projects / deliverable / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Merge branch 'for-rafael' of https://git.kernel.org/pub/scm/linux/kernel/git/mzx...
[deliverable/linux.git] / net / openvswitch / datapath.c
CommitLineData
ccb1352e 1/*
ad552007 2 * Copyright (c) 2007-2014 Nicira, Inc.
ccb1352e
JG		 3 *
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.
7 *
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.
12 *
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
16 * 02110-1301, USA
17 */
18
19#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
20
21#include <linux/init.h>
22#include <linux/module.h>
23#include <linux/if_arp.h>
24#include <linux/if_vlan.h>
25#include <linux/in.h>
26#include <linux/ip.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>
ccb1352e
JG		 39#include <linux/ethtool.h>
40#include <linux/wait.h>
ccb1352e
JG		 41#include <asm/div64.h>
42#include <linux/highmem.h>
43#include <linux/netfilter_bridge.h>
44#include <linux/netfilter_ipv4.h>
45#include <linux/inetdevice.h>
46#include <linux/list.h>
47#include <linux/openvswitch.h>
48#include <linux/rculist.h>
49#include <linux/dmi.h>
ccb1352e 50#include <net/genetlink.h>
46df7b81
PS		 51#include <net/net_namespace.h>
52#include <net/netns/generic.h>
ccb1352e
JG		 53
54#include "datapath.h"
55#include "flow.h"
e80857cc 56#include "flow_table.h"
e6445719 57#include "flow_netlink.h"
ccb1352e 58#include "vport-internal_dev.h"
cff63a52 59#include "vport-netdev.h"
ccb1352e 60
8e4e1713 61int ovs_net_id __read_mostly;
9ba559d9 62EXPORT_SYMBOL_GPL(ovs_net_id);
8e4e1713 63
0c200ef9
PS		 64static struct genl_family dp_packet_genl_family;
65static struct genl_family dp_flow_genl_family;
66static struct genl_family dp_datapath_genl_family;
67
74ed7ab9
JS		 68static const struct nla_policy flow_policy[];
69
48e48a70 70static const struct genl_multicast_group ovs_dp_flow_multicast_group = {
71 .name = OVS_FLOW_MCGROUP,
0c200ef9
PS		 72};
73
48e48a70 74static const struct genl_multicast_group ovs_dp_datapath_multicast_group = {
75 .name = OVS_DATAPATH_MCGROUP,
0c200ef9
PS		 76};
77
48e48a70 78static const struct genl_multicast_group ovs_dp_vport_multicast_group = {
79 .name = OVS_VPORT_MCGROUP,
0c200ef9
PS		 80};
81
fb5d1e9e
JR		 82/* Check if need to build a reply message.
83 * OVS userspace sets the NLM_F_ECHO flag if it needs the reply. */
9b67aa4a
SG		 84static bool ovs_must_notify(struct genl_family *family, struct genl_info *info,
85 unsigned int group)
fb5d1e9e
JR		 86{
87 return info->nlhdr->nlmsg_flags & NLM_F_ECHO ||
f8403a2e 88 genl_has_listeners(family, genl_info_net(info), group);
fb5d1e9e
JR		 89}
90
68eb5503 91static void ovs_notify(struct genl_family *family,
2a94fe48 92 struct sk_buff *skb, struct genl_info *info)
ed661185 93{
68eb5503 94 genl_notify(family, skb, genl_info_net(info), info->snd_portid,
2a94fe48 95 0, info->nlhdr, GFP_KERNEL);
ed661185
TG		 96}
97
ccb1352e
JG		 98/**
99 * DOC: Locking:
100 *
8e4e1713
PS		 101 * All writes e.g. Writes to device state (add/remove datapath, port, set
102 * operations on vports, etc.), Writes to other state (flow table
103 * modifications, set miscellaneous datapath parameters, etc.) are protected
104 * by ovs_lock.
ccb1352e
JG		 105 *
106 * Reads are protected by RCU.
107 *
108 * There are a few special cases (mostly stats) that have their own
109 * synchronization but they nest under all of above and don't interact with
110 * each other.
