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 */
407 struct genl_info info
= {
408 #if LINUX_VERSION_CODE >= KERNEL_VERSION(3,14,0)
409 .dst_sk
= net
->genl_sock
,
411 .snd_portid
= upcall_info
->portid
,
416 if (vlan_tx_tag_present(skb
)) {
417 nskb
= skb_clone(skb
, GFP_ATOMIC
);
421 nskb
= __vlan_put_tag(nskb
, nskb
->vlan_proto
, vlan_tx_tag_get(nskb
));
425 vlan_set_tci(nskb
, 0);
430 if (nla_attr_size(skb
->len
) > USHRT_MAX
) {
435 len
= upcall_msg_size(skb
, upcall_info
->userdata
);
436 user_skb
= genlmsg_new_unicast(len
, &info
, GFP_ATOMIC
);
442 upcall
= genlmsg_put(user_skb
, 0, 0, &dp_packet_genl_family
,
443 0, upcall_info
->cmd
);
444 upcall
->dp_ifindex
= dp_ifindex
;
446 nla
= nla_nest_start(user_skb
, OVS_PACKET_ATTR_KEY
);
447 ovs_nla_put_flow(upcall_info
->key
, upcall_info
->key
, user_skb
);
448 nla_nest_end(user_skb
, nla
);
450 if (upcall_info
->userdata
)
451 __nla_put(user_skb
, OVS_PACKET_ATTR_USERDATA
,
452 nla_len(upcall_info
->userdata
),
453 nla_data(upcall_info
->userdata
));
455 nla
= __nla_reserve(user_skb
, OVS_PACKET_ATTR_PACKET
, skb
->len
);
457 skb_copy_and_csum_dev(skb
, nla_data(nla
));
459 genlmsg_end(user_skb
, upcall
);
460 err
= genlmsg_unicast(net
, user_skb
, upcall_info
->portid
);
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(false);
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
;
643 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 ovs_flow_stats_get(flow
, &stats
, &used
, &tcp_flags
);
675 nla_put_u64(skb
, OVS_FLOW_ATTR_USED
, ovs_flow_used_time(used
)))
676 goto nla_put_failure
;
678 if (stats
.n_packets
&&
679 nla_put(skb
, OVS_FLOW_ATTR_STATS
, sizeof(struct ovs_flow_stats
), &stats
))
680 goto nla_put_failure
;
682 if ((u8
)ntohs(tcp_flags
) &&
683 nla_put_u8(skb
, OVS_FLOW_ATTR_TCP_FLAGS
, (u8
)ntohs(tcp_flags
)))
684 goto nla_put_failure
;
686 /* If OVS_FLOW_ATTR_ACTIONS doesn't fit, skip dumping the actions if
687 * this is the first flow to be dumped into 'skb'. This is unusual for
688 * Netlink but individual action lists can be longer than
689 * NLMSG_GOODSIZE and thus entirely undumpable if we didn't do this.
690 * The userspace caller can always fetch the actions separately if it
691 * really wants them. (Most userspace callers in fact don't care.)
693 * This can only fail for dump operations because the skb is always
694 * properly sized for single flows.
696 start
= nla_nest_start(skb
, OVS_FLOW_ATTR_ACTIONS
);
698 const struct sw_flow_actions
*sf_acts
;
700 sf_acts
= rcu_dereference_ovsl(flow
->sf_acts
);
702 err
= ovs_nla_put_actions(sf_acts
->actions
,
703 sf_acts
->actions_len
, skb
);
705 nla_nest_end(skb
, start
);
710 nla_nest_cancel(skb
, start
);
712 } else if (skb_orig_len
)
713 goto nla_put_failure
;
715 return genlmsg_end(skb
, ovs_header
);
720 genlmsg_cancel(skb
, ovs_header
);
724 static struct sk_buff
*ovs_flow_cmd_alloc_info(struct sw_flow
*flow
,
725 struct genl_info
*info
)
729 len
= ovs_flow_cmd_msg_size(ovsl_dereference(flow
->sf_acts
));
731 return genlmsg_new_unicast(len
, info
, GFP_KERNEL
);
734 static struct sk_buff
*ovs_flow_cmd_build_info(struct sw_flow
*flow
,
736 struct genl_info
*info
,
742 skb
= ovs_flow_cmd_alloc_info(flow
, info
);
744 return ERR_PTR(-ENOMEM
);
746 retval
= ovs_flow_cmd_fill_info(flow
, dp
, skb
, info
->snd_portid
,
747 info
->snd_seq
, 0, cmd
);
752 static int ovs_flow_cmd_new_or_set(struct sk_buff
*skb
, struct genl_info
*info
)
754 struct nlattr
**a
= info
->attrs
;
755 struct ovs_header
*ovs_header
= info
->userhdr
;
756 struct sw_flow_key key
