2 * Copyright (c) 2007-2013 Nicira, Inc.
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of version 2 of the GNU General Public
6 * License as published by the Free Software Foundation.
8 * This program is distributed in the hope that it will be useful, but
9 * WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
13 * You should have received a copy of the GNU General Public License
14 * along with this program; if not, write to the Free Software
15 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
19 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
21 #include <linux/init.h>
22 #include <linux/module.h>
23 #include <linux/if_arp.h>
24 #include <linux/if_vlan.h>
27 #include <linux/jhash.h>
28 #include <linux/delay.h>
29 #include <linux/time.h>
30 #include <linux/etherdevice.h>
31 #include <linux/genetlink.h>
32 #include <linux/kernel.h>
33 #include <linux/kthread.h>
34 #include <linux/mutex.h>
35 #include <linux/percpu.h>
36 #include <linux/rcupdate.h>
37 #include <linux/tcp.h>
38 #include <linux/udp.h>
39 #include <linux/version.h>
40 #include <linux/ethtool.h>
41 #include <linux/wait.h>
42 #include <asm/div64.h>
43 #include <linux/highmem.h>
44 #include <linux/netfilter_bridge.h>
45 #include <linux/netfilter_ipv4.h>
46 #include <linux/inetdevice.h>
47 #include <linux/list.h>
48 #include <linux/openvswitch.h>
49 #include <linux/rculist.h>
50 #include <linux/dmi.h>
51 #include <linux/genetlink.h>
52 #include <net/genetlink.h>
53 #include <net/genetlink.h>
54 #include <net/net_namespace.h>
55 #include <net/netns/generic.h>
59 #include "flow_netlink.h"
61 #include "vport-internal_dev.h"
62 #include "vport-netdev.h"
64 int ovs_net_id __read_mostly;
66 static void ovs_notify(struct sk_buff *skb, struct genl_info *info,
67 struct genl_multicast_group *grp)
69 genl_notify(skb, genl_info_net(info), info->snd_portid,
70 grp->id, info->nlhdr, GFP_KERNEL);
76 * All writes e.g. Writes to device state (add/remove datapath, port, set
77 * operations on vports, etc.), Writes to other state (flow table
78 * modifications, set miscellaneous datapath parameters, etc.) are protected
81 * Reads are protected by RCU.
83 * There are a few special cases (mostly stats) that have their own
84 * synchronization but they nest under all of above and don't interact with
87 * The RTNL lock nests inside ovs_mutex.
90 static DEFINE_MUTEX(ovs_mutex);
94 mutex_lock(&ovs_mutex);
99 mutex_unlock(&ovs_mutex);
102 #ifdef CONFIG_LOCKDEP
103 int lockdep_ovsl_is_held(void)
106 return lockdep_is_held(&ovs_mutex);
112 static struct vport *new_vport(const struct vport_parms *);
113 static int queue_gso_packets(struct net *, int dp_ifindex, struct sk_buff *,
114 const struct dp_upcall_info *);
115 static int queue_userspace_packet(struct net *, int dp_ifindex,
117 const struct dp_upcall_info *);
119 /* Must be called with rcu_read_lock or ovs_mutex. */
120 static struct datapath *get_dp(struct net *net, int dp_ifindex)
122 struct datapath *dp = NULL;
123 struct net_device *dev;
126 dev = dev_get_by_index_rcu(net, dp_ifindex);
128 struct vport *vport = ovs_internal_dev_get_vport(dev);
137 /* Must be called with rcu_read_lock or ovs_mutex. */
138 const char *ovs_dp_name(const struct datapath *dp)
140 struct vport *vport = ovs_vport_ovsl_rcu(dp, OVSP_LOCAL);
141 return vport->ops->get_name(vport);
144 static int get_dpifindex(struct datapath *dp)
151 local = ovs_vport_rcu(dp, OVSP_LOCAL);
153 ifindex = netdev_vport_priv(local)->dev->ifindex;
162 static void destroy_dp_rcu(struct rcu_head *rcu)
164 struct datapath *dp = container_of(rcu, struct datapath, rcu);
166 ovs_flow_tbl_destroy(&dp->table);
167 free_percpu(dp->stats_percpu);
168 release_net(ovs_dp_get_net(dp));
173 static struct hlist_head *vport_hash_bucket(const struct datapath *dp,
176 return &dp->ports[port_no & (DP_VPORT_HASH_BUCKETS - 1)];
179 struct vport *ovs_lookup_vport(const struct datapath *dp, u16 port_no)
182 struct hlist_head *head;
184 head = vport_hash_bucket(dp, port_no);
185 hlist_for_each_entry_rcu(vport, head, dp_hash_node) {
186 if (vport->port_no == port_no)
192 /* Called with ovs_mutex. */
193 static struct vport *new_vport(const struct vport_parms *parms)
197 vport = ovs_vport_add(parms);
198 if (!IS_ERR(vport)) {
199 struct datapath *dp = parms->dp;
200 struct hlist_head *head = vport_hash_bucket(dp, vport->port_no);
202 hlist_add_head_rcu(&vport->dp_hash_node, head);
207 void ovs_dp_detach_port(struct vport *p)
211 /* First drop references to device. */
212 hlist_del_rcu(&p->dp_hash_node);
214 /* Then destroy it. */
218 /* Must be called with rcu_read_lock. */
219 void ovs_dp_process_received_packet(struct vport *p, struct sk_buff *skb)
221 struct datapath *dp = p->dp;
222 struct sw_flow *flow;
223 struct dp_stats_percpu *stats;
224 struct sw_flow_key key;
229 stats = this_cpu_ptr(dp->stats_percpu);
231 /* Extract flow from 'skb' into 'key'. */
232 error = ovs_flow_extract(skb, p->port_no, &key);
233 if (unlikely(error)) {
239 flow = ovs_flow_tbl_lookup(&dp->table, &key, &n_mask_hit);
240 if (unlikely(!flow)) {
241 struct dp_upcall_info upcall;
243 upcall.cmd = OVS_PACKET_CMD_MISS;
245 upcall.userdata = NULL;
246 upcall.portid = p->upcall_portid;
247 ovs_dp_upcall(dp, skb, &upcall);
249 stats_counter = &stats->n_missed;
253 OVS_CB(skb)->flow = flow;
254 OVS_CB(skb)->pkt_key = &key;
256 stats_counter = &stats->n_hit;
257 ovs_flow_used(OVS_CB(skb)->flow, skb);
258 ovs_execute_actions(dp, skb);
261 /* Update datapath statistics. */
262 u64_stats_update_begin(&stats->sync);
264 stats->n_mask_hit += n_mask_hit;
265 u64_stats_update_end(&stats->sync);
268 static struct genl_family dp_packet_genl_family = {
269 .id = GENL_ID_GENERATE,
270 .hdrsize = sizeof(struct ovs_header),
271 .name = OVS_PACKET_FAMILY,
272 .version = OVS_PACKET_VERSION,
273 .maxattr = OVS_PACKET_ATTR_MAX,
278 int ovs_dp_upcall(struct datapath *dp, struct sk_buff *skb,
279 const struct dp_upcall_info *upcall_info)
281 struct dp_stats_percpu *stats;
285 if (upcall_info->portid == 0) {
290 dp_ifindex = get_dpifindex(dp);
296 if (!skb_is_gso(skb))
297 err = queue_userspace_packet(ovs_dp_get_net(dp), dp_ifindex, skb, upcall_info);
299 err = queue_gso_packets(ovs_dp_get_net(dp), dp_ifindex, skb, upcall_info);
306 stats = this_cpu_ptr(dp->stats_percpu);
308 u64_stats_update_begin(&stats->sync);
310 u64_stats_update_end(&stats->sync);
315 static int queue_gso_packets(struct net *net, int dp_ifindex,
317 const struct dp_upcall_info *upcall_info)
319 unsigned short gso_type = skb_shinfo(skb)->gso_type;
320 struct dp_upcall_info later_info;
321 struct sw_flow_key later_key;
322 struct sk_buff *segs, *nskb;
325 segs = __skb_gso_segment(skb, NETIF_F_SG | NETIF_F_HW_CSUM, false);
327 return PTR_ERR(segs);
329 /* Queue all of the segments. */
332 err = queue_userspace_packet(net, dp_ifindex, skb, upcall_info);
336 if (skb == segs && gso_type & SKB_GSO_UDP) {
337 /* The initial flow key extracted by ovs_flow_extract()
338 * in this case is for a first fragment, so we need to
339 * properly mark later fragments.
