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_stats(&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 ovs_flow_stats_update(OVS_CB(skb)->flow, skb);
257 ovs_execute_actions(dp, skb);
258 stats_counter = &stats->n_hit;
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 int ovs_packet_cmd_execute(struct sk_buff *skb, struct genl_info *info)
461 struct ovs_header *ovs_header = info->userhdr;
462 struct nlattr **a = info->attrs;
463 struct sw_flow_actions *acts;
464 struct sk_buff *packet;
465 struct sw_flow *flow;
472 if (!a[OVS_PACKET_ATTR_PACKET] || !a[OVS_PACKET_ATTR_KEY] ||
473 !a[OVS_PACKET_ATTR_ACTIONS])
476 len = nla_len(a[OVS_PACKET_ATTR_PACKET]);
477 packet = __dev_alloc_skb(NET_IP_ALIGN + len, GFP_KERNEL);
481 skb_reserve(packet, NET_IP_ALIGN);
483 nla_memcpy(__skb_put(packet, len), a[OVS_PACKET_ATTR_PACKET], len);
485 skb_reset_mac_header(packet);
486 eth = eth_hdr(packet);
488 /* Normally, setting the skb 'protocol' field would be handled by a
489 * call to eth_type_trans(), but it assumes there's a sending
490 * device, which we may not have. */
491 if (ntohs(eth->h_proto) >= ETH_P_802_3_MIN)
492 packet->protocol = eth->h_proto;
494 packet->protocol = htons(ETH_P_802_2);
496 /* Build an sw_flow for sending this packet. */
497 flow = ovs_flow_alloc();
502 err = ovs_flow_extract(packet, -1, &flow->key);
506 err = ovs_nla_get_flow_metadata(flow, a[OVS_PACKET_ATTR_KEY]);
509 acts = ovs_nla_alloc_flow_actions(nla_len(a[OVS_PACKET_ATTR_ACTIONS]));
514 err = ovs_nla_copy_actions(a[OVS_PACKET_ATTR_ACTIONS],
515 &flow->key, 0, &acts);
516 rcu_assign_pointer(flow->sf_acts, acts);
520 OVS_CB(packet)->flow = flow;
521 OVS_CB(packet)->pkt_key = &flow->key;
522 packet->priority = flow->key.phy.priority;
523 packet->mark = flow->key.phy.skb_mark;
526 dp = get_dp(sock_net(skb->sk), ovs_header->dp_ifindex);
532 err = ovs_execute_actions(dp, packet);
536 ovs_flow_free(flow, false);
542 ovs_flow_free(flow, false);
549 static const struct nla_policy packet_policy[OVS_PACKET_ATTR_MAX + 1] = {
550 [OVS_PACKET_ATTR_PACKET] = { .len = ETH_HLEN },
551 [OVS_PACKET_ATTR_KEY] = { .type = NLA_NESTED },
552 [OVS_PACKET_ATTR_ACTIONS] = { .type = NLA_NESTED },
555 static struct genl_ops dp_packet_genl_ops[] = {
556 { .cmd = OVS_PACKET_CMD_EXECUTE,
557 .flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
558 .policy = packet_policy,
559 .doit = ovs_packet_cmd_execute
563 static void get_dp_stats(struct datapath *dp, struct ovs_dp_stats *stats,
564 struct ovs_dp_megaflow_stats *mega_stats)
568 memset(mega_stats, 0, sizeof(*mega_stats));
570 stats->n_flows = ovs_flow_tbl_count(&dp->table);
571 mega_stats->n_masks = ovs_flow_tbl_num_masks(&dp->table);
573 stats->n_hit = stats->n_missed = stats->n_lost = 0;
575 for_each_possible_cpu(i) {
576 const struct dp_stats_percpu *percpu_stats;
577 struct dp_stats_percpu local_stats;
580 percpu_stats = per_cpu_ptr(dp->stats_percpu, i);
583 start = u64_stats_fetch_begin_bh(&percpu_stats->sync);
584 local_stats = *percpu_stats;
585 } while (u64_stats_fetch_retry_bh(&percpu_stats->sync, start));
587 stats->n_hit += local_stats.n_hit;
588 stats->n_missed += local_stats.n_missed;
589 stats->n_lost += local_stats.n_lost;
590 mega_stats->n_mask_hit += local_stats.n_mask_hit;
594 static const struct nla_policy flow_policy[OVS_FLOW_ATTR_MAX + 1] = {
595 [OVS_FLOW_ATTR_KEY] = { .type = NLA_NESTED },
596 [OVS_FLOW_ATTR_ACTIONS] = { .type = NLA_NESTED },
597 [OVS_FLOW_ATTR_CLEAR] = { .type = NLA_FLAG },
600 static struct genl_family dp_flow_genl_family = {
601 .id = GENL_ID_GENERATE,
602 .hdrsize = sizeof(struct ovs_header),
603 .name = OVS_FLOW_FAMILY,
604 .version = OVS_FLOW_VERSION,
605 .maxattr = OVS_FLOW_ATTR_MAX,
610 static struct genl_multicast_group ovs_dp_flow_multicast_group = {
611 .name = OVS_FLOW_MCGROUP
614 static size_t ovs_flow_cmd_msg_size(const struct sw_flow_actions *acts)
616 return NLMSG_ALIGN(sizeof(struct ovs_header))
617 + nla_total_size(key_attr_size()) /* OVS_FLOW_ATTR_KEY */
618 + nla_total_size(key_attr_size()) /* OVS_FLOW_ATTR_MASK */
619 + nla_total_size(sizeof(struct ovs_flow_stats)) /* OVS_FLOW_ATTR_STATS */
620 + nla_total_size(1) /* OVS_FLOW_ATTR_TCP_FLAGS */
621 + nla_total_size(8) /* OVS_FLOW_ATTR_USED */
622 + nla_total_size(acts->actions_len); /* OVS_FLOW_ATTR_ACTIONS */
625 /* Called with ovs_mutex. */
626 static int ovs_flow_cmd_fill_info(struct sw_flow *flow, struct datapath *dp,