, masked_key
;
757 struct sw_flow
*flow
= NULL
;
758 struct sw_flow_mask mask
;
759 struct sk_buff
*reply
;
761 struct sw_flow_actions
*acts
= NULL
;
762 struct sw_flow_match match
;
768 if (!a
[OVS_FLOW_ATTR_KEY
])
771 ovs_match_init(&match
, &key
, &mask
);
772 error
= ovs_nla_get_match(&match
, &exact_5tuple
,
773 a
[OVS_FLOW_ATTR_KEY
], a
[OVS_FLOW_ATTR_MASK
]);
777 /* Validate actions. */
778 if (a
[OVS_FLOW_ATTR_ACTIONS
]) {
779 acts
= ovs_nla_alloc_flow_actions(nla_len(a
[OVS_FLOW_ATTR_ACTIONS
]));
780 error
= PTR_ERR(acts
);
784 ovs_flow_mask_key(&masked_key
, &key
, &mask
);
785 error
= ovs_nla_copy_actions(a
[OVS_FLOW_ATTR_ACTIONS
],
786 &masked_key
, 0, &acts
);
788 OVS_NLERR("Flow actions may not be safe on all matching packets.\n");
791 } else if (info
->genlhdr
->cmd
== OVS_FLOW_CMD_NEW
) {
797 dp
= get_dp(sock_net(skb
->sk
), ovs_header
->dp_ifindex
);
802 /* Check if this is a duplicate flow */
803 flow
= ovs_flow_tbl_lookup(&dp
->table
, &key
);
805 /* Bail out if we're not allowed to create a new flow. */
807 if (info
->genlhdr
->cmd
== OVS_FLOW_CMD_SET
)
811 flow
= ovs_flow_alloc(!exact_5tuple
);
813 error
= PTR_ERR(flow
);
817 flow
->key
= masked_key
;
818 flow
->unmasked_key
= key
;
819 rcu_assign_pointer(flow
->sf_acts
, acts
);
821 /* Put flow in bucket. */
822 error
= ovs_flow_tbl_insert(&dp
->table
, flow
, &mask
);
828 reply
= ovs_flow_cmd_build_info(flow
, dp
, info
, OVS_FLOW_CMD_NEW
);
830 /* We found a matching flow. */
831 struct sw_flow_actions
*old_acts
;
833 /* Bail out if we're not allowed to modify an existing flow.
834 * We accept NLM_F_CREATE in place of the intended NLM_F_EXCL
835 * because Generic Netlink treats the latter as a dump
836 * request. We also accept NLM_F_EXCL in case that bug ever
840 if (info
->genlhdr
->cmd
== OVS_FLOW_CMD_NEW
&&
841 info
->nlhdr
->nlmsg_flags
& (NLM_F_CREATE
| NLM_F_EXCL
))
844 /* The unmasked key has to be the same for flow updates. */
846 if (!ovs_flow_cmp_unmasked_key(flow
, &match
)) {
847 OVS_NLERR("Flow modification message rejected, unmasked key does not match.\n");
851 /* Update actions. */
852 old_acts
= ovsl_dereference(flow
->sf_acts
);
853 rcu_assign_pointer(flow
->sf_acts
, acts
);
854 ovs_nla_free_flow_actions(old_acts
);
856 reply
= ovs_flow_cmd_build_info(flow
, dp
, info
, OVS_FLOW_CMD_NEW
);
859 if (a
[OVS_FLOW_ATTR_CLEAR
])
860 ovs_flow_stats_clear(flow
);
865 ovs_notify(reply
, info
, &ovs_dp_flow_multicast_group
);
867 netlink_set_err(sock_net(skb
->sk
)->genl_sock
, 0,
868 ovs_dp_flow_multicast_group
.id
, PTR_ERR(reply
));
872 ovs_flow_free(flow
, false);
881 static int ovs_flow_cmd_get(struct sk_buff
*skb
, struct genl_info
*info
)
883 struct nlattr
**a
= info
->attrs
;
884 struct ovs_header
*ovs_header
= info
->userhdr
;
885 struct sw_flow_key key
;
886 struct sk_buff
*reply
;
887 struct sw_flow
*flow
;
889 struct sw_flow_match match
;
892 if (!a
[OVS_FLOW_ATTR_KEY
]) {
893 OVS_NLERR("Flow get message rejected, Key attribute missing.\n");
897 ovs_match_init(&match
, &key
, NULL
);
898 err
= ovs_nla_get_match(&match
, NULL
, a
[OVS_FLOW_ATTR_KEY
], NULL
);
903 dp
= get_dp(sock_net(skb
->sk
), ovs_header
->dp_ifindex
);
909 flow
= ovs_flow_tbl_lookup(&dp
->table
, &key
);
910 if (!flow
|| !ovs_flow_cmp_unmasked_key(flow
, &match
)) {
915 reply
= ovs_flow_cmd_build_info(flow
, dp
, info
, OVS_FLOW_CMD_NEW
);
917 err
= PTR_ERR(reply
);
922 return genlmsg_reply(reply
, info
);
928 static int ovs_flow_cmd_del(struct sk_buff
*skb
, struct genl_info
*info
)
930 struct nlattr
**a
= info
->attrs