341 later_key = *upcall_info->key;
342 later_key.ip.frag = OVS_FRAG_TYPE_LATER;
344 later_info = *upcall_info;
345 later_info.key = &later_key;
346 upcall_info = &later_info;
348 } while ((skb = skb->next));
350 /* Free all of the segments. */
358 } while ((skb = nskb));
362 static size_t key_attr_size(void)
364 return nla_total_size(4) /* OVS_KEY_ATTR_PRIORITY */
365 + nla_total_size(0) /* OVS_KEY_ATTR_TUNNEL */
366 + nla_total_size(8) /* OVS_TUNNEL_KEY_ATTR_ID */
367 + nla_total_size(4) /* OVS_TUNNEL_KEY_ATTR_IPV4_SRC */
368 + nla_total_size(4) /* OVS_TUNNEL_KEY_ATTR_IPV4_DST */
369 + nla_total_size(1) /* OVS_TUNNEL_KEY_ATTR_TOS */
370 + nla_total_size(1) /* OVS_TUNNEL_KEY_ATTR_TTL */
371 + nla_total_size(0) /* OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT */
372 + nla_total_size(0) /* OVS_TUNNEL_KEY_ATTR_CSUM */
373 + nla_total_size(4) /* OVS_KEY_ATTR_IN_PORT */
374 + nla_total_size(4) /* OVS_KEY_ATTR_SKB_MARK */
375 + nla_total_size(12) /* OVS_KEY_ATTR_ETHERNET */
376 + nla_total_size(2) /* OVS_KEY_ATTR_ETHERTYPE */
377 + nla_total_size(4) /* OVS_KEY_ATTR_8021Q */
378 + nla_total_size(0) /* OVS_KEY_ATTR_ENCAP */
379 + nla_total_size(2) /* OVS_KEY_ATTR_ETHERTYPE */
380 + nla_total_size(40) /* OVS_KEY_ATTR_IPV6 */
381 + nla_total_size(2) /* OVS_KEY_ATTR_ICMPV6 */
382 + nla_total_size(28); /* OVS_KEY_ATTR_ND */
385 static size_t upcall_msg_size(const struct sk_buff *skb,
386 const struct nlattr *userdata)
388 size_t size = NLMSG_ALIGN(sizeof(struct ovs_header))
389 + nla_total_size(skb->len) /* OVS_PACKET_ATTR_PACKET */
390 + nla_total_size(key_attr_size()); /* OVS_PACKET_ATTR_KEY */
392 /* OVS_PACKET_ATTR_USERDATA */
394 size += NLA_ALIGN(userdata->nla_len);
399 static int queue_userspace_packet(struct net *net, int dp_ifindex,
401 const struct dp_upcall_info *upcall_info)
403 struct ovs_header *upcall;
404 struct sk_buff *nskb = NULL;
405 struct sk_buff *user_skb; /* to be queued to userspace */
409 if (vlan_tx_tag_present(skb)) {
410 nskb = skb_clone(skb, GFP_ATOMIC);
414 nskb = __vlan_put_tag(nskb, nskb->vlan_proto, vlan_tx_tag_get(nskb));
418 vlan_set_tci(nskb, 0);
423 if (nla_attr_size(skb->len) > USHRT_MAX) {
428 user_skb = genlmsg_new(upcall_msg_size(skb, upcall_info->userdata), GFP_ATOMIC);
434 upcall = genlmsg_put(user_skb, 0, 0, &dp_packet_genl_family,
435 0, upcall_info->cmd);
436 upcall->dp_ifindex = dp_ifindex;
438 nla = nla_nest_start(user_skb, OVS_PACKET_ATTR_KEY);
439 ovs_nla_put_flow(upcall_info->key, upcall_info->key, user_skb);
440 nla_nest_end(user_skb, nla);
442 if (upcall_info->userdata)
443 __nla_put(user_skb, OVS_PACKET_ATTR_USERDATA,
444 nla_len(upcall_info->userdata),
445 nla_data(upcall_info->userdata));
447 nla = __nla_reserve(user_skb, OVS_PACKET_ATTR_PACKET, skb->len);
449 skb_copy_and_csum_dev(skb, nla_data(nla));
451 genlmsg_end(user_skb, upcall);
452 err = genlmsg_unicast(net, user_skb, upcall_info->portid);
459 static void clear_stats(struct sw_flow *flow)
463 flow->packet_count = 0;
464 flow->byte_count = 0;
467 static int ovs_packet_cmd_execute(struct sk_buff *skb, struct genl_info *info)
469 struct ovs_header *ovs_header = info->userhdr;
470 struct nlattr **a = info->attrs;
471 struct sw_flow_actions *acts;
472 struct sk_buff *packet;
473 struct sw_flow *flow;
480 if (!a[OVS_PACKET_ATTR_PACKET] || !a[OVS_PACKET_ATTR_KEY] ||
481 !a[OVS_PACKET_ATTR_ACTIONS])
484 len = nla_len(a[OVS_PACKET_ATTR_PACKET]);
485 packet = __dev_alloc_skb(NET_IP_ALIGN + len, GFP_KERNEL);
489 skb_reserve(packet, NET_IP_ALIGN);
491 nla_memcpy(__skb_put(packet, len), a[OVS_PACKET_ATTR_PACKET], len);
493 skb_reset_mac_header(packet);
494 eth = eth_hdr(packet);
496 /* Normally, setting the skb 'protocol' field would be handled by a
497 * call to eth_type_trans(), but it assumes there's a sending
498 * device, which we may not have. */
499 if (ntohs(eth->h_proto) >= ETH_P_802_3_MIN)
500 packet->protocol = eth->h_proto;
502 packet->protocol = htons(ETH_P_802_2);
504 /* Build an sw_flow for sending this packet. */
505 flow = ovs_flow_alloc();
510 err = ovs_flow_extract(packet, -1, &flow->key);
514 err = ovs_nla_get_flow_metadata(flow, a[OVS_PACKET_ATTR_KEY]);
517 acts = ovs_nla_alloc_flow_actions(nla_len(a[OVS_PACKET_ATTR_ACTIONS]));
522 err = ovs_nla_copy_actions(a[OVS_PACKET_ATTR_ACTIONS],
523 &flow->key, 0, &acts);
524 rcu_assign_pointer(flow->sf_acts, acts);
528 OVS_CB(packet)->flow = flow;