627 struct sk_buff *skb, u32 portid,
628 u32 seq, u32 flags, u8 cmd)
630 const int skb_orig_len = skb->len;
631 struct nlattr *start;
632 struct sw_flow_stats flow_stats;
633 struct ovs_header *ovs_header;
637 ovs_header = genlmsg_put(skb, portid, seq, &dp_flow_genl_family, flags, cmd);
641 ovs_header->dp_ifindex = get_dpifindex(dp);
644 nla = nla_nest_start(skb, OVS_FLOW_ATTR_KEY);
646 goto nla_put_failure;
648 err = ovs_nla_put_flow(&flow->unmasked_key, &flow->unmasked_key, skb);
651 nla_nest_end(skb, nla);
653 nla = nla_nest_start(skb, OVS_FLOW_ATTR_MASK);
655 goto nla_put_failure;
657 err = ovs_nla_put_flow(&flow->key, &flow->mask->key, skb);
661 nla_nest_end(skb, nla);
663 ovs_flow_stats_get(flow, &flow_stats);
664 if (flow_stats.used &&
665 nla_put_u64(skb, OVS_FLOW_ATTR_USED, ovs_flow_used_time(flow_stats.used)))
666 goto nla_put_failure;
668 if (flow_stats.packet_count) {
669 struct ovs_flow_stats stats = {
670 .n_packets = flow_stats.packet_count,
671 .n_bytes = flow_stats.byte_count,
674 if (nla_put(skb, OVS_FLOW_ATTR_STATS,
675 sizeof(struct ovs_flow_stats), &stats))
676 goto nla_put_failure;
679 if ((u8)ntohs(flow_stats.tcp_flags) &&
680 nla_put_u8(skb, OVS_FLOW_ATTR_TCP_FLAGS,
681 (u8)ntohs(flow_stats.tcp_flags)))
682 goto nla_put_failure;
684 /* If OVS_FLOW_ATTR_ACTIONS doesn't fit, skip dumping the actions if
685 * this is the first flow to be dumped into 'skb'. This is unusual for
686 * Netlink but individual action lists can be longer than
687 * NLMSG_GOODSIZE and thus entirely undumpable if we didn't do this.
688 * The userspace caller can always fetch the actions separately if it
689 * really wants them. (Most userspace callers in fact don't care.)
691 * This can only fail for dump operations because the skb is always
692 * properly sized for single flows.
694 start = nla_nest_start(skb, OVS_FLOW_ATTR_ACTIONS);
696 const struct sw_flow_actions *sf_acts;
698 sf_acts = rcu_dereference_check(flow->sf_acts,
699 lockdep_ovsl_is_held());
701 err = ovs_nla_put_actions(sf_acts->actions,
702 sf_acts->actions_len, skb);
704 nla_nest_end(skb, start);
709 nla_nest_cancel(skb, start);
711 } else if (skb_orig_len)
712 goto nla_put_failure;
714 return genlmsg_end(skb, ovs_header);
719 genlmsg_cancel(skb, ovs_header);
723 static struct sk_buff *ovs_flow_cmd_alloc_info(struct sw_flow *flow)
725 const struct sw_flow_actions *sf_acts;
727 sf_acts = ovsl_dereference(flow->sf_acts);
729 return genlmsg_new(ovs_flow_cmd_msg_size(sf_acts), GFP_KERNEL);
732 static struct sk_buff *ovs_flow_cmd_build_info(struct sw_flow *flow,
734 u32 portid, u32 seq, u8 cmd)
739 skb = ovs_flow_cmd_alloc_info(flow);
741 return ERR_PTR(-ENOMEM);
743 retval = ovs_flow_cmd_fill_info(flow, dp, skb, portid, seq, 0, cmd);
748 static int ovs_flow_cmd_new_or_set(struct sk_buff *skb, struct genl_info *info)
750 struct nlattr **a = info->attrs;
751 struct ovs_header *ovs_header = info->userhdr;
752 struct sw_flow_key key, masked_key;
753 struct sw_flow *flow = NULL;
754 struct sw_flow_mask mask;
755 struct sk_buff *reply;
757 struct sw_flow_actions *acts = NULL;
758 struct sw_flow_match match;
763 if (!a[OVS_FLOW_ATTR_KEY])
766 ovs_match_init(&match, &key, &mask);
767 error = ovs_nla_get_match(&match,
768 a[OVS_FLOW_ATTR_KEY], a[OVS_FLOW_ATTR_MASK]);
772 /* Validate actions. */
773 if (a[OVS_FLOW_ATTR_ACTIONS]) {
774 acts = ovs_nla_alloc_flow_actions(nla_len(a[OVS_FLOW_ATTR_ACTIONS]));
775 error = PTR_ERR(acts);
779 ovs_flow_mask_key(&masked_key, &key, &mask);
780 error = ovs_nla_copy_actions(a[OVS_FLOW_ATTR_ACTIONS],
781 &masked_key, 0, &acts);
783 OVS_NLERR("Flow actions may not be safe on all matching packets.\n");
786 } else if (info->genlhdr->cmd == OVS_FLOW_CMD_NEW) {
792 dp = get_dp(sock_net(skb->sk), ovs_header->dp_ifindex);
797 /* Check if this is a duplicate flow */
798 flow = ovs_flow_tbl_lookup(&dp->table, &key);
800 /* Bail out if we're not allowed to create a new flow. */
802 if (info->genlhdr->cmd == OVS_FLOW_CMD_SET)
806 flow = ovs_flow_alloc();
808 error = PTR_ERR(flow);
812 flow->key = masked_key;
813 flow->unmasked_key = key;
814 rcu_assign_pointer(flow->sf_acts, acts);
816 /* Put flow in bucket. */
817 error = ovs_flow_tbl_insert(&dp->table, flow, &mask);
823 reply = ovs_flow_cmd_build_info(flow, dp, info->snd_portid,
824 info->snd_seq, OVS_FLOW_CMD_NEW);
826 /* We found a matching flow. */
827 struct sw_flow_actions *old_acts;
829 /* Bail out if we're not allowed to modify an existing flow.