;
931 struct ovs_header
*ovs_header
= info
->userhdr
;
932 struct sw_flow_key key
;
933 struct sk_buff
*reply
;
934 struct sw_flow
*flow
;
936 struct sw_flow_match match
;
940 dp
= get_dp(sock_net(skb
->sk
), ovs_header
->dp_ifindex
);
946 if (!a
[OVS_FLOW_ATTR_KEY
]) {
947 err
= ovs_flow_tbl_flush(&dp
->table
);
951 ovs_match_init(&match
, &key
, NULL
);
952 err
= ovs_nla_get_match(&match
, NULL
, a
[OVS_FLOW_ATTR_KEY
], NULL
);
956 flow
= ovs_flow_tbl_lookup(&dp
->table
, &key
);
957 if (!flow
|| !ovs_flow_cmp_unmasked_key(flow
, &match
)) {
962 reply
= ovs_flow_cmd_alloc_info(flow
, info
);
968 ovs_flow_tbl_remove(&dp
->table
, flow
);
970 err
= ovs_flow_cmd_fill_info(flow
, dp
, reply
, info
->snd_portid
,
971 info
->snd_seq
, 0, OVS_FLOW_CMD_DEL
);
974 ovs_flow_free(flow
, true);
977 ovs_notify(reply
, info
, &ovs_dp_flow_multicast_group
);
984 static int ovs_flow_cmd_dump(struct sk_buff
*skb
, struct netlink_callback
*cb
)
986 struct ovs_header
*ovs_header
= genlmsg_data(nlmsg_data(cb
->nlh
));
987 struct table_instance
*ti
;
991 dp
= get_dp(sock_net(skb
->sk
), ovs_header
->dp_ifindex
);
997 ti
= rcu_dereference(dp
->table
.ti
);
999 struct sw_flow
*flow
;
1002 bucket
= cb
->args
[0];
1004 flow
= ovs_flow_tbl_dump_next(ti
, &bucket
, &obj
);
1008 if (ovs_flow_cmd_fill_info(flow
, dp
, skb
,
1009 NETLINK_CB(cb
->skb
).portid
,
1010 cb
->nlh
->nlmsg_seq
, NLM_F_MULTI
,
1011 OVS_FLOW_CMD_NEW
) < 0)
1014 cb
->args
[0] = bucket
;
1021 static struct genl_ops dp_flow_genl_ops
[] = {
1022 { .cmd
= OVS_FLOW_CMD_NEW
,
1023 .flags
= GENL_ADMIN_PERM
, /* Requires CAP_NET_ADMIN privilege. */
1024 .policy
= flow_policy
,
1025 .doit
= ovs_flow_cmd_new_or_set
1027 { .cmd
= OVS_FLOW_CMD_DEL
,
1028 .flags
= GENL_ADMIN_PERM
, /* Requires CAP_NET_ADMIN privilege. */
1029 .policy
= flow_policy
,
1030 .doit
= ovs_flow_cmd_del
1032 { .cmd
= OVS_FLOW_CMD_GET
,
1033 .flags
= 0, /* OK for unprivileged users. */
1034 .policy
= flow_policy
,
1035 .doit
= ovs_flow_cmd_get
,
1036 .dumpit
= ovs_flow_cmd_dump
1038 { .cmd
= OVS_FLOW_CMD_SET
,
1039 .flags
= GENL_ADMIN_PERM
, /* Requires CAP_NET_ADMIN privilege. */
1040 .policy
= flow_policy
,
1041 .doit
= ovs_flow_cmd_new_or_set
,
1045 static const struct nla_policy datapath_policy
[OVS_DP_ATTR_MAX
+ 1] = {
1046 [OVS_DP_ATTR_NAME
] = { .type
= NLA_NUL_STRING
, .len
= IFNAMSIZ
- 1 },
1047 [OVS_DP_ATTR_UPCALL_PID
] = { .type
= NLA_U32
},
1050 static struct genl_family dp_datapath_genl_family
= {
1051 .id
= GENL_ID_GENERATE
,
1052 .hdrsize
= sizeof(struct ovs_header
),
1053 .name
= OVS_DATAPATH_FAMILY
,
1054 .version
= OVS_DATAPATH_VERSION
,
1055 .maxattr
= OVS_DP_ATTR_MAX
,
1060 static struct genl_multicast_group ovs_dp_datapath_multicast_group
= {
1061 .name
= OVS_DATAPATH_MCGROUP
1064 static size_t ovs_dp_cmd_msg_size(void)
1066 size_t msgsize
= NLMSG_ALIGN(sizeof(struct ovs_header
));
1068 msgsize
+= nla_total_size(IFNAMSIZ
);
1069 msgsize
+= nla_total_size(sizeof(struct ovs_dp_stats
));
1070 msgsize
+= nla_total_size(sizeof(struct ovs_dp_megaflow_stats
));
1075 static int ovs_dp_cmd_fill_info(struct datapath
*dp
, struct sk_buff
*skb
,
1076 u32 portid
, u32 seq
, u32 flags
, u8 cmd
)
1078 struct ovs_header
*ovs_header
;
1079 struct ovs_dp_stats dp_stats
;
1080 struct ovs_dp_megaflow_stats dp_megaflow_stats
;
1083 ovs_header
= genlmsg_put(skb
, portid
, seq
, &dp_datapath_genl_family
,
1088 ovs_header
->dp_ifindex
= get_dpifindex(dp
);
1091 err
= nla_put_string(skb
, OVS_DP_ATTR_NAME
, ovs_dp_name(dp
));
1094 goto nla_put_failure
;
1096 get_dp_stats(dp
, &dp_stats