529 OVS_CB(packet)->pkt_key = &flow->key;
530 packet->priority = flow->key.phy.priority;
531 packet->mark = flow->key.phy.skb_mark;
534 dp = get_dp(sock_net(skb->sk), ovs_header->dp_ifindex);
540 err = ovs_execute_actions(dp, packet);
544 ovs_flow_free(flow, false);
550 ovs_flow_free(flow, false);
557 static const struct nla_policy packet_policy[OVS_PACKET_ATTR_MAX + 1] = {
558 [OVS_PACKET_ATTR_PACKET] = { .len = ETH_HLEN },
559 [OVS_PACKET_ATTR_KEY] = { .type = NLA_NESTED },
560 [OVS_PACKET_ATTR_ACTIONS] = { .type = NLA_NESTED },
563 static struct genl_ops dp_packet_genl_ops[] = {
564 { .cmd = OVS_PACKET_CMD_EXECUTE,
565 .flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
566 .policy = packet_policy,
567 .doit = ovs_packet_cmd_execute
571 static void get_dp_stats(struct datapath *dp, struct ovs_dp_stats *stats,
572 struct ovs_dp_megaflow_stats *mega_stats)
576 memset(mega_stats, 0, sizeof(*mega_stats));
578 stats->n_flows = ovs_flow_tbl_count(&dp->table);
579 mega_stats->n_masks = ovs_flow_tbl_num_masks(&dp->table);
581 stats->n_hit = stats->n_missed = stats->n_lost = 0;
583 for_each_possible_cpu(i) {
584 const struct dp_stats_percpu *percpu_stats;
585 struct dp_stats_percpu local_stats;
588 percpu_stats = per_cpu_ptr(dp->stats_percpu, i);
591 start = u64_stats_fetch_begin_bh(&percpu_stats->sync);
592 local_stats = *percpu_stats;
593 } while (u64_stats_fetch_retry_bh(&percpu_stats->sync, start));
595 stats->n_hit += local_stats.n_hit;
596 stats->n_missed += local_stats.n_missed;
597 stats->n_lost += local_stats.n_lost;
598 mega_stats->n_mask_hit += local_stats.n_mask_hit;
602 static const struct nla_policy flow_policy[OVS_FLOW_ATTR_MAX + 1] = {
603 [OVS_FLOW_ATTR_KEY] = { .type = NLA_NESTED },
604 [OVS_FLOW_ATTR_ACTIONS] = { .type = NLA_NESTED },
605 [OVS_FLOW_ATTR_CLEAR] = { .type = NLA_FLAG },
608 static struct genl_family dp_flow_genl_family = {
609 .id = GENL_ID_GENERATE,
610 .hdrsize = sizeof(struct ovs_header),
611 .name = OVS_FLOW_FAMILY,
612 .version = OVS_FLOW_VERSION,
613 .maxattr = OVS_FLOW_ATTR_MAX,
618 static struct genl_multicast_group ovs_dp_flow_multicast_group = {
619 .name = OVS_FLOW_MCGROUP
622 static size_t ovs_flow_cmd_msg_size(const struct sw_flow_actions *acts)
624 return NLMSG_ALIGN(sizeof(struct ovs_header))
625 + nla_total_size(key_attr_size()) /* OVS_FLOW_ATTR_KEY */
626 + nla_total_size(key_attr_size()) /* OVS_FLOW_ATTR_MASK */
627 + nla_total_size(sizeof(struct ovs_flow_stats)) /* OVS_FLOW_ATTR_STATS */
628 + nla_total_size(1) /* OVS_FLOW_ATTR_TCP_FLAGS */
629 + nla_total_size(8) /* OVS_FLOW_ATTR_USED */
630 + nla_total_size(acts->actions_len); /* OVS_FLOW_ATTR_ACTIONS */
633 /* Called with ovs_mutex. */
634 static int ovs_flow_cmd_fill_info(struct sw_flow *flow, struct datapath *dp,
635 struct sk_buff *skb, u32 portid,
636 u32 seq, u32 flags, u8 cmd)
638 const int skb_orig_len = skb->len;
639 struct nlattr *start;
640 struct ovs_flow_stats stats;
641 struct ovs_header *ovs_header;
647 ovs_header = genlmsg_put(skb, portid, seq, &dp_flow_genl_family, flags, cmd);
651 ovs_header->dp_ifindex = get_dpifindex(dp);
654 nla = nla_nest_start(skb, OVS_FLOW_ATTR_KEY);
656 goto nla_put_failure;
658 err = ovs_nla_put_flow(&flow->unmasked_key, &flow->unmasked_key, skb);
661 nla_nest_end(skb, nla);
663 nla = nla_nest_start(skb, OVS_FLOW_ATTR_MASK);
665 goto nla_put_failure;
667 err = ovs_nla_put_flow(&flow->key, &flow->mask->key, skb);
671 nla_nest_end(skb, nla);
673 spin_lock_bh(&flow->lock);
675 stats.n_packets = flow->packet_count;
676 stats.n_bytes = flow->byte_count;
677 tcp_flags = flow->tcp_flags;
678 spin_unlock_bh(&flow->lock);
681 nla_put_u64(skb, OVS_FLOW_ATTR_USED, ovs_flow_used_time(used)))
682 goto nla_put_failure;
684 if (stats.n_packets &&
685 nla_put(skb, OVS_FLOW_ATTR_STATS,
686 sizeof(struct ovs_flow_stats), &stats))
687 goto nla_put_failure;
690 nla_put_u8(skb, OVS_FLOW_ATTR_TCP_FLAGS, tcp_flags))
691 goto nla_put_failure;
693 /* If OVS_FLOW_ATTR_ACTIONS doesn't fit, skip dumping the actions if
694 * this is the first flow to be dumped into 'skb'. This is unusual for
695 * Netlink but individual action lists can be longer than
696 * NLMSG_GOODSIZE and thus entirely undumpable if we didn't do this.
697 * The userspace caller can always fetch the actions separately if it
698 * really wants them. (Most userspace callers in fact don't care.)