830 * We accept NLM_F_CREATE in place of the intended NLM_F_EXCL
831 * because Generic Netlink treats the latter as a dump
832 * request. We also accept NLM_F_EXCL in case that bug ever
836 if (info->genlhdr->cmd == OVS_FLOW_CMD_NEW &&
837 info->nlhdr->nlmsg_flags & (NLM_F_CREATE | NLM_F_EXCL))
840 /* The unmasked key has to be the same for flow updates. */
842 if (!ovs_flow_cmp_unmasked_key(flow, &match)) {
843 OVS_NLERR("Flow modification message rejected, unmasked key does not match.\n");
847 /* Update actions. */
848 old_acts = ovsl_dereference(flow->sf_acts);
849 rcu_assign_pointer(flow->sf_acts, acts);
850 ovs_nla_free_flow_actions(old_acts);
852 reply = ovs_flow_cmd_build_info(flow, dp, info->snd_portid,
853 info->snd_seq, OVS_FLOW_CMD_NEW);
856 if (a[OVS_FLOW_ATTR_CLEAR])
857 ovs_flow_stats_clear(flow);
862 ovs_notify(reply, info, &ovs_dp_flow_multicast_group);
864 netlink_set_err(sock_net(skb->sk)->genl_sock, 0,
865 ovs_dp_flow_multicast_group.id, PTR_ERR(reply));
869 ovs_flow_free(flow, false);
878 static int ovs_flow_cmd_get(struct sk_buff *skb, struct genl_info *info)
880 struct nlattr **a = info->attrs;
881 struct ovs_header *ovs_header = info->userhdr;
882 struct sw_flow_key key;
883 struct sk_buff *reply;
884 struct sw_flow *flow;
886 struct sw_flow_match match;
889 if (!a[OVS_FLOW_ATTR_KEY]) {
890 OVS_NLERR("Flow get message rejected, Key attribute missing.\n");
894 ovs_match_init(&match, &key, NULL);
895 err = ovs_nla_get_match(&match, a[OVS_FLOW_ATTR_KEY], NULL);
900 dp = get_dp(sock_net(skb->sk), ovs_header->dp_ifindex);
906 flow = ovs_flow_tbl_lookup(&dp->table, &key);
907 if (!flow || !ovs_flow_cmp_unmasked_key(flow, &match)) {
912 reply = ovs_flow_cmd_build_info(flow, dp, info->snd_portid,
913 info->snd_seq, OVS_FLOW_CMD_NEW);
915 err = PTR_ERR(reply);
920 return genlmsg_reply(reply, info);
926 static int ovs_flow_cmd_del(struct sk_buff *skb, struct genl_info *info)
928 struct nlattr **a = info->attrs;
929 struct ovs_header *ovs_header = info->userhdr;
930 struct sw_flow_key key;
931 struct sk_buff *reply;
932 struct sw_flow *flow;
934 struct sw_flow_match match;
938 dp = get_dp(sock_net(skb->sk), ovs_header->dp_ifindex);
944 if (!a[OVS_FLOW_ATTR_KEY]) {
945 err = ovs_flow_tbl_flush(&dp->table);
949 ovs_match_init(&match, &key, NULL);
950 err = ovs_nla_get_match(&match, a[OVS_FLOW_ATTR_KEY], NULL);
954 flow = ovs_flow_tbl_lookup(&dp->table, &key);
955 if (!flow || !ovs_flow_cmp_unmasked_key(flow, &match)) {
960 reply = ovs_flow_cmd_alloc_info(flow);
966 ovs_flow_tbl_remove(&dp->table, flow);
968 err = ovs_flow_cmd_fill_info(flow, dp, reply, info->snd_portid,
969 info->snd_seq, 0, OVS_FLOW_CMD_DEL);
972 ovs_flow_free(flow, true);
975 ovs_notify(reply, info, &ovs_dp_flow_multicast_group);
982 static int ovs_flow_cmd_dump(struct sk_buff *skb, struct netlink_callback *cb)
984 struct ovs_header *ovs_header = genlmsg_data(nlmsg_data(cb->nlh));
985 struct table_instance *ti;
989 dp = get_dp(sock_net(skb->sk), ovs_header->dp_ifindex);
995 ti = rcu_dereference(dp->table.ti);
997 struct sw_flow *flow;
1000 bucket = cb->args[0];
1002 flow = ovs_flow_tbl_dump_next(ti, &bucket, &obj);
1006 if (ovs_flow_cmd_fill_info(flow, dp, skb,
1007 NETLINK_CB(cb->skb).portid,
1008 cb->nlh->nlmsg_seq, NLM_F_MULTI,
1009 OVS_FLOW_CMD_NEW) < 0)
1012 cb->args[0] = bucket;
1019 static struct genl_ops dp_flow_genl_ops[] = {
1020 { .cmd = OVS_FLOW_CMD_NEW,
1021 .flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
1022 .policy = flow_policy,
1023 .doit = ovs_flow_cmd_new_or_set
1025 { .cmd = OVS_FLOW_CMD_DEL,