8e4e1713
PS		 111 *
112 * The RTNL lock nests inside ovs_mutex.
ccb1352e
JG		 113 */
114
8e4e1713
PS		 115static DEFINE_MUTEX(ovs_mutex);
116
117void ovs_lock(void)
118{
119 mutex_lock(&ovs_mutex);
120}
121
122void ovs_unlock(void)
123{
124 mutex_unlock(&ovs_mutex);
125}
126
127#ifdef CONFIG_LOCKDEP
128int lockdep_ovsl_is_held(void)
129{
130 if (debug_locks)
131 return lockdep_is_held(&ovs_mutex);
132 else
133 return 1;
134}
9ba559d9 135EXPORT_SYMBOL_GPL(lockdep_ovsl_is_held);
8e4e1713
PS		 136#endif
137
ccb1352e 138static struct vport *new_vport(const struct vport_parms *);
8055a89c 139static int queue_gso_packets(struct datapath *dp, struct sk_buff *,
e8eedb85 140 const struct sw_flow_key *,
ccb1352e 141 const struct dp_upcall_info *);
8055a89c 142static int queue_userspace_packet(struct datapath *dp, struct sk_buff *,
e8eedb85 143 const struct sw_flow_key *,
ccb1352e
JG		 144 const struct dp_upcall_info *);
145
cc3a5ae6
AZ		 146/* Must be called with rcu_read_lock. */
147static struct datapath *get_dp_rcu(struct net *net, int dp_ifindex)
ccb1352e 148{
cc3a5ae6 149 struct net_device *dev = dev_get_by_index_rcu(net, dp_ifindex);
ccb1352e 150
ccb1352e
JG		 151 if (dev) {
152 struct vport *vport = ovs_internal_dev_get_vport(dev);
153 if (vport)
cc3a5ae6 154 return vport->dp;
ccb1352e 155 }
cc3a5ae6
AZ		 156
157 return NULL;
158}
159
160/* The caller must hold either ovs_mutex or rcu_read_lock to keep the
161 * returned dp pointer valid.
162 */
163static inline struct datapath *get_dp(struct net *net, int dp_ifindex)
164{
165 struct datapath *dp;
166
167 WARN_ON_ONCE(!rcu_read_lock_held() && !lockdep_ovsl_is_held());
168 rcu_read_lock();
169 dp = get_dp_rcu(net, dp_ifindex);
ccb1352e
JG		 170 rcu_read_unlock();
171
172 return dp;
173}
174
8e4e1713 175/* Must be called with rcu_read_lock or ovs_mutex. */
971427f3 176const char *ovs_dp_name(const struct datapath *dp)
ccb1352e 177{
8e4e1713 178 struct vport *vport = ovs_vport_ovsl_rcu(dp, OVSP_LOCAL);
c9db965c 179 return ovs_vport_name(vport);
ccb1352e
JG		 180}
181
12eb18f7 182static int get_dpifindex(const struct datapath *dp)
ccb1352e
JG		 183{
184 struct vport *local;
185 int ifindex;
186
187 rcu_read_lock();
188
15eac2a7 189 local = ovs_vport_rcu(dp, OVSP_LOCAL);
ccb1352e 190 if (local)
be4ace6e 191 ifindex = local->dev->ifindex;
ccb1352e
JG		 192 else
193 ifindex = 0;
194
195 rcu_read_unlock();
196
197 return ifindex;
198}
199
200static void destroy_dp_rcu(struct rcu_head *rcu)
201{
202 struct datapath *dp = container_of(rcu, struct datapath, rcu);
203
9b996e54 204 ovs_flow_tbl_destroy(&dp->table);
ccb1352e 205 free_percpu(dp->stats_percpu);
15eac2a7 206 kfree(dp->ports);
ccb1352e
JG		 207 kfree(dp);
208}
209
15eac2a7
PS		 210static struct hlist_head *vport_hash_bucket(const struct datapath *dp,
211 u16 port_no)
212{
213 return &dp->ports[port_no & (DP_VPORT_HASH_BUCKETS - 1)];
214}
215
bb6f9a70 216/* Called with ovs_mutex or RCU read lock. */
15eac2a7
PS		 217struct vport *ovs_lookup_vport(const struct datapath *dp, u16 port_no)