700 * This can only fail for dump operations because the skb is always
701 * properly sized for single flows.
703 start = nla_nest_start(skb, OVS_FLOW_ATTR_ACTIONS);
705 const struct sw_flow_actions *sf_acts;
707 sf_acts = rcu_dereference_check(flow->sf_acts,
708 lockdep_ovsl_is_held());
710 err = ovs_nla_put_actions(sf_acts->actions,
711 sf_acts->actions_len, skb);
713 nla_nest_end(skb, start);
718 nla_nest_cancel(skb, start);
720 } else if (skb_orig_len)
721 goto nla_put_failure;
723 return genlmsg_end(skb, ovs_header);
728 genlmsg_cancel(skb, ovs_header);
732 static struct sk_buff *ovs_flow_cmd_alloc_info(struct sw_flow *flow)
734 const struct sw_flow_actions *sf_acts;
736 sf_acts = ovsl_dereference(flow->sf_acts);
738 return genlmsg_new(ovs_flow_cmd_msg_size(sf_acts), GFP_KERNEL);
741 static struct sk_buff *ovs_flow_cmd_build_info(struct sw_flow *flow,
743 u32 portid, u32 seq, u8 cmd)
748 skb = ovs_flow_cmd_alloc_info(flow);
750 return ERR_PTR(-ENOMEM);
752 retval = ovs_flow_cmd_fill_info(flow, dp, skb, portid, seq, 0, cmd);
757 static struct sw_flow *__ovs_flow_tbl_lookup(struct flow_table *tbl,
758 const struct sw_flow_key *key)
760 u32 __always_unused n_mask_hit;
762 return ovs_flow_tbl_lookup(tbl, key, &n_mask_hit);
765 static int ovs_flow_cmd_new_or_set(struct sk_buff *skb, struct genl_info *info)
767 struct nlattr **a = info->attrs;
768 struct ovs_header *ovs_header = info->userhdr;
769 struct sw_flow_key key, masked_key;
770 struct sw_flow *flow = NULL;
771 struct sw_flow_mask mask;
772 struct sk_buff *reply;
774 struct sw_flow_actions *acts = NULL;
775 struct sw_flow_match match;
780 if (!a[OVS_FLOW_ATTR_KEY])
783 ovs_match_init(&match, &key, &mask);
784 error = ovs_nla_get_match(&match,
785 a[OVS_FLOW_ATTR_KEY], a[OVS_FLOW_ATTR_MASK]);
789 /* Validate actions. */
790 if (a[OVS_FLOW_ATTR_ACTIONS]) {
791 acts = ovs_nla_alloc_flow_actions(nla_len(a[OVS_FLOW_ATTR_ACTIONS]));
792 error = PTR_ERR(acts);
796 ovs_flow_mask_key(&masked_key, &key, &mask);
797 error = ovs_nla_copy_actions(a[OVS_FLOW_ATTR_ACTIONS],
798 &masked_key, 0, &acts);
800 OVS_NLERR("Flow actions may not be safe on all matching packets.\n");
803 } else if (info->genlhdr->cmd == OVS_FLOW_CMD_NEW) {
809 dp = get_dp(sock_net(skb->sk), ovs_header->dp_ifindex);
814 /* Check if this is a duplicate flow */
815 flow = __ovs_flow_tbl_lookup(&dp->table, &key);
817 /* Bail out if we're not allowed to create a new flow. */
819 if (info->genlhdr->cmd == OVS_FLOW_CMD_SET)
823 flow = ovs_flow_alloc();
825 error = PTR_ERR(flow);
830 flow->key = masked_key;
831 flow->unmasked_key = key;
832 rcu_assign_pointer(flow->sf_acts, acts);
834 /* Put flow in bucket. */
835 error = ovs_flow_tbl_insert(&dp->table, flow, &mask);
841 reply = ovs_flow_cmd_build_info(flow, dp, info->snd_portid,
842 info->snd_seq, OVS_FLOW_CMD_NEW);
844 /* We found a matching flow. */
845 struct sw_flow_actions *old_acts;
847 /* Bail out if we're not allowed to modify an existing flow.
848 * We accept NLM_F_CREATE in place of the intended NLM_F_EXCL
849 * because Generic Netlink treats the latter as a dump
850 * request. We also accept NLM_F_EXCL in case that bug ever
854 if (info->genlhdr->cmd == OVS_FLOW_CMD_NEW &&
855 info->nlhdr->nlmsg_flags & (NLM_F_CREATE | NLM_F_EXCL))
858 /* The unmasked key has to be the same for flow updates. */
860 if (!ovs_flow_cmp_unmasked_key(flow, &match)) {
861 OVS_NLERR("Flow modification message rejected, unmasked key does not match.\n");
865 /* Update actions. */
866 old_acts = ovsl_dereference(flow->sf_acts);
867 rcu_assign_pointer(flow->sf_acts, acts);
868 ovs_nla_free_flow_actions(old_acts);
870 reply = ovs_flow_cmd_build_info(flow, dp, info->snd_portid,
871 info->snd_seq, OVS_FLOW_CMD_NEW);
874 if (a[OVS_FLOW_ATTR_CLEAR]) {
875 spin_lock_bh(&flow->lock);
877 spin_unlock_bh(&flow->lock);
883 ovs_notify(reply, info, &ovs_dp_flow_multicast_group);
885 netlink_set_err(sock_net(skb->sk)->genl_sock, 0,
886 ovs_dp_flow_multicast_group.id, PTR_ERR(reply));
890 ovs_flow_free(flow, false);
899 static int ovs_flow_cmd_get(struct sk_buff *skb, struct genl_info *info)
901 struct nlattr **a = info->attrs;
902 struct ovs_header *ovs_header = info->userhdr;
903 struct sw_flow_key key;
904 struct sk_buff *reply;
905 struct sw_flow *flow;
907 struct sw_flow_match match;
910 if (!a[OVS_FLOW_ATTR_KEY]) {
911 OVS_NLERR("Flow get message rejected, Key attribute missing.\n");
915 ovs_match_init(&match, &key, NULL);
916 err = ovs_nla_get_match(&match, a[OVS_FLOW_ATTR_KEY], NULL);
921 dp = get_dp(sock_net(skb->sk), ovs_header->dp_ifindex);
927 flow = __ovs_flow_tbl_lookup(&dp->table, &key);
928 if (!flow || !ovs_flow_cmp_unmasked_key(flow, &match)) {
933 reply = ovs_flow_cmd_build_info(flow, dp, info->snd_portid,
934 info->snd_seq, OVS_FLOW_CMD_NEW);
936 err = PTR_ERR(reply);
941 return genlmsg_reply(reply, info);
947 static int ovs_flow_cmd_del(struct sk_buff *skb, struct genl_info *info)