1026 .flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
1027 .policy = flow_policy,
1028 .doit = ovs_flow_cmd_del
1030 { .cmd = OVS_FLOW_CMD_GET,
1031 .flags = 0, /* OK for unprivileged users. */
1032 .policy = flow_policy,
1033 .doit = ovs_flow_cmd_get,
1034 .dumpit = ovs_flow_cmd_dump
1036 { .cmd = OVS_FLOW_CMD_SET,
1037 .flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
1038 .policy = flow_policy,
1039 .doit = ovs_flow_cmd_new_or_set,
1043 static const struct nla_policy datapath_policy[OVS_DP_ATTR_MAX + 1] = {
1044 [OVS_DP_ATTR_NAME] = { .type = NLA_NUL_STRING, .len = IFNAMSIZ - 1 },
1045 [OVS_DP_ATTR_UPCALL_PID] = { .type = NLA_U32 },
1048 static struct genl_family dp_datapath_genl_family = {
1049 .id = GENL_ID_GENERATE,
1050 .hdrsize = sizeof(struct ovs_header),
1051 .name = OVS_DATAPATH_FAMILY,
1052 .version = OVS_DATAPATH_VERSION,
1053 .maxattr = OVS_DP_ATTR_MAX,
1058 static struct genl_multicast_group ovs_dp_datapath_multicast_group = {
1059 .name = OVS_DATAPATH_MCGROUP
1062 static size_t ovs_dp_cmd_msg_size(void)
1064 size_t msgsize = NLMSG_ALIGN(sizeof(struct ovs_header));
1066 msgsize += nla_total_size(IFNAMSIZ);
1067 msgsize += nla_total_size(sizeof(struct ovs_dp_stats));
1068 msgsize += nla_total_size(sizeof(struct ovs_dp_megaflow_stats));
1073 static int ovs_dp_cmd_fill_info(struct datapath *dp, struct sk_buff *skb,
1074 u32 portid, u32 seq, u32 flags, u8 cmd)
1076 struct ovs_header *ovs_header;
1077 struct ovs_dp_stats dp_stats;
1078 struct ovs_dp_megaflow_stats dp_megaflow_stats;
1081 ovs_header = genlmsg_put(skb, portid, seq, &dp_datapath_genl_family,
1086 ovs_header->dp_ifindex = get_dpifindex(dp);
1089 err = nla_put_string(skb, OVS_DP_ATTR_NAME, ovs_dp_name(dp));
1092 goto nla_put_failure;
1094 get_dp_stats(dp, &dp_stats, &dp_megaflow_stats);
1095 if (nla_put(skb, OVS_DP_ATTR_STATS, sizeof(struct ovs_dp_stats),
1097 goto nla_put_failure;
1099 if (nla_put(skb, OVS_DP_ATTR_MEGAFLOW_STATS,
1100 sizeof(struct ovs_dp_megaflow_stats),
1101 &dp_megaflow_stats))
1102 goto nla_put_failure;
1104 return genlmsg_end(skb, ovs_header);
1107 genlmsg_cancel(skb, ovs_header);
1112 static struct sk_buff *ovs_dp_cmd_build_info(struct datapath *dp, u32 portid,
1115 struct sk_buff *skb;
1118 skb = genlmsg_new(ovs_dp_cmd_msg_size(), GFP_KERNEL);
1120 return ERR_PTR(-ENOMEM);
1122 retval = ovs_dp_cmd_fill_info(dp, skb, portid, seq, 0, cmd);
1125 return ERR_PTR(retval);
1130 /* Called with ovs_mutex. */
1131 static struct datapath *lookup_datapath(struct net *net,
1132 struct ovs_header *ovs_header,
1133 struct nlattr *a[OVS_DP_ATTR_MAX + 1])
1135 struct datapath *dp;
1137 if (!a[OVS_DP_ATTR_NAME])
1138 dp = get_dp(net, ovs_header->dp_ifindex);
1140 struct vport *vport;
1143 vport = ovs_vport_locate(net, nla_data(a[OVS_DP_ATTR_NAME]));
1144 dp = vport && vport->port_no == OVSP_LOCAL ? vport->dp : NULL;
1147 return dp ? dp : ERR_PTR(-ENODEV);
1150 static int ovs_dp_cmd_new(struct sk_buff *skb, struct genl_info *info)
1152 struct nlattr **a = info->attrs;
1153 struct vport_parms parms;
1154 struct sk_buff *reply;
1155 struct datapath *dp;
1156 struct vport *vport;
1157 struct ovs_net *ovs_net;
1161 if (!a[OVS_DP_ATTR_NAME] || !a[OVS_DP_ATTR_UPCALL_PID])
1167 dp = kzalloc(sizeof(*dp), GFP_KERNEL);
1169 goto err_unlock_ovs;
1171 ovs_dp_set_net(dp, hold_net(sock_net(skb->sk)));
1173 /* Allocate table. */
1174 err = ovs_flow_tbl_init(&dp->table);
1178 dp->stats_percpu = alloc_percpu(struct dp_stats_percpu);
1179 if (!dp->stats_percpu) {
1181 goto err_destroy_table;
1184 dp->ports = kmalloc(DP_VPORT_HASH_BUCKETS * sizeof(struct hlist_head),