, &dp_megaflow_stats
);
1097 if (nla_put(skb
, OVS_DP_ATTR_STATS
, sizeof(struct ovs_dp_stats
),
1099 goto nla_put_failure
;
1101 if (nla_put(skb
, OVS_DP_ATTR_MEGAFLOW_STATS
,
1102 sizeof(struct ovs_dp_megaflow_stats
),
1103 &dp_megaflow_stats
))
1104 goto nla_put_failure
;
1106 return genlmsg_end(skb
, ovs_header
);
1109 genlmsg_cancel(skb
, ovs_header
);
1114 static struct sk_buff
*ovs_dp_cmd_build_info(struct datapath
*dp
,
1115 struct genl_info
*info
, u8 cmd
)
1117 struct sk_buff
*skb
;
1120 skb
= genlmsg_new_unicast(ovs_dp_cmd_msg_size(), info
, GFP_KERNEL
);
1122 return ERR_PTR(-ENOMEM
);
1124 retval
= ovs_dp_cmd_fill_info(dp
, skb
, info
->snd_portid
, info
->snd_seq
, 0, cmd
);
1127 return ERR_PTR(retval
);
1132 /* Called with ovs_mutex. */
1133 static struct datapath
*lookup_datapath(struct net
*net
,
1134 struct ovs_header
*ovs_header
,
1135 struct nlattr
*a
[OVS_DP_ATTR_MAX
+ 1])
1137 struct datapath
*dp
;
1139 if (!a
[OVS_DP_ATTR_NAME
])
1140 dp
= get_dp(net
, ovs_header
->dp_ifindex
);
1142 struct vport
*vport
;
1145 vport
= ovs_vport_locate(net
, nla_data(a
[OVS_DP_ATTR_NAME
]));
1146 dp
= vport
&& vport
->port_no
== OVSP_LOCAL
? vport
->dp
: NULL
;
1149 return dp
? dp
: ERR_PTR(-ENODEV
);
1152 static int ovs_dp_cmd_new(struct sk_buff
*skb
, struct genl_info
*info
)
1154 struct nlattr
**a
= info
->attrs
;
1155 struct vport_parms parms
;
1156 struct sk_buff
*reply
;
1157 struct datapath
*dp
;
1158 struct vport
*vport
;
1159 struct ovs_net
*ovs_net
;
1163 if (!a
[OVS_DP_ATTR_NAME
] || !a
[OVS_DP_ATTR_UPCALL_PID
])
1169 dp
= kzalloc(sizeof(*dp
), GFP_KERNEL
);
1171 goto err_unlock_ovs
;
1173 ovs_dp_set_net(dp
, hold_net(sock_net(skb
->sk
)));
1175 /* Allocate table. */
1176 err
= ovs_flow_tbl_init(&dp
->table
);
1180 dp
->stats_percpu
= alloc_percpu(struct dp_stats_percpu
);
1181 if (!dp
->stats_percpu
) {
1183 goto err_destroy_table
;
1186 dp
->ports
= kmalloc(DP_VPORT_HASH_BUCKETS
* sizeof(struct hlist_head
),
1190 goto err_destroy_percpu
;
1193 for (i
= 0; i
< DP_VPORT_HASH_BUCKETS
; i
++)
1194 INIT_HLIST_HEAD(&dp
->ports
[i
]);
1196 /* Set up our datapath device. */
1197 parms
.name
= nla_data(a
[OVS_DP_ATTR_NAME
]);
1198 parms
.type
= OVS_VPORT_TYPE_INTERNAL
;
1199 parms
.options
= NULL
;
1201 parms
.port_no
= OVSP_LOCAL
;
1202 parms
.upcall_portid
= nla_get_u32(a
[OVS_DP_ATTR_UPCALL_PID
]);
1204 vport
= new_vport(&parms
);
1205 if (IS_ERR(vport
)) {
1206 err
= PTR_ERR(vport
);
1210 goto err_destroy_ports_array
;
1213 reply
= ovs_dp_cmd_build_info(dp
, info
, OVS_DP_CMD_NEW
);
1214 err
= PTR_ERR(reply
);
1216 goto err_destroy_local_port
;
1218 ovs_net
= net_generic(ovs_dp_get_net(dp
), ovs_net_id
);
1219 list_add_tail_rcu(&dp
->list_node
, &ovs_net
->dps
);
1223 ovs_notify(reply
, info
, &ovs_dp_datapath_multicast_group
);
1226 err_destroy_local_port
:
1227 ovs_dp_detach_port(ovs_vport_ovsl(dp
, OVSP_LOCAL
));
1228 err_destroy_ports_array
:
1231 free_percpu(dp
->stats_percpu
);
1233 ovs_flow_tbl_destroy(&dp
->table
);
1235 release_net(ovs_dp_get_net(dp
));
1243 /* Called with ovs_mutex. */
1244 static void __dp_destroy(struct datapath
*dp
)
1248 for (i
= 0; i
< DP_VPORT_HASH_BUCKETS
; i
++) {
1249 struct vport
*vport
;
1250 struct hlist_node
*n
;
1252 hlist_for_each_entry_safe(vport
, n
, &dp
->ports
[i
], dp_hash_node
)
1253 if (vport
->port_no
!= OVSP_LOCAL
)
1254 ovs_dp_detach_port(vport
);
1257 list_del_rcu(&dp
->list_node
);
1259 /* OVSP_LOCAL is datapath internal port. We need to make sure that
1260 * all port in datapath are destroyed first before freeing datapath.
