On 2/17/16 11:05 AM, David Miller wrote:
From: David Ahern <d...@cumulusnetworks.com> Date: Sat, 13 Feb 2016 14:23:27 -0800@@ -3427,31 +3493,52 @@ static int addrconf_ifdown(struct net_device *dev, int how) write_lock_bh(&idev->lock); } - while (!list_empty(&idev->addr_list)) { - ifa = list_first_entry(&idev->addr_list, - struct inet6_ifaddr, if_list); - addrconf_del_dad_work(ifa); + INIT_LIST_HEAD(&del_list); + list_for_each_entry_safe(ifa, tmp, &idev->addr_list, if_list) { + bool keep_ifa = false; - list_del(&ifa->if_list); + if (!how && keep_addr && ifa->user_managed) + keep_ifa = true;I think it would make sense to evaluate "!how && keep_addr" outside the loop. The only thing that changes is ifa->user_managed on each iteration. But I also want some more documentation in what you are doing here. I understand the address flushing on ifdown avoidance, but all of this user_managed logic is not mentioned at all. Why do you need it? What role does it play in achieving your goal?
Per prior comment user_managed will go away in favor of checking IFA_F_PERMANENT. We are only keeping permanent addresses which in past versions of the patch were marked with user_managed flag but it is redundant.