1188 goto err_destroy_percpu;
1191 for (i = 0; i < DP_VPORT_HASH_BUCKETS; i++)
1192 INIT_HLIST_HEAD(&dp->ports[i]);
1194 /* Set up our datapath device. */
1195 parms.name = nla_data(a[OVS_DP_ATTR_NAME]);
1196 parms.type = OVS_VPORT_TYPE_INTERNAL;
1197 parms.options = NULL;
1199 parms.port_no = OVSP_LOCAL;
1200 parms.upcall_portid = nla_get_u32(a[OVS_DP_ATTR_UPCALL_PID]);
1202 vport = new_vport(&parms);
1203 if (IS_ERR(vport)) {
1204 err = PTR_ERR(vport);
1208 goto err_destroy_ports_array;
1211 reply = ovs_dp_cmd_build_info(dp, info->snd_portid,
1212 info->snd_seq, OVS_DP_CMD_NEW);
1213 err = PTR_ERR(reply);
1215 goto err_destroy_local_port;
1217 ovs_net = net_generic(ovs_dp_get_net(dp), ovs_net_id);
1218 list_add_tail_rcu(&dp->list_node, &ovs_net->dps);
1222 ovs_notify(reply, info, &ovs_dp_datapath_multicast_group);
1225 err_destroy_local_port:
1226 ovs_dp_detach_port(ovs_vport_ovsl(dp, OVSP_LOCAL));
1227 err_destroy_ports_array:
1230 free_percpu(dp->stats_percpu);
1232 ovs_flow_tbl_destroy(&dp->table);
1234 release_net(ovs_dp_get_net(dp));
1242 /* Called with ovs_mutex. */
1243 static void __dp_destroy(struct datapath *dp)
1247 for (i = 0; i < DP_VPORT_HASH_BUCKETS; i++) {
1248 struct vport *vport;
1249 struct hlist_node *n;
1251 hlist_for_each_entry_safe(vport, n, &dp->ports[i], dp_hash_node)
1252 if (vport->port_no != OVSP_LOCAL)
1253 ovs_dp_detach_port(vport);
1256 list_del_rcu(&dp->list_node);
1258 /* OVSP_LOCAL is datapath internal port. We need to make sure that
1259 * all port in datapath are destroyed first before freeing datapath.
1261 ovs_dp_detach_port(ovs_vport_ovsl(dp, OVSP_LOCAL));
1263 call_rcu(&dp->rcu, destroy_dp_rcu);
1266 static int ovs_dp_cmd_del(struct sk_buff *skb, struct genl_info *info)
1268 struct sk_buff *reply;
1269 struct datapath *dp;
1273 dp = lookup_datapath(sock_net(skb->sk), info->userhdr, info->attrs);
1278 reply = ovs_dp_cmd_build_info(dp, info->snd_portid,
1279 info->snd_seq, OVS_DP_CMD_DEL);
1280 err = PTR_ERR(reply);
1287 ovs_notify(reply, info, &ovs_dp_datapath_multicast_group);
1295 static int ovs_dp_cmd_set(struct sk_buff *skb, struct genl_info *info)
1297 struct sk_buff *reply;
1298 struct datapath *dp;
1302 dp = lookup_datapath(sock_net(skb->sk), info->userhdr, info->attrs);
1307 reply = ovs_dp_cmd_build_info(dp, info->snd_portid,
1308 info->snd_seq, OVS_DP_CMD_NEW);
1309 if (IS_ERR(reply)) {
1310 err = PTR_ERR(reply);
1311 netlink_set_err(sock_net(skb->sk)->genl_sock, 0,
1312 ovs_dp_datapath_multicast_group.id, err);
1318 ovs_notify(reply, info, &ovs_dp_datapath_multicast_group);
1326 static int ovs_dp_cmd_get(struct sk_buff *skb, struct genl_info *info)
1328 struct sk_buff *reply;
1329 struct datapath *dp;
1333 dp = lookup_datapath(sock_net(skb->sk), info->userhdr, info->attrs);
1339 reply = ovs_dp_cmd_build_info(dp, info->snd_portid,
1340 info->snd_seq, OVS_DP_CMD_NEW);
1341 if (IS_ERR(reply)) {
1342 err = PTR_ERR(reply);
1347 return genlmsg_reply(reply, info);
1354 static int ovs_dp_cmd_dump(struct sk_buff *skb, struct netlink_callback *cb)
1356 struct ovs_net *ovs_net = net_generic(sock_net(skb->sk), ovs_net_id);
1357 struct datapath *dp;
1358 int skip = cb->args[0];
1362 list_for_each_entry_rcu(dp, &ovs_net->dps, list_node) {
1364 ovs_dp_cmd_fill_info(dp, skb, NETLINK_CB(cb->skb).portid,
1365 cb->nlh->nlmsg_seq, NLM_F_MULTI,
1366 OVS_DP_CMD_NEW) < 0)
1377 static struct genl_ops dp_datapath_genl_ops[] = {
1378 { .cmd = OVS_DP_CMD_NEW,
1379 .flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
1380 .policy = datapath_policy,
1381 .doit = ovs_dp_cmd_new
1383 { .cmd = OVS_DP_CMD_DEL,