218{
219 struct vport *vport;
15eac2a7
PS		 220 struct hlist_head *head;
221
222 head = vport_hash_bucket(dp, port_no);
b67bfe0d 223 hlist_for_each_entry_rcu(vport, head, dp_hash_node) {
15eac2a7
PS		 224 if (vport->port_no == port_no)
225 return vport;
226 }
227 return NULL;
228}
229
8e4e1713 230/* Called with ovs_mutex. */
ccb1352e
JG		 231static struct vport *new_vport(const struct vport_parms *parms)
232{
233 struct vport *vport;
234
235 vport = ovs_vport_add(parms);
236 if (!IS_ERR(vport)) {
237 struct datapath *dp = parms->dp;
15eac2a7 238 struct hlist_head *head = vport_hash_bucket(dp, vport->port_no);
ccb1352e 239
15eac2a7 240 hlist_add_head_rcu(&vport->dp_hash_node, head);
ccb1352e 241 }
ccb1352e
JG		 242 return vport;
243}
244
ccb1352e
JG		 245void ovs_dp_detach_port(struct vport *p)
246{
8e4e1713 247 ASSERT_OVSL();
ccb1352e
JG		 248
249 /* First drop references to device. */
15eac2a7 250 hlist_del_rcu(&p->dp_hash_node);
ccb1352e
JG		 251
252 /* Then destroy it. */
253 ovs_vport_del(p);
254}
255
256/* Must be called with rcu_read_lock. */
8c8b1b83 257void ovs_dp_process_packet(struct sk_buff *skb, struct sw_flow_key *key)
ccb1352e 258{
83c8df26 259 const struct vport *p = OVS_CB(skb)->input_vport;
ccb1352e
JG		 260 struct datapath *dp = p->dp;
261 struct sw_flow *flow;
d98612b8 262 struct sw_flow_actions *sf_acts;
ccb1352e 263 struct dp_stats_percpu *stats;
ccb1352e 264 u64 *stats_counter;
1bd7116f 265 u32 n_mask_hit;
ccb1352e 266
404f2f10 267 stats = this_cpu_ptr(dp->stats_percpu);
ccb1352e 268
ccb1352e 269 /* Look up flow. */
8c8b1b83 270 flow = ovs_flow_tbl_lookup_stats(&dp->table, key, &n_mask_hit);
ccb1352e
JG		 271 if (unlikely(!flow)) {
272 struct dp_upcall_info upcall;
8c8b1b83 273 int error;
ccb1352e 274
ccea7445 275 memset(&upcall, 0, sizeof(upcall));
ccb1352e 276 upcall.cmd = OVS_PACKET_CMD_MISS;
5cd667b0 277 upcall.portid = ovs_vport_find_upcall_portid(p, skb);
7f8a436e 278 upcall.mru = OVS_CB(skb)->mru;
e8eedb85 279 error = ovs_dp_upcall(dp, skb, key, &upcall);
c5eba0b6
LR		 280 if (unlikely(error))
281 kfree_skb(skb);
282 else
283 consume_skb(skb);
ccb1352e
JG		 284 stats_counter = &stats->n_missed;
285 goto out;
286 }
287
d98612b8
LJ		 288 ovs_flow_stats_update(flow, key->tp.flags, skb);
289 sf_acts = rcu_dereference(flow->sf_acts);
290 ovs_execute_actions(dp, skb, sf_acts, key);
ccb1352e 291
e298e505 292 stats_counter = &stats->n_hit;
ccb1352e
JG		 293
294out:
295 /* Update datapath statistics. */
df9d9fdf 296 u64_stats_update_begin(&stats->syncp);
ccb1352e 297 (*stats_counter)++;
1bd7116f 298 stats->n_mask_hit += n_mask_hit;
df9d9fdf 299 u64_stats_update_end(&stats->syncp);
ccb1352e
JG		 300}
301
ccb1352e 302int ovs_dp_upcall(struct datapath *dp, struct sk_buff *skb,
e8eedb85 303 const struct sw_flow_key *key,
46df7b81 304 const struct dp_upcall_info *upcall_info)
ccb1352e
JG		 305{
306 struct dp_stats_percpu *stats;
ccb1352e
JG		 307 int err;
308
15e47304 309 if (upcall_info->portid == 0) {
ccb1352e
JG		 310 err = -ENOTCONN;
311 goto err;