1262 ovs_dp_detach_port(ovs_vport_ovsl(dp
, OVSP_LOCAL
));
1264 call_rcu(&dp
->rcu
, destroy_dp_rcu
);
1267 static int ovs_dp_cmd_del(struct sk_buff
*skb
, struct genl_info
*info
)
1269 struct sk_buff
*reply
;
1270 struct datapath
*dp
;
1274 dp
= lookup_datapath(sock_net(skb
->sk
), info
->userhdr
, info
->attrs
);
1279 reply
= ovs_dp_cmd_build_info(dp
, info
, 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
, OVS_DP_CMD_NEW
);
1308 if (IS_ERR(reply
)) {
1309 err
= PTR_ERR(reply
);
1310 netlink_set_err(sock_net(skb
->sk
)->genl_sock
, 0,
1311 ovs_dp_datapath_multicast_group
.id
, err
);
1317 ovs_notify(reply
, info
, &ovs_dp_datapath_multicast_group
);
1325 static int ovs_dp_cmd_get(struct sk_buff
*skb
, struct genl_info
*info
)
1327 struct sk_buff
*reply
;
1328 struct datapath
*dp
;
1332 dp
= lookup_datapath(sock_net(skb
->sk
), info
->userhdr
, info
->attrs
);
1338 reply
= ovs_dp_cmd_build_info(dp
, info
, OVS_DP_CMD_NEW
);
1339 if (IS_ERR(reply
)) {
1340 err
= PTR_ERR(reply
);
1345 return genlmsg_reply(reply
, info
);
1352 static int ovs_dp_cmd_dump(struct sk_buff
*skb
, struct netlink_callback
*cb
)
1354 struct ovs_net
*ovs_net
= net_generic(sock_net(skb
->sk
), ovs_net_id
);
1355 struct datapath
*dp
;
1356 int skip
= cb
->args
[0];
1360 list_for_each_entry_rcu(dp
, &ovs_net
->dps
, list_node
) {
1362 ovs_dp_cmd_fill_info(dp
, skb
, NETLINK_CB(cb
->skb
).portid
,
1363 cb
->nlh
->nlmsg_seq
, NLM_F_MULTI
,
1364 OVS_DP_CMD_NEW
) < 0)
1375 static struct genl_ops dp_datapath_genl_ops
[] = {
1376 { .cmd
= OVS_DP_CMD_NEW
,
1377 .flags
= GENL_ADMIN_PERM
, /* Requires CAP_NET_ADMIN privilege. */
1378 .policy
= datapath_policy
,
1379 .doit
= ovs_dp_cmd_new
1381 { .cmd
= OVS_DP_CMD_DEL
,
1382 .flags
= GENL_ADMIN_PERM
, /* Requires CAP_NET_ADMIN privilege. */
1383 .policy
= datapath_policy
,
1384 .doit
= ovs_dp_cmd_del
1386 { .cmd
= OVS_DP_CMD_GET
,
1387 .flags
= 0, /* OK for unprivileged users. */
1388 .policy
= datapath_policy
,
1389 .doit
= ovs_dp_cmd_get
,
1390 .dumpit
= ovs_dp_cmd_dump
1392 { .cmd
= OVS_DP_CMD_SET
,
1393 .flags
= GENL_ADMIN_PERM
, /* Requires CAP_NET_ADMIN privilege. */
1394 .policy
= datapath_policy
,
1395 .doit
= ovs_dp_cmd_set
,
1399 static const struct nla_policy vport_policy
[OVS_VPORT_ATTR_MAX
+ 1] = {
1400 [OVS_VPORT_ATTR_NAME
] = { .type
= NLA_NUL_STRING
, .len
= IFNAMSIZ
- 1 },
1401 [OVS_VPORT_ATTR_STATS
] = { .len
= sizeof(struct ovs_vport_stats
) },
1402 [OVS_VPORT_ATTR_PORT_NO
] = { .type
= NLA_U32
},
1403 [OVS_VPORT_ATTR_TYPE
] = { .type
= NLA_U32
},
1404 [OVS_VPORT_ATTR_UPCALL_PID
] = { .type
= NLA_U32
},
1405 [OVS_VPORT_ATTR_OPTIONS
] = { .type
= NLA_NESTED
},
1408 static struct genl_family dp_vport_genl_family
= {
1409 .id
= GENL_ID_GENERATE
,
1410 .hdrsize
= sizeof(struct ovs_header
),
1411 .name
= OVS_VPORT_FAMILY
,
1412 .version
= OVS_VPORT_VERSION
,
1413 .maxattr
= OVS_VPORT_ATTR_MAX
,
1418 struct genl_multicast_group ovs_dp_vport_multicast_group
= {
1419 .name
= OVS_VPORT_MCGROUP
1422 /* Called with ovs_mutex or RCU read lock. */
1423 static int ovs_vport_cmd_fill_info(struct vport
*vport
, struct sk_buff
*skb
,
1424 u32 portid
, u32 seq
, u32 flags
, u8 cmd
)
1426 struct ovs_header
*ovs_header
;
1427 struct ovs_vport_stats vport_stats
