route

How IPv6 addresses are flushed on link down

IPv6协议栈里， 当一个网口被down之后，网口上对应的IP地址也会一起被flush掉。
面对IPv6跟IPv4不同的行为方式，内核提供了一个规避的开关。
在4.6内核之后提供了一个开关，用来避免IPv6地址别清理掉。
这个开关既有全局的设置，也有每个interface粒度的单独开关。

1
2
3
4
5
6
7
keep_addr_on_down - INTEGER
    Keep all IPv6 addresses on an interface down event. If set static
    global addresses with no expiration time are not flushed.
      >0 : enabled
       0 : system default
      <0 : disabled
    Default: 0 (addresses are removed)

call stack

1
2
-> addrconf_init
-> -> register_netdevice_notifier(ipv6_dev_notf)
1
2
-> addrconf_notify
-> ->

register call back function

IPv6 register a callback function for link down message.

1
7301         register_netdevice_notifier(&ipv6_dev_notf);
1
2
3
4
3713 static struct notifier_block ipv6_dev_notf = {
3714         .notifier_call = addrconf_notify,
3715         .priority = ADDRCONF_NOTIFY_PRIORITY,
3716 };
1
2
3
4
5
6
7
8
9
10
3514 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3515                            void *ptr)
3516 {
3668         case NETDEV_DOWN:
3669         case NETDEV_UNREGISTER:
3670                 /*
3671                  *      Remove all addresses from this interface.
3672                  */
3673                 addrconf_ifdown(dev, event != NETDEV_DOWN);
3674                 break;

遍历hash链里配置在本idev下的所有ipv6 地址节点ifa，并把它们从系统里删除。

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
3737 static int addrconf_ifdown(struct net_device *dev, bool unregister)
3738 {
...
3784         /* Step 2: clear hash table */
3785         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3786                 struct hlist_head *h = &net->ipv6.inet6_addr_lst[i];
3787
3788                 spin_lock_bh(&net->ipv6.addrconf_hash_lock);
3789 restart:
3790                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3791                         if (ifa->idev == idev) {
3792                                 addrconf_del_dad_work(ifa);
3793                                 /* combined flag + permanent flag decide if
3794                                  * address is retained on a down event
3795                                  */
3796                                 if (!keep_addr ||
3797                                     !(ifa->flags & IFA_F_PERMANENT) ||
3798                                     addr_is_local(&ifa->addr)) {
3799                                         hlist_del_init_rcu(&ifa->addr_lst);
3800                                         goto restart;
3801                                 }
3802                         }
3803                 }
3804                 spin_unlock_bh(&net->ipv6.addrconf_hash_lock);
3805         }

keep_addr_on_down

面对IPv6跟IPv4不同的行为方式，内核提供了一个规避的开关。
在4.6内核之后提供了一个开关，用来避免IPv6地址别清理掉。
这个开关既有全局的设置，也有每个interface粒度的单独开关。
系统在flush前会看各个开关的并联效果。

比较详细的介绍和讨论见：
https://serverfault.com/questions/842542/why-are-ipv6-addresses-flushed-on-link-down

1
2
3
4
5
6
7
8
9
10
11
12
3771         /* combine the user config with event to determine if permanent
3772          * addresses are to be removed from address hash table
3773          */
3774         if (!unregister && !idev->cnf.disable_ipv6) {
3775                 /* aggregate the system setting and interface setting */
3776                 int _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3777
3778                 if (!_keep_addr)
3779                         _keep_addr = idev->cnf.keep_addr_on_down;
3780
3781                 keep_addr = (_keep_addr > 0);
3782         }