949 struct nlattr **a = info->attrs;
950 struct ovs_header *ovs_header = info->userhdr;
951 struct sw_flow_key key;
952 struct sk_buff *reply;
953 struct sw_flow *flow;
955 struct sw_flow_match match;
959 dp = get_dp(sock_net(skb->sk), ovs_header->dp_ifindex);
965 if (!a[OVS_FLOW_ATTR_KEY]) {
966 err = ovs_flow_tbl_flush(&dp->table);
970 ovs_match_init(&match, &key, NULL);
971 err = ovs_nla_get_match(&match, a[OVS_FLOW_ATTR_KEY], NULL);
975 flow = __ovs_flow_tbl_lookup(&dp->table, &key);
976 if (!flow || !ovs_flow_cmp_unmasked_key(flow, &match)) {
981 reply = ovs_flow_cmd_alloc_info(flow);
987 ovs_flow_tbl_remove(&dp->table, flow);
989 err = ovs_flow_cmd_fill_info(flow, dp, reply, info->snd_portid,
990 info->snd_seq, 0, OVS_FLOW_CMD_DEL);
993 ovs_flow_free(flow, true);
996 ovs_notify(reply, info, &ovs_dp_flow_multicast_group);
1003 static int ovs_flow_cmd_dump(struct sk_buff *skb, struct netlink_callback *cb)
1005 struct ovs_header *ovs_header = genlmsg_data(nlmsg_data(cb->nlh));
1006 struct table_instance *ti;
1007 struct datapath *dp;
1010 dp = get_dp(sock_net(skb->sk), ovs_header->dp_ifindex);
1016 ti = rcu_dereference(dp->table.ti);
1018 struct sw_flow *flow;
1021 bucket = cb->args[0];
1023 flow = ovs_flow_tbl_dump_next(ti, &bucket, &obj);
1027 if (ovs_flow_cmd_fill_info(flow, dp, skb,
1028 NETLINK_CB(cb->skb).portid,
1029 cb->nlh->nlmsg_seq, NLM_F_MULTI,
1030 OVS_FLOW_CMD_NEW) < 0)
1033 cb->args[0] = bucket;
1040 static struct genl_ops dp_flow_genl_ops[] = {
1041 { .cmd = OVS_FLOW_CMD_NEW,
1042 .flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
1043 .policy = flow_policy,
1044 .doit = ovs_flow_cmd_new_or_set
1046 { .cmd = OVS_FLOW_CMD_DEL,
1047 .flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
1048 .policy = flow_policy,
1049 .doit = ovs_flow_cmd_del
1051 { .cmd = OVS_FLOW_CMD_GET,
1052 .flags = 0, /* OK for unprivileged users. */
1053 .policy = flow_policy,
1054 .doit = ovs_flow_cmd_get,
1055 .dumpit = ovs_flow_cmd_dump
1057 { .cmd = OVS_FLOW_CMD_SET,
1058 .flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
1059 .policy = flow_policy,
1060 .doit = ovs_flow_cmd_new_or_set,
1064 static const struct nla_policy datapath_policy[OVS_DP_ATTR_MAX + 1] = {
1065 [OVS_DP_ATTR_NAME] = { .type = NLA_NUL_STRING, .len = IFNAMSIZ - 1 },
1066 [OVS_DP_ATTR_UPCALL_PID] = { .type = NLA_U32 },
1069 static struct genl_family dp_datapath_genl_family = {
1070 .id = GENL_ID_GENERATE,
1071 .hdrsize = sizeof(struct ovs_header),
1072 .name = OVS_DATAPATH_FAMILY,
1073 .version = OVS_DATAPATH_VERSION,
1074 .maxattr = OVS_DP_ATTR_MAX,
1079 static struct genl_multicast_group ovs_dp_datapath_multicast_group = {
1080 .name = OVS_DATAPATH_MCGROUP
1083 static size_t ovs_dp_cmd_msg_size(void)
1085 size_t msgsize = NLMSG_ALIGN(sizeof(struct ovs_header));
1087 msgsize += nla_total_size(IFNAMSIZ);
1088 msgsize += nla_total_size(sizeof(struct ovs_dp_stats));
1089 msgsize += nla_total_size(sizeof(struct ovs_dp_megaflow_stats));
1094 static int ovs_dp_cmd_fill_info(struct datapath *dp, struct sk_buff *skb,
1095 u32 portid, u32 seq, u32 flags, u8 cmd)
1097 struct ovs_header *ovs_header;
1098 struct ovs_dp_stats dp_stats;
1099 struct ovs_dp_megaflow_stats dp_megaflow_stats;
1102 ovs_header = genlmsg_put(skb, portid, seq, &dp_datapath_genl_family,
1107 ovs_header->dp_ifindex = get_dpifindex(dp);
1110 err = nla_put_string(skb, OVS_DP_ATTR_NAME, ovs_dp_name(dp));
1113 goto nla_put_failure;
1115 get_dp_stats(dp, &dp_stats, &dp_megaflow_stats);
1116 if (nla_put(skb, OVS_DP_ATTR_STATS, sizeof(struct ovs_dp_stats),
1118 goto nla_put_failure;
1120 if (nla_put(skb, OVS_DP_ATTR_MEGAFLOW_STATS,
1121 sizeof(struct ovs_dp_megaflow_stats),
1122 &dp_megaflow_stats))
1123 goto nla_put_failure;
1125 return genlmsg_end(skb, ovs_header);
1128 genlmsg_cancel(skb, ovs_header);
1133 static struct sk_buff *ovs_dp_cmd_build_info(struct datapath *dp, u32 portid,
1136 struct sk_buff *skb;
1139 skb = genlmsg_new(ovs_dp_cmd_msg_size(), GFP_KERNEL);
1141 return ERR_PTR(-ENOMEM);
1143 retval = ovs_dp_cmd_fill_info(dp, skb, portid, seq, 0, cmd);
1146 return ERR_PTR(retval);
1151 /* Called with ovs_mutex. */
1152 static struct datapath *lookup_datapath(struct net *net,
1153 struct ovs_header *ovs_header,
1154 struct nlattr *a[OVS_DP_ATTR_MAX + 1])
1156 struct datapath *dp;
1158 if (!a[OVS_DP_ATTR_NAME])
1159 dp = get_dp(net, ovs_header->dp_ifindex);
1161 struct vport *vport;
1164 vport = ovs_vport_locate(net, nla_data(a[OVS_DP_ATTR_NAME]));
1165 dp = vport && vport->port_no == OVSP_LOCAL ? vport->dp : NULL;
1168 return dp ? dp : ERR_PTR(-ENODEV);
1171 static int ovs_dp_cmd_new(struct sk_buff *skb, struct genl_info *info)
1173 struct nlattr **a = info->attrs;
1174 struct vport_parms parms;
1175 struct sk_buff *reply;
1176 struct datapath *dp;
1177 struct vport *vport;
1178 struct ovs_net *ovs_net;
1182 if (!a[OVS_DP_ATTR_NAME] || !a[OVS_DP_ATTR_UPCALL_PID])