;
1430 ovs_header
= genlmsg_put(skb
, portid
, seq
, &dp_vport_genl_family
,
1435 ovs_header
->dp_ifindex
= get_dpifindex(vport
->dp
);
1437 if (nla_put_u32(skb
, OVS_VPORT_ATTR_PORT_NO
, vport
->port_no
) ||
1438 nla_put_u32(skb
, OVS_VPORT_ATTR_TYPE
, vport
->ops
->type
) ||
1439 nla_put_string(skb
, OVS_VPORT_ATTR_NAME
, vport
->ops
->get_name(vport
)) ||
1440 nla_put_u32(skb
, OVS_VPORT_ATTR_UPCALL_PID
, vport
->upcall_portid
))
1441 goto nla_put_failure
;
1443 ovs_vport_get_stats(vport
, &vport_stats
);
1444 if (nla_put(skb
, OVS_VPORT_ATTR_STATS
, sizeof(struct ovs_vport_stats
),
1446 goto nla_put_failure
;
1448 err
= ovs_vport_get_options(vport
, skb
);
1449 if (err
== -EMSGSIZE
)
1452 return genlmsg_end(skb
, ovs_header
);
1457 genlmsg_cancel(skb
, ovs_header
);
1461 /* Called with ovs_mutex or RCU read lock. */
1462 struct sk_buff
*ovs_vport_cmd_build_info(struct vport
*vport
, u32 portid
,
1465 struct sk_buff
*skb
;
1468 skb
= nlmsg_new(NLMSG_DEFAULT_SIZE
, GFP_ATOMIC
);
1470 return ERR_PTR(-ENOMEM
);
1472 retval
= ovs_vport_cmd_fill_info(vport
, skb
, portid
, seq
, 0, cmd
);
1478 /* Called with ovs_mutex or RCU read lock. */
1479 static struct vport
*lookup_vport(struct net
*net
,
1480 struct ovs_header
*ovs_header
,
1481 struct nlattr
*a
[OVS_VPORT_ATTR_MAX
+ 1])
1483 struct datapath
*dp
;
1484 struct vport
*vport
;
1486 if (a
[OVS_VPORT_ATTR_NAME
]) {
1487 vport
= ovs_vport_locate(net
, nla_data(a
[OVS_VPORT_ATTR_NAME
]));
1489 return ERR_PTR(-ENODEV
);
1490 if (ovs_header
->dp_ifindex
&&
1491 ovs_header
->dp_ifindex
!= get_dpifindex(vport
->dp
))
1492 return ERR_PTR(-ENODEV
);
1494 } else if (a
[OVS_VPORT_ATTR_PORT_NO
]) {
1495 u32 port_no
= nla_get_u32(a
[OVS_VPORT_ATTR_PORT_NO
]);
1497 if (port_no
>= DP_MAX_PORTS
)
1498 return ERR_PTR(-EFBIG
);
1500 dp
= get_dp(net
, ovs_header
->dp_ifindex
);
1502 return ERR_PTR(-ENODEV
);
1504 vport
= ovs_vport_ovsl_rcu(dp
, port_no
);
1506 return ERR_PTR(-ENODEV
);
1509 return ERR_PTR(-EINVAL
);
1512 static int ovs_vport_cmd_new(struct sk_buff
*skb
, struct genl_info
*info
)
1514 struct nlattr
**a
= info
->attrs
;
1515 struct ovs_header
*ovs_header
= info
->userhdr
;
1516 struct vport_parms parms
;
1517 struct sk_buff
*reply
;
1518 struct vport
*vport
;
1519 struct datapath
*dp
;
1524 if (!a
[OVS_VPORT_ATTR_NAME
] || !a
[OVS_VPORT_ATTR_TYPE
] ||
1525 !a
[OVS_VPORT_ATTR_UPCALL_PID
])
1529 dp
= get_dp(sock_net(skb
->sk
), ovs_header
->dp_ifindex
);
1534 if (a
[OVS_VPORT_ATTR_PORT_NO
]) {
1535 port_no
= nla_get_u32(a
[OVS_VPORT_ATTR_PORT_NO
]);
1538 if (port_no
>= DP_MAX_PORTS
)
1541 vport
= ovs_vport_ovsl(dp
, port_no
);
1546 for (port_no
= 1; ; port_no
++) {
1547 if (port_no
>= DP_MAX_PORTS
) {
1551 vport
= ovs_vport_ovsl(dp
, port_no
);
1557 parms
.name
= nla_data(a
[OVS_VPORT_ATTR_NAME
]);
1558 parms
.type
= nla_get_u32(a
[OVS_VPORT_ATTR_TYPE
]);
1559 parms
.options
= a
[OVS_VPORT_ATTR_OPTIONS
];
1561 parms
.port_no
= port_no
;
1562 parms
.upcall_portid
= nla_get_u32(a
[OVS_VPORT_ATTR_UPCALL_PID
]);
1564 vport
= new_vport(&parms
);
1565 err
= PTR_ERR(vport
);
1570 if (a
[OVS_VPORT_ATTR_STATS
])
1571 ovs_vport_set_stats(vport
, nla_data(a
[OVS_VPORT_ATTR_STATS
]));
1573 reply
