]> www.infradead.org Git - users/dwmw2/linux.git/commitdiff
ipv4: Convert RTM_NEWADDR to per-netns RTNL.
authorKuniyuki Iwashima <kuniyu@amazon.com>
Mon, 21 Oct 2024 18:32:32 +0000 (11:32 -0700)
committerPaolo Abeni <pabeni@redhat.com>
Tue, 29 Oct 2024 10:54:57 +0000 (11:54 +0100)
The address hash table and GC are already namespacified.

Let's push down RTNL into inet_rtm_newaddr() as rtnl_net_lock().

Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
net/ipv4/devinet.c

index 8db84c70ebedb07e21fa2b071b182a047f9df448..7f24bc38981b62a03994b562ccb2edf4949b5ea5 100644 (file)
@@ -974,8 +974,6 @@ static int inet_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
        struct in_ifaddr *ifa;
        int ret;
 
-       ASSERT_RTNL();
-
        ret = inet_validate_rtm(nlh, tb, extack, &valid_lft, &prefered_lft);
        if (ret < 0)
                return ret;
@@ -983,9 +981,13 @@ static int inet_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
        if (!nla_get_in_addr(tb[IFA_LOCAL]))
                return 0;
 
+       rtnl_net_lock(net);
+
        ifa = inet_rtm_to_ifa(net, nlh, tb, extack);
-       if (IS_ERR(ifa))
-               return PTR_ERR(ifa);
+       if (IS_ERR(ifa)) {
+               ret = PTR_ERR(ifa);
+               goto unlock;
+       }
 
        ifa_existing = find_matching_ifa(ifa);
        if (!ifa_existing) {
@@ -998,11 +1000,11 @@ static int inet_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
                        if (ret < 0) {
                                NL_SET_ERR_MSG(extack, "ipv4: Multicast auto join failed");
                                inet_free_ifa(ifa);
-                               return ret;
+                               goto unlock;
                        }
                }
-               return __inet_insert_ifa(ifa, nlh, NETLINK_CB(skb).portid,
-                                        extack);
+
+               ret = __inet_insert_ifa(ifa, nlh, NETLINK_CB(skb).portid, extack);
        } else {
                u32 new_metric = ifa->ifa_rt_priority;
                u8 new_proto = ifa->ifa_proto;
@@ -1012,7 +1014,8 @@ static int inet_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
                if (nlh->nlmsg_flags & NLM_F_EXCL ||
                    !(nlh->nlmsg_flags & NLM_F_REPLACE)) {
                        NL_SET_ERR_MSG(extack, "ipv4: Address already assigned");
-                       return -EEXIST;
+                       ret = -EEXIST;
+                       goto unlock;
                }
                ifa = ifa_existing;
 
@@ -1029,7 +1032,11 @@ static int inet_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
                                   &net->ipv4.addr_chk_work, 0);
                rtmsg_ifa(RTM_NEWADDR, ifa, nlh, NETLINK_CB(skb).portid);
        }
-       return 0;
+
+unlock:
+       rtnl_net_unlock(net);
+
+       return ret;
 }
 
 /*
@@ -2823,7 +2830,8 @@ static struct rtnl_af_ops inet_af_ops __read_mostly = {
 };
 
 static const struct rtnl_msg_handler devinet_rtnl_msg_handlers[] __initconst = {
-       {.protocol = PF_INET, .msgtype = RTM_NEWADDR, .doit = inet_rtm_newaddr},
+       {.protocol = PF_INET, .msgtype = RTM_NEWADDR, .doit = inet_rtm_newaddr,
+        .flags = RTNL_FLAG_DOIT_PERNET},
        {.protocol = PF_INET, .msgtype = RTM_DELADDR, .doit = inet_rtm_deladdr},
        {.protocol = PF_INET, .msgtype = RTM_GETADDR, .dumpit = inet_dump_ifaddr,
         .flags = RTNL_FLAG_DUMP_UNLOCKED | RTNL_FLAG_DUMP_SPLIT_NLM_DONE},