst_init_metrics() in ip6_route_net_init()
>> 3) ip6_route_dev_notify(), which is supposed to initialize it after
>> loopback registers
>>
>> Unfortunately the last one still happens in a wrong order because
>> we expect to initialize net->ipv6.ip6_null_entry->rt6i_idev to
>> net->loopback_dev's idev, so we have to do that after we add
>> idev to it. However, this notifier has priority == 0 same as
>> ipv6_dev_notf, and ipv6_dev_notf is registered after
>> ip6_route_dev_notifier so it is called actually after
>> ip6_route_dev_notifier.
>>
>> Fix it by specifying a smaller priority for ip6_route_dev_notifier.
>>
>> Signed-off-by: Cong Wang <xiyou.wangc...@gmail.com>
>> ---
>> net/ipv6/route.c | 2 +-
>> 1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/net/ipv6/route.c b/net/ipv6/route.c
>> index 2f11366..4dbf7e2 100644
>> --- a/net/ipv6/route.c
>> +++ b/net/ipv6/route.c
>> @@ -4024,7 +4024,7 @@ static struct pernet_operations ip6_route_net_late_ops
>> = {
>>
>> static struct notifier_block ip6_route_dev_notifier = {
>> .notifier_call = ip6_route_dev_notify,
>> - .priority = 0,
>> + .priority = -10, /* Must be called after addrconf_notify!! */
>> };
>>
>> void __init ip6_route_init_special_entries(void)
>>
>
> And I see a refcnt problem with this change:
>
> root@kenny-jessie2:~# unshare -n
> root@kenny-jessie2:~# logout
> root@kenny-jessie2:~# unshare -n
>
> Message from syslogd@kenny-jessie2 at May 4 07:04:38 ...
> kernel:[ 62.581552] unregister_netdevice: waiting for lo to become
> free. Usage count = 1
Ah, looks like we need to put the refcnt for UNREGISTER too.
Will send v2 to include your ADDRCONF_NOTIFY_PRIORITY suggestion.
