rtnl: Add support for netdev event to link messages
authorVlad Yasevich <vyasevich@gmail.com>
Tue, 4 Apr 2017 13:23:42 +0000 (09:23 -0400)
committerDavid S. Miller <davem@davemloft.net>
Wed, 5 Apr 2017 15:14:14 +0000 (08:14 -0700)
When netdev events happen, a rtnetlink_event() handler will send
messages for every event in it's white list.  These messages contain
current information about a particular device, but they do not include
the iformation about which event just happened.  The consumer of
the message has to try to infer this information.  In some cases
(ex: NETDEV_NOTIFY_PEERS), that is not possible.

This patch adds a new extension to RTM_NEWLINK message called IFLA_EVENT
that would have an encoding of the which event triggered this
message.  This would allow the the message consumer to easily determine
if it is interested in a particular event or not.

Signed-off-by: Vladislav Yasevich <vyasevic@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
include/linux/rtnetlink.h
include/uapi/linux/if_link.h
net/core/dev.c
net/core/rtnetlink.c

index 57e54847b0b9e96442b886b9cb05a67948a24c55..0459018173cff14c4e30254f692ff8a73dcec299 100644 (file)
@@ -18,7 +18,8 @@ extern int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst,
 
 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned change, gfp_t flags);
 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
-                                      unsigned change, gfp_t flags);
+                                      unsigned change, unsigned long event,
+                                      gfp_t flags);
 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev,
                       gfp_t flags);
 
index 8b405afb23763498bf0b879ab319dcc9bf94f4b9..97f6d302f6278470a86ec656e0f7bf2909dbb9cc 100644 (file)
@@ -157,6 +157,7 @@ enum {
        IFLA_GSO_MAX_SIZE,
        IFLA_PAD,
        IFLA_XDP,
+       IFLA_EVENT,
        __IFLA_MAX
 };
 
@@ -899,4 +900,24 @@ enum {
 
 #define IFLA_XDP_MAX (__IFLA_XDP_MAX - 1)
 
+enum {
+       IFLA_EVENT_UNSPEC,
+       IFLA_EVENT_REBOOT,
+       IFLA_EVENT_CHANGE_MTU,
+       IFLA_EVENT_CHANGE_ADDR,
+       IFLA_EVENT_CHANGE_NAME,
+       IFLA_EVENT_FEAT_CHANGE,
+       IFLA_EVENT_BONDING_FAILOVER,
+       IFLA_EVENT_POST_TYPE_CHANGE,
+       IFLA_EVENT_NOTIFY_PEERS,
+       IFLA_EVENT_CHANGE_UPPER,
+       IFLA_EVENT_RESEND_IGMP,
+       IFLA_EVENT_PRE_CHANGE_MTU,
+       IFLA_EVENT_CHANGE_INFO_DATA,
+       IFLA_EVENT_PRE_CHANGE_UPPER,
+       IFLA_EVENT_CHANGE_LOWER_STATE,
+       IFLA_EVENT_UDP_TUNNEL_PUSH_INFO,
+       IFLA_EVENT_CHANGE_TX_QUEUE_LEN,
+};
+
 #endif /* _UAPI_LINUX_IF_LINK_H */
index ef9fe60ee294b0e2503456f68136440646c86344..7efb4178ffef7c991df07602e8d4ee683af02355 100644 (file)
@@ -6840,7 +6840,7 @@ static void rollback_registered_many(struct list_head *head)
 
                if (!dev->rtnl_link_ops ||
                    dev->rtnl_link_state == RTNL_LINK_INITIALIZED)
-                       skb = rtmsg_ifinfo_build_skb(RTM_DELLINK, dev, ~0U,
+                       skb = rtmsg_ifinfo_build_skb(RTM_DELLINK, dev, ~0U, 0,
                                                     GFP_KERNEL);
 
                /*
index 58419da7961bfeb2481d7eb7f9c13a88171bb0cb..b2bd4c9ee8600aecaaa74e585ed9e38a8a314e78 100644 (file)
@@ -944,6 +944,7 @@ static noinline size_t if_nlmsg_size(const struct net_device *dev,
               + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_SWITCH_ID */
               + nla_total_size(IFNAMSIZ) /* IFLA_PHYS_PORT_NAME */
               + rtnl_xdp_size(dev) /* IFLA_XDP */
+              + nla_total_size(4)  /* IFLA_EVENT */
               + nla_total_size(1); /* IFLA_PROTO_DOWN */
 