1188 dp = kzalloc(sizeof(*dp), GFP_KERNEL);
1190 goto err_unlock_ovs;
1192 ovs_dp_set_net(dp, hold_net(sock_net(skb->sk)));
1194 /* Allocate table. */
1195 err = ovs_flow_tbl_init(&dp->table);
1199 dp->stats_percpu = alloc_percpu(struct dp_stats_percpu);
1200 if (!dp->stats_percpu) {
1202 goto err_destroy_table;
1205 dp->ports = kmalloc(DP_VPORT_HASH_BUCKETS * sizeof(struct hlist_head),
1209 goto err_destroy_percpu;
1212 for (i = 0; i < DP_VPORT_HASH_BUCKETS; i++)
1213 INIT_HLIST_HEAD(&dp->ports[i]);
1215 /* Set up our datapath device. */
1216 parms.name = nla_data(a[OVS_DP_ATTR_NAME]);
1217 parms.type = OVS_VPORT_TYPE_INTERNAL;
1218 parms.options = NULL;
1220 parms.port_no = OVSP_LOCAL;
1221 parms.upcall_portid = nla_get_u32(a[OVS_DP_ATTR_UPCALL_PID]);
1223 vport = new_vport(&parms);
1224 if (IS_ERR(vport)) {
1225 err = PTR_ERR(vport);
1229 goto err_destroy_ports_array;
1232 reply = ovs_dp_cmd_build_info(dp, info->snd_portid,
1233 info->snd_seq, OVS_DP_CMD_NEW);
1234 err = PTR_ERR(reply);
1236 goto err_destroy_local_port;
1238 ovs_net = net_generic(ovs_dp_get_net(dp), ovs_net_id);
1239 list_add_tail_rcu(&dp->list_node, &ovs_net->dps);
1243 ovs_notify(reply, info, &ovs_dp_datapath_multicast_group);
1246 err_destroy_local_port:
1247 ovs_dp_detach_port(ovs_vport_ovsl(dp, OVSP_LOCAL));
1248 err_destroy_ports_array:
1251 free_percpu(dp->stats_percpu);
1253 ovs_flow_tbl_destroy(&dp->table);
1255 release_net(ovs_dp_get_net(dp));
1263 /* Called with ovs_mutex. */
1264 static void __dp_destroy(struct datapath *dp)
1268 for (i = 0; i < DP_VPORT_HASH_BUCKETS; i++) {
1269 struct vport *vport;
1270 struct hlist_node *n;
1272 hlist_for_each_entry_safe(vport, n, &dp->ports[i], dp_hash_node)
1273 if (vport->port_no != OVSP_LOCAL)
1274 ovs_dp_detach_port(vport);
1277 list_del_rcu(&dp->list_node);
1279 /* OVSP_LOCAL is datapath internal port. We need to make sure that
1280 * all port in datapath are destroyed first before freeing datapath.
1282 ovs_dp_detach_port(ovs_vport_ovsl(dp, OVSP_LOCAL));
1284 call_rcu(&dp->rcu, destroy_dp_rcu);
1287 static int ovs_dp_cmd_del(struct sk_buff *skb, struct genl_info *info)
1289 struct sk_buff *reply;
1290 struct datapath *dp;
1294 dp = lookup_datapath(sock_net(skb->sk), info->userhdr, info->attrs);
1299 reply = ovs_dp_cmd_build_info(dp, info->snd_portid,
1300 info->snd_seq, OVS_DP_CMD_DEL);
1301 err = PTR_ERR(reply);
1308 ovs_notify(reply, info, &ovs_dp_datapath_multicast_group);
1316 static int ovs_dp_cmd_set(struct sk_buff *skb, struct genl_info *info)
1318 struct sk_buff *reply;
1319 struct datapath *dp;
1323 dp = lookup_datapath(sock_net(skb->sk), info->userhdr, info->attrs);
1328 reply = ovs_dp_cmd_build_info(dp, info->snd_portid,
1329 info->snd_seq, OVS_DP_CMD_NEW);
1330 if (IS_ERR(reply)) {
1331 err = PTR_ERR(reply);
1332 netlink_set_err(sock_net(skb->sk)->genl_sock, 0,
1333 ovs_dp_datapath_multicast_group.id, err);
1339 ovs_notify(reply, info, &ovs_dp_datapath_multicast_group);
1347 static int ovs_dp_cmd_get(struct sk_buff *skb, struct genl_info *info)
1349 struct sk_buff *reply;
1350 struct datapath *dp;
1354 dp = lookup_datapath(sock_net(skb->sk), info->userhdr, info->attrs);
1360 reply = ovs_dp_cmd_build_info(dp, info->snd_portid,
1361 info->snd_seq, OVS_DP_CMD_NEW);
1362 if (IS_ERR(reply)) {
1363 err = PTR_ERR(reply);
1368 return genlmsg_reply(reply, info);
1375 static int ovs_dp_cmd_dump(struct sk_buff *skb, struct netlink_callback *cb)
1377 struct ovs_net *ovs_net = net_generic(sock_net(skb->sk), ovs_net_id);
1378 struct datapath *dp;
1379 int skip = cb->args[0];
1383 list_for_each_entry_rcu(dp, &ovs_net->dps, list_node) {
1385 ovs_dp_cmd_fill_info(dp, skb, NETLINK_CB(cb->skb).portid,
1386 cb->nlh->nlmsg_seq, NLM_F_MULTI,
1387 OVS_DP_CMD_NEW) < 0)
1398 static struct genl_ops dp_datapath_genl_ops[] = {
1399 { .cmd = OVS_DP_CMD_NEW,
1400 .flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
1401 .policy = datapath_policy,
1402 .doit = ovs_dp_cmd_new
1404 { .cmd = OVS_DP_CMD_DEL,
1405 .flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
1406 .policy = datapath_policy,
1407 .doit = ovs_dp_cmd_del
1409 { .cmd = OVS_DP_CMD_GET,
1410 .flags = 0, /* OK for unprivileged users. */
1411 .policy = datapath_policy,
1412 .doit = ovs_dp_cmd_get,
1413 .dumpit = ovs_dp_cmd_dump
1415 { .cmd = OVS_DP_CMD_SET,
1416 .flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
1417 .policy = datapath_policy,
1418 .doit = ovs_dp_cmd_set,
1422 static const struct nla_policy vport_policy[OVS_VPORT_ATTR_MAX + 1] = {
1423 [OVS_VPORT_ATTR_NAME] = { .type = NLA_NUL_STRING, .len = IFNAMSIZ - 1 },
1424 [OVS_VPORT_ATTR_STATS] = { .len = sizeof(struct ovs_vport_stats) },
1425 [OVS_VPORT_ATTR_PORT_NO] = { .type = NLA_U32 },
1426 [OVS_VPORT_ATTR_TYPE] = { .type = NLA_U32 },
1427 [OVS_VPORT_ATTR_UPCALL_PID] = { .type = NLA_U32 },
1428 [OVS_VPORT_ATTR_OPTIONS] = { .type = NLA_NESTED },
1431 static struct genl_family dp_vport_genl_family = {