= ovs_vport_cmd_build_info(vport
, info
->snd_portid
, info
->snd_seq
,
1575 if (IS_ERR(reply
)) {
1576 err
= PTR_ERR(reply
);
1577 ovs_dp_detach_port(vport
);
1581 ovs_notify(reply
, info
, &ovs_dp_vport_multicast_group
);
1589 static int ovs_vport_cmd_set(struct sk_buff
*skb
, struct genl_info
*info
)
1591 struct nlattr
**a
= info
->attrs
;
1592 struct sk_buff
*reply
;
1593 struct vport
*vport
;
1597 vport
= lookup_vport(sock_net(skb
->sk
), info
->userhdr
, a
);
1598 err
= PTR_ERR(vport
);
1602 if (a
[OVS_VPORT_ATTR_TYPE
] &&
1603 nla_get_u32(a
[OVS_VPORT_ATTR_TYPE
]) != vport
->ops
->type
) {
1608 reply
= nlmsg_new(NLMSG_DEFAULT_SIZE
, GFP_KERNEL
);
1614 if (a
[OVS_VPORT_ATTR_OPTIONS
]) {
1615 err
= ovs_vport_set_options(vport
, a
[OVS_VPORT_ATTR_OPTIONS
]);
1620 if (a
[OVS_VPORT_ATTR_STATS
])
1621 ovs_vport_set_stats(vport
, nla_data(a
[OVS_VPORT_ATTR_STATS
]));
1623 if (a
[OVS_VPORT_ATTR_UPCALL_PID
])
1624 vport
->upcall_portid
= nla_get_u32(a
[OVS_VPORT_ATTR_UPCALL_PID
]);
1626 err
= ovs_vport_cmd_fill_info(vport
, reply
, info
->snd_portid
,
1627 info
->snd_seq
, 0, OVS_VPORT_CMD_NEW
);
1631 ovs_notify(reply
, info
, &ovs_dp_vport_multicast_group
);
1641 static int ovs_vport_cmd_del(struct sk_buff
*skb
, struct genl_info
*info
)
1643 struct nlattr
**a
= info
->attrs
;
1644 struct sk_buff
*reply
;
1645 struct vport
*vport
;
1649 vport
= lookup_vport(sock_net(skb
->sk
), info
->userhdr
, a
);
1650 err
= PTR_ERR(vport
);
1654 if (vport
->port_no
== OVSP_LOCAL
) {
1659 reply
= ovs_vport_cmd_build_info(vport
, info
->snd_portid
,
1660 info
->snd_seq
, OVS_VPORT_CMD_DEL
);
1661 err
= PTR_ERR(reply
);
1666 ovs_dp_detach_port(vport
);
1668 ovs_notify(reply
, info
, &ovs_dp_vport_multicast_group
);
1675 static int ovs_vport_cmd_get(struct sk_buff
*skb
, struct genl_info
*info
)
1677 struct nlattr
**a
= info
->attrs
;
1678 struct ovs_header
*ovs_header
= info
->userhdr
;
1679 struct sk_buff
*reply
;
1680 struct vport
*vport
;
1684 vport
= lookup_vport(sock_net(skb
->sk
), ovs_header
, a
);
1685 err
= PTR_ERR(vport
);
1689 reply
= ovs_vport_cmd_build_info(vport
, info
->snd_portid
,
1690 info
->snd_seq
, OVS_VPORT_CMD_NEW
);
1691 err
= PTR_ERR(reply
);
1697 return genlmsg_reply(reply
, info
);
1704 static int ovs_vport_cmd_dump(struct sk_buff
*skb
, struct netlink_callback
*cb
)
1706 struct ovs_header
*ovs_header
= genlmsg_data(nlmsg_data(cb
->nlh
));
1707 struct datapath
*dp
;
1708 int bucket
= cb
->args
[0], skip
= cb
->args
[1];
1711 dp
= get_dp(sock_net(skb
->sk
), ovs_header
->dp_ifindex
);
1716 for (i
= bucket
; i
< DP_VPORT_HASH_BUCKETS
; i
++) {
1717 struct vport
*vport
;
1720 hlist_for_each_entry_rcu(vport
, &dp
->ports
[i
], dp_hash_node
) {
1722 ovs_vport_cmd_fill_info(vport
, skb
,
1723 NETLINK_CB(cb
->skb
).portid
,
1726 OVS_VPORT_CMD_NEW
) < 0)
1742 static struct genl_ops dp_vport_genl_ops
[] = {
1743 { .cmd
= OVS_VPORT_CMD_NEW
,
1744 .flags
= GENL_ADMIN_PERM
, /* Requires CAP_NET_ADMIN privilege. */
1745 .policy
= vport_policy
,
1746 .doit
= ovs_vport_cmd_new
1748 { .cmd
= OVS_VPORT_CMD_DEL
,
1749 .flags
= GENL_ADMIN_PERM
, /* Requires CAP_NET_ADMIN privilege. */
1750 .policy
= vport_policy
,
1751 .doit
= ovs_vport_cmd_del
1753 { .cmd
= OVS_VPORT_CMD_GET
,