1384 .flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
1385 .policy = datapath_policy,
1386 .doit = ovs_dp_cmd_del
1388 { .cmd = OVS_DP_CMD_GET,
1389 .flags = 0, /* OK for unprivileged users. */
1390 .policy = datapath_policy,
1391 .doit = ovs_dp_cmd_get,
1392 .dumpit = ovs_dp_cmd_dump
1394 { .cmd = OVS_DP_CMD_SET,
1395 .flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
1396 .policy = datapath_policy,
1397 .doit = ovs_dp_cmd_set,
1401 static const struct nla_policy vport_policy[OVS_VPORT_ATTR_MAX + 1] = {
1402 [OVS_VPORT_ATTR_NAME] = { .type = NLA_NUL_STRING, .len = IFNAMSIZ - 1 },
1403 [OVS_VPORT_ATTR_STATS] = { .len = sizeof(struct ovs_vport_stats) },
1404 [OVS_VPORT_ATTR_PORT_NO] = { .type = NLA_U32 },
1405 [OVS_VPORT_ATTR_TYPE] = { .type = NLA_U32 },
1406 [OVS_VPORT_ATTR_UPCALL_PID] = { .type = NLA_U32 },
1407 [OVS_VPORT_ATTR_OPTIONS] = { .type = NLA_NESTED },
1410 static struct genl_family dp_vport_genl_family = {
1411 .id = GENL_ID_GENERATE,
1412 .hdrsize = sizeof(struct ovs_header),
1413 .name = OVS_VPORT_FAMILY,
1414 .version = OVS_VPORT_VERSION,
1415 .maxattr = OVS_VPORT_ATTR_MAX,
1420 struct genl_multicast_group ovs_dp_vport_multicast_group = {
1421 .name = OVS_VPORT_MCGROUP
1424 /* Called with ovs_mutex or RCU read lock. */
1425 static int ovs_vport_cmd_fill_info(struct vport *vport, struct sk_buff *skb,
1426 u32 portid, u32 seq, u32 flags, u8 cmd)
1428 struct ovs_header *ovs_header;
1429 struct ovs_vport_stats vport_stats;
1432 ovs_header = genlmsg_put(skb, portid, seq, &dp_vport_genl_family,
1437 ovs_header->dp_ifindex = get_dpifindex(vport->dp);
1439 if (nla_put_u32(skb, OVS_VPORT_ATTR_PORT_NO, vport->port_no) ||
1440 nla_put_u32(skb, OVS_VPORT_ATTR_TYPE, vport->ops->type) ||
1441 nla_put_string(skb, OVS_VPORT_ATTR_NAME, vport->ops->get_name(vport)) ||
1442 nla_put_u32(skb, OVS_VPORT_ATTR_UPCALL_PID, vport->upcall_portid))
1443 goto nla_put_failure;
1445 ovs_vport_get_stats(vport, &vport_stats);
1446 if (nla_put(skb, OVS_VPORT_ATTR_STATS, sizeof(struct ovs_vport_stats),
1448 goto nla_put_failure;
1450 err = ovs_vport_get_options(vport, skb);
1451 if (err == -EMSGSIZE)
1454 return genlmsg_end(skb, ovs_header);
1459 genlmsg_cancel(skb, ovs_header);
1463 /* Called with ovs_mutex or RCU read lock. */
1464 struct sk_buff *ovs_vport_cmd_build_info(struct vport *vport, u32 portid,
1467 struct sk_buff *skb;
1470 skb = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_ATOMIC);
1472 return ERR_PTR(-ENOMEM);
1474 retval = ovs_vport_cmd_fill_info(vport, skb, portid, seq, 0, cmd);
1480 /* Called with ovs_mutex or RCU read lock. */
1481 static struct vport *lookup_vport(struct net *net,
1482 struct ovs_header *ovs_header,
1483 struct nlattr *a[OVS_VPORT_ATTR_MAX + 1])
1485 struct datapath *dp;
1486 struct vport *vport;
1488 if (a[OVS_VPORT_ATTR_NAME]) {
1489 vport = ovs_vport_locate(net, nla_data(a[OVS_VPORT_ATTR_NAME]));
1491 return ERR_PTR(-ENODEV);
1492 if (ovs_header->dp_ifindex &&
1493 ovs_header->dp_ifindex != get_dpifindex(vport->dp))
1494 return ERR_PTR(-ENODEV);
1496 } else if (a[OVS_VPORT_ATTR_PORT_NO]) {
1497 u32 port_no = nla_get_u32(a[OVS_VPORT_ATTR_PORT_NO]);
1499 if (port_no >= DP_MAX_PORTS)
1500 return ERR_PTR(-EFBIG);
1502 dp = get_dp(net, ovs_header->dp_ifindex);
1504 return ERR_PTR(-ENODEV);
1506 vport = ovs_vport_ovsl_rcu(dp, port_no);
1508 return ERR_PTR(-ENODEV);
1511 return ERR_PTR(-EINVAL);
1514 static int ovs_vport_cmd_new(struct sk_buff *skb, struct genl_info *info)
1516 struct nlattr **a = info->attrs;
1517 struct ovs_header *ovs_header = info->userhdr;