 }
@@ -1276,9 +1277,70 @@ err_cancel:
        return err;
 }
 
+static int rtnl_fill_link_event(struct sk_buff *skb, unsigned long event)
+{
+       u32 rtnl_event;
+
+       switch (event) {
+       case NETDEV_REBOOT:
+               rtnl_event = IFLA_EVENT_REBOOT;
+               break;
+       case NETDEV_CHANGEMTU:
+               rtnl_event = IFLA_EVENT_CHANGE_MTU;
+               break;
+       case NETDEV_CHANGEADDR:
+               rtnl_event = IFLA_EVENT_CHANGE_ADDR;
+               break;
+       case NETDEV_CHANGENAME:
+               rtnl_event = IFLA_EVENT_CHANGE_NAME;
+               break;
+       case NETDEV_FEAT_CHANGE:
+               rtnl_event = IFLA_EVENT_FEAT_CHANGE;
+               break;
+       case NETDEV_BONDING_FAILOVER:
+               rtnl_event = IFLA_EVENT_BONDING_FAILOVER;
+               break;
+       case NETDEV_POST_TYPE_CHANGE:
+               rtnl_event = IFLA_EVENT_POST_TYPE_CHANGE;
+               break;
+       case NETDEV_NOTIFY_PEERS:
+               rtnl_event = IFLA_EVENT_NOTIFY_PEERS;
+               break;
+       case NETDEV_CHANGEUPPER:
+               rtnl_event = IFLA_EVENT_CHANGE_UPPER;
+               break;
+       case NETDEV_RESEND_IGMP:
+               rtnl_event = IFLA_EVENT_RESEND_IGMP;
+               break;
+       case NETDEV_PRECHANGEMTU:
+               rtnl_event = IFLA_EVENT_PRE_CHANGE_MTU;
+               break;
+       case NETDEV_CHANGEINFODATA:
+               rtnl_event = IFLA_EVENT_CHANGE_INFO_DATA;
+               break;
+       case NETDEV_PRECHANGEUPPER:
+               rtnl_event = IFLA_EVENT_PRE_CHANGE_UPPER;
+               break;
+       case NETDEV_CHANGELOWERSTATE:
+               rtnl_event = IFLA_EVENT_CHANGE_LOWER_STATE;
+               break;
+       case NETDEV_UDP_TUNNEL_PUSH_INFO:
+               rtnl_event = IFLA_EVENT_UDP_TUNNEL_PUSH_INFO;
+               break;
+       case NETDEV_CHANGE_TX_QUEUE_LEN:
+               rtnl_event = IFLA_EVENT_CHANGE_TX_QUEUE_LEN;
+               break;
+       default:
+               return 0;
+       }
+
+       return nla_put_u32(skb, IFLA_EVENT, rtnl_event);
+}
+
 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
                            int type, u32 pid, u32 seq, u32 change,
-                           unsigned int flags, u32 ext_filter_mask)
+                           unsigned int flags, u32 ext_filter_mask,
+                           unsigned long event)
 {
        struct ifinfomsg *ifm;
        struct nlmsghdr *nlh;
@@ -1327,6 +1389,9 @@ static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
            nla_put_u8(skb, IFLA_PROTO_DOWN, dev->proto_down))
                goto nla_put_failure;
 
+       if (rtnl_fill_link_event(skb, event))
+               goto nla_put_failure;
+
        if (rtnl_fill_link_ifmap(skb, dev))
                goto nla_put_failure;
 
@@ -1461,6 +1526,7 @@ static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
        [IFLA_LINK_NETNSID]     = { .type = NLA_S32 },
        [IFLA_PROTO_DOWN]       = { .type = NLA_U8 },
        [IFLA_XDP]              = { .type = NLA_NESTED },
+       [IFLA_EVENT]            = { .type = NLA_U32 },
 };
 
 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
@@ -1619,7 +1685,7 @@ static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
                                               NETLINK_CB(cb->skb).portid,
                                               cb->nlh->nlmsg_seq, 0,
                                               flags,
-                                              ext_filter_mask);
+                                              ext_filter_mask, 0);
                        /* If we ran out of room on the first message,
                         * we're in trouble
                         */
@@ -2710,7 +2776,7 @@ static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh)
                return -ENOBUFS;
 
        err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid,
-                              nlh->nlmsg_seq, 0, 0, ext_filter_mask);
+                              nlh->nlmsg_seq, 0, 0, ext_filter_mask, 0);
        if (err < 0) {
                /* -EMSGSIZE implies BUG in if_nlmsg_size */
                WARN_ON(err == -EMSGSIZE);
@@ -2782,7 +2848,8 @@ static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
 }
 
 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
-                                      unsigned int change, gfp_t flags)
+                                      unsigned int change,
+                                      unsigned long event, gfp_t flags)
 {
        struct net *net = dev_net(dev);
        struct sk_buff *skb;
@@ -2793,7 +2860,7 @@ struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
        if (skb == NULL)
                goto errout;
 
-       err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
+       err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0, event);
        if (err < 0) {
                /* -EMSGSIZE implies BUG in if_nlmsg_size() */
                WARN_ON(err == -EMSGSIZE);
@@ -2814,18 +2881,25 @@ void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
        rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
 }
 
-void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
-                 gfp_t flags)
+static void rtmsg_ifinfo_event(int type, struct net_device *dev,
+                              unsigned int change, unsigned long event,
+                              gfp_t flags)
 {
        struct sk_buff *skb;
 
        if (dev->reg_state != NETREG_REGISTERED)
                return;
 
-       skb = rtmsg_ifinfo_build_skb(type, dev, change, flags);
+       skb = rtmsg_ifinfo_build_skb(type, dev, change, event, flags);
        if (skb)
                rtmsg_ifinfo_send(skb, dev, flags);
 }
+
+void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
+                 gfp_t flags)
+{
+       rtmsg_ifinfo_event(type, dev, change, 0, flags);
+}
 EXPORT_SYMBOL(rtmsg_ifinfo);
 
 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
@@ -4132,7 +4206,7 @@ static int rtnetlink_event(struct notifier_block *this, unsigned long event, voi
        case NETDEV_CHANGELOWERSTATE:
        case NETDEV_UDP_TUNNEL_PUSH_INFO:
        case NETDEV_CHANGE_TX_QUEUE_LEN:
-               rtmsg_ifinfo(RTM_NEWLINK, dev, 0, GFP_KERNEL);
+               rtmsg_ifinfo_event(RTM_NEWLINK, dev, 0, event, GFP_KERNEL);
                break;
        default:
                break;