1754 .flags
= 0, /* OK for unprivileged users. */
1755 .policy
= vport_policy
,
1756 .doit
= ovs_vport_cmd_get
,
1757 .dumpit
= ovs_vport_cmd_dump
1759 { .cmd
= OVS_VPORT_CMD_SET
,
1760 .flags
= GENL_ADMIN_PERM
, /* Requires CAP_NET_ADMIN privilege. */
1761 .policy
= vport_policy
,
1762 .doit
= ovs_vport_cmd_set
,
1766 struct genl_family_and_ops
{
1767 struct genl_family
*family
;
1768 struct genl_ops
*ops
;
1770 struct genl_multicast_group
*group
;
1773 static const struct genl_family_and_ops dp_genl_families
[] = {
1774 { &dp_datapath_genl_family
,
1775 dp_datapath_genl_ops
, ARRAY_SIZE(dp_datapath_genl_ops
),
1776 &ovs_dp_datapath_multicast_group
},
1777 { &dp_vport_genl_family
,
1778 dp_vport_genl_ops
, ARRAY_SIZE(dp_vport_genl_ops
),
1779 &ovs_dp_vport_multicast_group
},
1780 { &dp_flow_genl_family
,
1781 dp_flow_genl_ops
, ARRAY_SIZE(dp_flow_genl_ops
),
1782 &ovs_dp_flow_multicast_group
},
1783 { &dp_packet_genl_family
,
1784 dp_packet_genl_ops
, ARRAY_SIZE(dp_packet_genl_ops
),
1788 static void dp_unregister_genl(int n_families
)
1792 for (i
= 0; i
< n_families
; i
++)
1793 genl_unregister_family(dp_genl_families
[i
].family
);
1796 static int dp_register_genl(void)
1803 for (i
= 0; i
< ARRAY_SIZE(dp_genl_families
); i
++) {
1804 const struct genl_family_and_ops
*f
= &dp_genl_families
[i
];
1806 err
= genl_register_family_with_ops(f
->family
, f
->ops
,
1813 err
= genl_register_mc_group(f
->family
, f
->group
);
1822 dp_unregister_genl(n_registered
);
1826 static int __net_init
ovs_init_net(struct net
*net
)
1828 struct ovs_net
*ovs_net
= net_generic(net
, ovs_net_id
);
1830 INIT_LIST_HEAD(&ovs_net
->dps
);
1831 INIT_WORK(&ovs_net
->dp_notify_work
, ovs_dp_notify_wq
);
1835 static void __net_exit
ovs_exit_net(struct net
*net
)
1837 struct datapath
*dp
, *dp_next
;
1838 struct ovs_net
*ovs_net
= net_generic(net
, ovs_net_id
);
1841 list_for_each_entry_safe(dp
, dp_next
, &ovs_net
->dps
, list_node
)
1845 cancel_work_sync(&ovs_net
->dp_notify_work
);
1848 static struct pernet_operations ovs_net_ops
= {
1849 .init
= ovs_init_net
,
1850 .exit
= ovs_exit_net
,
1852 .size
= sizeof(struct ovs_net
),
1855 DEFINE_COMPAT_PNET_REG_FUNC(device
);
1857 static int __init
dp_init(void)
1861 BUILD_BUG_ON(sizeof(struct ovs_skb_cb
) > FIELD_SIZEOF(struct sk_buff
, cb
));
1863 pr_info("Open vSwitch switching datapath %s, built "__DATE__
" "__TIME__
"\n",
1866 err
= ovs_flow_init();
1870 err
= ovs_vport_init();
1872 goto error_flow_exit
;
1874 err
= register_pernet_device(&ovs_net_ops
);
1876 goto error_vport_exit
;
1878 err
= register_netdevice_notifier(&ovs_dp_device_notifier
);
1880 goto error_netns_exit
;
1882 err
= dp_register_genl();
1884 goto error_unreg_notifier
;
1888 error_unreg_notifier
:
1889 unregister_netdevice_notifier(&ovs_dp_device_notifier
);
1891 unregister_pernet_device(&ovs_net_ops
);
1900 static void dp_cleanup(void)
1902 dp_unregister_genl(ARRAY_SIZE(dp_genl_families
));
1903 unregister_netdevice_notifier(&ovs_dp_device_notifier
);
1904 unregister_pernet_device(&ovs_net_ops
);
1910 module_init(dp_init
);
1911 module_exit(dp_cleanup
);
1913 MODULE_DESCRIPTION("Open vSwitch switching datapath");
1914 MODULE_LICENSE("GPL");
1915 MODULE_VERSION(VERSION
);