1518 struct vport_parms parms;
1519 struct sk_buff *reply;
1520 struct vport *vport;
1521 struct datapath *dp;
1526 if (!a[OVS_VPORT_ATTR_NAME] || !a[OVS_VPORT_ATTR_TYPE] ||
1527 !a[OVS_VPORT_ATTR_UPCALL_PID])
1531 dp = get_dp(sock_net(skb->sk), ovs_header->dp_ifindex);
1536 if (a[OVS_VPORT_ATTR_PORT_NO]) {
1537 port_no = nla_get_u32(a[OVS_VPORT_ATTR_PORT_NO]);
1540 if (port_no >= DP_MAX_PORTS)
1543 vport = ovs_vport_ovsl(dp, port_no);
1548 for (port_no = 1; ; port_no++) {
1549 if (port_no >= DP_MAX_PORTS) {
1553 vport = ovs_vport_ovsl(dp, port_no);
1559 parms.name = nla_data(a[OVS_VPORT_ATTR_NAME]);
1560 parms.type = nla_get_u32(a[OVS_VPORT_ATTR_TYPE]);
1561 parms.options = a[OVS_VPORT_ATTR_OPTIONS];
1563 parms.port_no = port_no;
1564 parms.upcall_portid = nla_get_u32(a[OVS_VPORT_ATTR_UPCALL_PID]);
1566 vport = new_vport(&parms);
1567 err = PTR_ERR(vport);
1572 if (a[OVS_VPORT_ATTR_STATS])
1573 ovs_vport_set_stats(vport, nla_data(a[OVS_VPORT_ATTR_STATS]));
1575 reply = ovs_vport_cmd_build_info(vport, info->snd_portid, info->snd_seq,
1577 if (IS_ERR(reply)) {
1578 err = PTR_ERR(reply);
1579 ovs_dp_detach_port(vport);
1583 ovs_notify(reply, info, &ovs_dp_vport_multicast_group);
1591 static int ovs_vport_cmd_set(struct sk_buff *skb, struct genl_info *info)
1593 struct nlattr **a = info->attrs;
1594 struct sk_buff *reply;
1595 struct vport *vport;
1599 vport = lookup_vport(sock_net(skb->sk), info->userhdr, a);
1600 err = PTR_ERR(vport);
1604 if (a[OVS_VPORT_ATTR_TYPE] &&
1605 nla_get_u32(a[OVS_VPORT_ATTR_TYPE]) != vport->ops->type) {
1610 reply = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
1616 if (a[OVS_VPORT_ATTR_OPTIONS]) {
1617 err = ovs_vport_set_options(vport, a[OVS_VPORT_ATTR_OPTIONS]);
1622 if (a[OVS_VPORT_ATTR_STATS])
1623 ovs_vport_set_stats(vport, nla_data(a[OVS_VPORT_ATTR_STATS]));
1625 if (a[OVS_VPORT_ATTR_UPCALL_PID])
1626 vport->upcall_portid = nla_get_u32(a[OVS_VPORT_ATTR_UPCALL_PID]);
1628 err = ovs_vport_cmd_fill_info(vport, reply, info->snd_portid,
1629 info->snd_seq, 0, OVS_VPORT_CMD_NEW);
1633 ovs_notify(reply, info, &ovs_dp_vport_multicast_group);
1643 static int ovs_vport_cmd_del(struct sk_buff *skb, struct genl_info *info)
1645 struct nlattr **a = info->attrs;
1646 struct sk_buff *reply;
1647 struct vport *vport;
1651 vport = lookup_vport(sock_net(skb->sk), info->userhdr, a);
1652 err = PTR_ERR(vport);
1656 if (vport->port_no == OVSP_LOCAL) {
1661 reply = ovs_vport_cmd_build_info(vport, info->snd_portid,
1662 info->snd_seq, OVS_VPORT_CMD_DEL);
1663 err = PTR_ERR(reply);
1668 ovs_dp_detach_port(vport);
1670 ovs_notify(reply, info, &ovs_dp_vport_multicast_group);
1677 static int ovs_vport_cmd_get(struct sk_buff *skb, struct genl_info *info)
1679 struct nlattr **a = info->attrs;
1680 struct ovs_header *ovs_header = info->userhdr;
1681 struct sk_buff *reply;
1682 struct vport *vport;
1686 vport = lookup_vport(sock_net(skb->sk), ovs_header, a);
1687 err = PTR_ERR(vport);
1691 reply = ovs_vport_cmd_build_info(vport, info->snd_portid,
1692 info->snd_seq, OVS_VPORT_CMD_NEW);
1693 err = PTR_ERR(reply);
1699 return genlmsg_reply(reply, info);
1706 static int ovs_vport_cmd_dump(struct sk_buff *skb, struct netlink_callback *cb)
1708 struct ovs_header *ovs_header = genlmsg_data(nlmsg_data(cb->nlh));
1709 struct datapath *dp;
1710 int bucket = cb->args[0], skip = cb->args[1];
1713 dp = get_dp(sock_net(skb->sk), ovs_header->dp_ifindex);
1718 for (i = bucket; i < DP_VPORT_HASH_BUCKETS; i++) {
1719 struct vport *vport;
1722 hlist_for_each_entry_rcu(vport, &dp->ports[i], dp_hash_node) {
1724 ovs_vport_cmd_fill_info(vport, skb,