1432 .id = GENL_ID_GENERATE,
1433 .hdrsize = sizeof(struct ovs_header),
1434 .name = OVS_VPORT_FAMILY,
1435 .version = OVS_VPORT_VERSION,
1436 .maxattr = OVS_VPORT_ATTR_MAX,
1441 struct genl_multicast_group ovs_dp_vport_multicast_group = {
1442 .name = OVS_VPORT_MCGROUP
1445 /* Called with ovs_mutex or RCU read lock. */
1446 static int ovs_vport_cmd_fill_info(struct vport *vport, struct sk_buff *skb,
1447 u32 portid, u32 seq, u32 flags, u8 cmd)
1449 struct ovs_header *ovs_header;
1450 struct ovs_vport_stats vport_stats;
1453 ovs_header = genlmsg_put(skb, portid, seq, &dp_vport_genl_family,
1458 ovs_header->dp_ifindex = get_dpifindex(vport->dp);
1460 if (nla_put_u32(skb, OVS_VPORT_ATTR_PORT_NO, vport->port_no) ||
1461 nla_put_u32(skb, OVS_VPORT_ATTR_TYPE, vport->ops->type) ||
1462 nla_put_string(skb, OVS_VPORT_ATTR_NAME, vport->ops->get_name(vport)) ||
1463 nla_put_u32(skb, OVS_VPORT_ATTR_UPCALL_PID, vport->upcall_portid))
1464 goto nla_put_failure;
1466 ovs_vport_get_stats(vport, &vport_stats);
1467 if (nla_put(skb, OVS_VPORT_ATTR_STATS, sizeof(struct ovs_vport_stats),
1469 goto nla_put_failure;
1471 err = ovs_vport_get_options(vport, skb);
1472 if (err == -EMSGSIZE)
1475 return genlmsg_end(skb, ovs_header);
1480 genlmsg_cancel(skb, ovs_header);
1484 /* Called with ovs_mutex or RCU read lock. */
1485 struct sk_buff *ovs_vport_cmd_build_info(struct vport *vport, u32 portid,
1488 struct sk_buff *skb;
1491 skb = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_ATOMIC);
1493 return ERR_PTR(-ENOMEM);
1495 retval = ovs_vport_cmd_fill_info(vport, skb, portid, seq, 0, cmd);
1501 /* Called with ovs_mutex or RCU read lock. */
1502 static struct vport *lookup_vport(struct net *net,
1503 struct ovs_header *ovs_header,
1504 struct nlattr *a[OVS_VPORT_ATTR_MAX + 1])
1506 struct datapath *dp;
1507 struct vport *vport;
1509 if (a[OVS_VPORT_ATTR_NAME]) {
1510 vport = ovs_vport_locate(net, nla_data(a[OVS_VPORT_ATTR_NAME]));
1512 return ERR_PTR(-ENODEV);
1513 if (ovs_header->dp_ifindex &&
1514 ovs_header->dp_ifindex != get_dpifindex(vport->dp))
1515 return ERR_PTR(-ENODEV);
1517 } else if (a[OVS_VPORT_ATTR_PORT_NO]) {
1518 u32 port_no = nla_get_u32(a[OVS_VPORT_ATTR_PORT_NO]);
1520 if (port_no >= DP_MAX_PORTS)
1521 return ERR_PTR(-EFBIG);
1523 dp = get_dp(net, ovs_header->dp_ifindex);
1525 return ERR_PTR(-ENODEV);
1527 vport = ovs_vport_ovsl_rcu(dp, port_no);
1529 return ERR_PTR(-ENODEV);
1532 return ERR_PTR(-EINVAL);
1535 static int ovs_vport_cmd_new(struct sk_buff *skb, struct genl_info *info)
1537 struct nlattr **a = info->attrs;
1538 struct ovs_header *ovs_header = info->userhdr;
1539 struct vport_parms parms;
1540 struct sk_buff *reply;
1541 struct vport *vport;
1542 struct datapath *dp;
1547 if (!a[OVS_VPORT_ATTR_NAME] || !a[OVS_VPORT_ATTR_TYPE] ||
1548 !a[OVS_VPORT_ATTR_UPCALL_PID])
1552 dp = get_dp(sock_net(skb->sk), ovs_header->dp_ifindex);
1557 if (a[OVS_VPORT_ATTR_PORT_NO]) {
1558 port_no = nla_get_u32(a[OVS_VPORT_ATTR_PORT_NO]);
1561 if (port_no >= DP_MAX_PORTS)
1564 vport = ovs_vport_ovsl(dp, port_no);
1569 for (port_no = 1; ; port_no++) {
1570 if (port_no >= DP_MAX_PORTS) {
1574 vport = ovs_vport_ovsl(dp, port_no);
1580 parms.name = nla_data(a[OVS_VPORT_ATTR_NAME]);
1581 parms.type = nla_get_u32(a[OVS_VPORT_ATTR_TYPE]);
1582 parms.options = a[OVS_VPORT_ATTR_OPTIONS];
1584 parms.port_no = port_no;
1585 parms.upcall_portid = nla_get_u32(a[OVS_VPORT_ATTR_UPCALL_PID]);
1587 vport = new_vport(&parms);
1588 err = PTR_ERR(vport);
1593 if (a[OVS_VPORT_ATTR_STATS])
1594 ovs_vport_set_stats(vport, nla_data(a[OVS_VPORT_ATTR_STATS]));
1596 reply = ovs_vport_cmd_build_info(vport, info->snd_portid, info->snd_seq,
1598 if (IS_ERR(reply)) {
1599 err = PTR_ERR(reply);
1600 ovs_dp_detach_port(vport);
1604 ovs_notify(reply, info, &ovs_dp_vport_multicast_group);
1612 static int ovs_vport_cmd_set(struct sk_buff *skb, struct genl_info *info)
1614 struct nlattr **a = info->attrs;
1615 struct sk_buff *reply;
1616 struct vport *vport;
1620 vport = lookup_vport(sock_net(skb->sk), info->userhdr, a);
1621 err = PTR_ERR(vport);
1625 if (a[OVS_VPORT_ATTR_TYPE] &&
1626 nla_get_u32(a[OVS_VPORT_ATTR_TYPE]) != vport->ops->type) {
1631 reply = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
1637 if (a[OVS_VPORT_ATTR_OPTIONS]) {
1638 err = ovs_vport_set_options(vport, a[OVS_VPORT_ATTR_OPTIONS]);
1643 if (a[OVS_VPORT_ATTR_STATS])
1644 ovs_vport_set_stats(vport, nla_data(a[OVS_VPORT_ATTR_STATS]));
1646 if (a[OVS_VPORT_ATTR_UPCALL_PID])
1647 vport->upcall_portid = nla_get_u32(a[OVS_VPORT_ATTR_UPCALL_PID]);
1649 err = ovs_vport_cmd_fill_info(vport, reply, info->snd_portid,
1650 info->snd_seq, 0, OVS_VPORT_CMD_NEW);
1654 ovs_notify(reply, info, &ovs_dp_vport_multicast_group);
1664 static int ovs_vport_cmd_del(struct sk_buff *skb, struct genl_info *info)
1666 struct nlattr **a = info->attrs;
1667 struct sk_buff *reply;
1668 struct vport *vport;