1725 NETLINK_CB(cb->skb).portid,
1728 OVS_VPORT_CMD_NEW) < 0)
1744 static struct genl_ops dp_vport_genl_ops[] = {
1745 { .cmd = OVS_VPORT_CMD_NEW,
1746 .flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
1747 .policy = vport_policy,
1748 .doit = ovs_vport_cmd_new
1750 { .cmd = OVS_VPORT_CMD_DEL,
1751 .flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
1752 .policy = vport_policy,
1753 .doit = ovs_vport_cmd_del
1755 { .cmd = OVS_VPORT_CMD_GET,
1756 .flags = 0, /* OK for unprivileged users. */
1757 .policy = vport_policy,
1758 .doit = ovs_vport_cmd_get,
1759 .dumpit = ovs_vport_cmd_dump
1761 { .cmd = OVS_VPORT_CMD_SET,
1762 .flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
1763 .policy = vport_policy,
1764 .doit = ovs_vport_cmd_set,
1768 struct genl_family_and_ops {
1769 struct genl_family *family;
1770 struct genl_ops *ops;
1772 struct genl_multicast_group *group;
1775 static const struct genl_family_and_ops dp_genl_families[] = {
1776 { &dp_datapath_genl_family,
1777 dp_datapath_genl_ops, ARRAY_SIZE(dp_datapath_genl_ops),
1778 &ovs_dp_datapath_multicast_group },
1779 { &dp_vport_genl_family,
1780 dp_vport_genl_ops, ARRAY_SIZE(dp_vport_genl_ops),
1781 &ovs_dp_vport_multicast_group },
1782 { &dp_flow_genl_family,
1783 dp_flow_genl_ops, ARRAY_SIZE(dp_flow_genl_ops),
1784 &ovs_dp_flow_multicast_group },
1785 { &dp_packet_genl_family,
1786 dp_packet_genl_ops, ARRAY_SIZE(dp_packet_genl_ops),
1790 static void dp_unregister_genl(int n_families)
1794 for (i = 0; i < n_families; i++)
1795 genl_unregister_family(dp_genl_families[i].family);
1798 static int dp_register_genl(void)
1805 for (i = 0; i < ARRAY_SIZE(dp_genl_families); i++) {
1806 const struct genl_family_and_ops *f = &dp_genl_families[i];
1808 err = genl_register_family_with_ops(f->family, f->ops,
1815 err = genl_register_mc_group(f->family, f->group);
1824 dp_unregister_genl(n_registered);
1828 static int __net_init ovs_init_net(struct net *net)
1830 struct ovs_net *ovs_net = net_generic(net, ovs_net_id);
1832 INIT_LIST_HEAD(&ovs_net->dps);
1833 INIT_WORK(&ovs_net->dp_notify_work, ovs_dp_notify_wq);
1837 static void __net_exit ovs_exit_net(struct net *net)
1839 struct datapath *dp, *dp_next;
1840 struct ovs_net *ovs_net = net_generic(net, ovs_net_id);
1843 list_for_each_entry_safe(dp, dp_next, &ovs_net->dps, list_node)
1847 cancel_work_sync(&ovs_net->dp_notify_work);
1850 static struct pernet_operations ovs_net_ops = {
1851 .init = ovs_init_net,
1852 .exit = ovs_exit_net,
1854 .size = sizeof(struct ovs_net),
1857 DEFINE_COMPAT_PNET_REG_FUNC(device);
1859 static int __init dp_init(void)
1863 BUILD_BUG_ON(sizeof(struct ovs_skb_cb) > FIELD_SIZEOF(struct sk_buff, cb));
1865 pr_info("Open vSwitch switching datapath %s, built "__DATE__" "__TIME__"\n",
1868 err = ovs_flow_init();
1872 err = ovs_vport_init();
1874 goto error_flow_exit;
1876 err = register_pernet_device(&ovs_net_ops);
1878 goto error_vport_exit;
1880 err = register_netdevice_notifier(&ovs_dp_device_notifier);
1882 goto error_netns_exit;
1884 err = dp_register_genl();
1886 goto error_unreg_notifier;
1890 error_unreg_notifier:
1891 unregister_netdevice_notifier(&ovs_dp_device_notifier);
1893 unregister_pernet_device(&ovs_net_ops);
1902 static void dp_cleanup(void)
1904 dp_unregister_genl(ARRAY_SIZE(dp_genl_families));
1905 unregister_netdevice_notifier(&ovs_dp_device_notifier);
1906 unregister_pernet_device(&ovs_net_ops);
1912 module_init(dp_init);
1913 module_exit(dp_cleanup);
1915 MODULE_DESCRIPTION("Open vSwitch switching datapath");
1916 MODULE_LICENSE("GPL");
1917 MODULE_VERSION(VERSION);