1672 vport = lookup_vport(sock_net(skb->sk), info->userhdr, a);
1673 err = PTR_ERR(vport);
1677 if (vport->port_no == OVSP_LOCAL) {
1682 reply = ovs_vport_cmd_build_info(vport, info->snd_portid,
1683 info->snd_seq, OVS_VPORT_CMD_DEL);
1684 err = PTR_ERR(reply);
1689 ovs_dp_detach_port(vport);
1691 ovs_notify(reply, info, &ovs_dp_vport_multicast_group);
1698 static int ovs_vport_cmd_get(struct sk_buff *skb, struct genl_info *info)
1700 struct nlattr **a = info->attrs;
1701 struct ovs_header *ovs_header = info->userhdr;
1702 struct sk_buff *reply;
1703 struct vport *vport;
1707 vport = lookup_vport(sock_net(skb->sk), ovs_header, a);
1708 err = PTR_ERR(vport);
1712 reply = ovs_vport_cmd_build_info(vport, info->snd_portid,
1713 info->snd_seq, OVS_VPORT_CMD_NEW);
1714 err = PTR_ERR(reply);
1720 return genlmsg_reply(reply, info);
1727 static int ovs_vport_cmd_dump(struct sk_buff *skb, struct netlink_callback *cb)
1729 struct ovs_header *ovs_header = genlmsg_data(nlmsg_data(cb->nlh));
1730 struct datapath *dp;
1731 int bucket = cb->args[0], skip = cb->args[1];
1734 dp = get_dp(sock_net(skb->sk), ovs_header->dp_ifindex);
1739 for (i = bucket; i < DP_VPORT_HASH_BUCKETS; i++) {
1740 struct vport *vport;
1743 hlist_for_each_entry_rcu(vport, &dp->ports[i], dp_hash_node) {
1745 ovs_vport_cmd_fill_info(vport, skb,
1746 NETLINK_CB(cb->skb).portid,
1749 OVS_VPORT_CMD_NEW) < 0)
1765 static struct genl_ops dp_vport_genl_ops[] = {
1766 { .cmd = OVS_VPORT_CMD_NEW,
1767 .flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
1768 .policy = vport_policy,
1769 .doit = ovs_vport_cmd_new
1771 { .cmd = OVS_VPORT_CMD_DEL,
1772 .flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
1773 .policy = vport_policy,
1774 .doit = ovs_vport_cmd_del
1776 { .cmd = OVS_VPORT_CMD_GET,
1777 .flags = 0, /* OK for unprivileged users. */
1778 .policy = vport_policy,
1779 .doit = ovs_vport_cmd_get,
1780 .dumpit = ovs_vport_cmd_dump
1782 { .cmd = OVS_VPORT_CMD_SET,
1783 .flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
1784 .policy = vport_policy,
1785 .doit = ovs_vport_cmd_set,
1789 struct genl_family_and_ops {
1790 struct genl_family *family;
1791 struct genl_ops *ops;
1793 struct genl_multicast_group *group;
1796 static const struct genl_family_and_ops dp_genl_families[] = {
1797 { &dp_datapath_genl_family,
1798 dp_datapath_genl_ops, ARRAY_SIZE(dp_datapath_genl_ops),
1799 &ovs_dp_datapath_multicast_group },
1800 { &dp_vport_genl_family,
1801 dp_vport_genl_ops, ARRAY_SIZE(dp_vport_genl_ops),
1802 &ovs_dp_vport_multicast_group },
1803 { &dp_flow_genl_family,
1804 dp_flow_genl_ops, ARRAY_SIZE(dp_flow_genl_ops),
1805 &ovs_dp_flow_multicast_group },
1806 { &dp_packet_genl_family,
1807 dp_packet_genl_ops, ARRAY_SIZE(dp_packet_genl_ops),
1811 static void dp_unregister_genl(int n_families)
1815 for (i = 0; i < n_families; i++)
1816 genl_unregister_family(dp_genl_families[i].family);
1819 static int dp_register_genl(void)
1826 for (i = 0; i < ARRAY_SIZE(dp_genl_families); i++) {
1827 const struct genl_family_and_ops *f = &dp_genl_families[i];
1829 err = genl_register_family_with_ops(f->family, f->ops,
1836 err = genl_register_mc_group(f->family, f->group);
1845 dp_unregister_genl(n_registered);
1849 static int __net_init ovs_init_net(struct net *net)
1851 struct ovs_net *ovs_net = net_generic(net, ovs_net_id);
1853 INIT_LIST_HEAD(&ovs_net->dps);
1854 INIT_WORK(&ovs_net->dp_notify_work, ovs_dp_notify_wq);
1858 static void __net_exit ovs_exit_net(struct net *net)
1860 struct datapath *dp, *dp_next;
1861 struct ovs_net *ovs_net = net_generic(net, ovs_net_id);
1864 list_for_each_entry_safe(dp, dp_next, &ovs_net->dps, list_node)
1868 cancel_work_sync(&ovs_net->dp_notify_work);
1871 static struct pernet_operations ovs_net_ops = {
1872 .init = ovs_init_net,
1873 .exit = ovs_exit_net,
1875 .size = sizeof(struct ovs_net),
1878 DEFINE_COMPAT_PNET_REG_FUNC(device);
1880 static int __init dp_init(void)
1884 BUILD_BUG_ON(sizeof(struct ovs_skb_cb) > FIELD_SIZEOF(struct sk_buff, cb));
1886 pr_info("Open vSwitch switching datapath %s, built "__DATE__" "__TIME__"\n",
1889 err = ovs_flow_init();
1893 err = ovs_vport_init();
1895 goto error_flow_exit;
1897 err = register_pernet_device(&ovs_net_ops);
1899 goto error_vport_exit;
1901 err = register_netdevice_notifier(&ovs_dp_device_notifier);
1903 goto error_netns_exit;
1905 err = dp_register_genl();
1907 goto error_unreg_notifier;
1911 error_unreg_notifier:
1912 unregister_netdevice_notifier(&ovs_dp_device_notifier);
1914 unregister_pernet_device(&ovs_net_ops);
1923 static void dp_cleanup(void)
1925 dp_unregister_genl(ARRAY_SIZE(dp_genl_families));
1926 unregister_netdevice_notifier(&ovs_dp_device_notifier);
1927 unregister_pernet_device(&ovs_net_ops);
1933 module_init(dp_init);
1934 module_exit(dp_cleanup);
1936 MODULE_DESCRIPTION("Open vSwitch switching datapath");
1937 MODULE_LICENSE("GPL");
1938 MODULE_VERSION(VERSION);