312 }
313
ccb1352e 314 if (!skb_is_gso(skb))
e8eedb85 315 err = queue_userspace_packet(dp, skb, key, upcall_info);
ccb1352e 316 else
e8eedb85 317 err = queue_gso_packets(dp, skb, key, upcall_info);
ccb1352e
JG		 318 if (err)
319 goto err;
320
321 return 0;
322
323err:
404f2f10 324 stats = this_cpu_ptr(dp->stats_percpu);
ccb1352e 325
df9d9fdf 326 u64_stats_update_begin(&stats->syncp);
ccb1352e 327 stats->n_lost++;
df9d9fdf 328 u64_stats_update_end(&stats->syncp);
ccb1352e
JG		 329
330 return err;
331}
332
8055a89c 333static int queue_gso_packets(struct datapath *dp, struct sk_buff *skb,
e8eedb85 334 const struct sw_flow_key *key,
ccb1352e
JG		 335 const struct dp_upcall_info *upcall_info)
336{
a1b5d0dd 337 unsigned short gso_type = skb_shinfo(skb)->gso_type;
ccb1352e
JG		 338 struct sw_flow_key later_key;
339 struct sk_buff *segs, *nskb;
e8eedb85 340 struct ovs_skb_cb ovs_cb;
ccb1352e
JG		 341 int err;
342
e8eedb85 343 ovs_cb = *OVS_CB(skb);
09c5e605 344 segs = __skb_gso_segment(skb, NETIF_F_SG, false);
e8eedb85 345 *OVS_CB(skb) = ovs_cb;
92e5dfc3
PS		 346 if (IS_ERR(segs))
347 return PTR_ERR(segs);
330966e5
FW		 348 if (segs == NULL)
349 return -EINVAL;
ccb1352e 350
e8eedb85
PS		 351 if (gso_type & SKB_GSO_UDP) {
352 /* The initial flow key extracted by ovs_flow_key_extract()
353 * in this case is for a first fragment, so we need to
354 * properly mark later fragments.
355 */
356 later_key = *key;
357 later_key.ip.frag = OVS_FRAG_TYPE_LATER;
358 }
359
ccb1352e
JG		 360 /* Queue all of the segments. */
361 skb = segs;
362 do {
e8eedb85
PS		 363 *OVS_CB(skb) = ovs_cb;
364 if (gso_type & SKB_GSO_UDP && skb != segs)
365 key = &later_key;
366
367 err = queue_userspace_packet(dp, skb, key, upcall_info);
ccb1352e
JG		 368 if (err)
369 break;
370
ccb1352e
JG		 371 } while ((skb = skb->next));
372
373 /* Free all of the segments. */
374 skb = segs;
375 do {
376 nskb = skb->next;
377 if (err)
378 kfree_skb(skb);
379 else
380 consume_skb(skb);
381 } while ((skb = nskb));
382 return err;
383}
384
8f0aad6f 385static size_t upcall_msg_size(const struct dp_upcall_info *upcall_info,
bda56f14 386 unsigned int hdrlen)
c3ff8cfe
TG		 387{
388 size_t size = NLMSG_ALIGN(sizeof(struct ovs_header))
bda56f14 389 + nla_total_size(hdrlen) /* OVS_PACKET_ATTR_PACKET */
41af73e9 390 + nla_total_size(ovs_key_attr_size()); /* OVS_PACKET_ATTR_KEY */
c3ff8cfe
TG		 391
392 /* OVS_PACKET_ATTR_USERDATA */
8f0aad6f
WZ		 393 if (upcall_info->userdata)
394 size += NLA_ALIGN(upcall_info->userdata->nla_len);
395
396 /* OVS_PACKET_ATTR_EGRESS_TUN_KEY */
397 if (upcall_info->egress_tun_info)
398 size += nla_total_size(ovs_tun_key_attr_size());
c3ff8cfe 399
ccea7445
NM		 400 /* OVS_PACKET_ATTR_ACTIONS */
401 if (upcall_info->actions_len)
402 size += nla_total_size(upcall_info->actions_len);
403
7f8a436e
JS		 404 /* OVS_PACKET_ATTR_MRU */
405 if (upcall_info->mru)
406 size += nla_total_size(sizeof(upcall_info->mru));
407
c3ff8cfe
TG		 408 return size;
409}
410
7f8a436e
JS		 411static void pad_packet(struct datapath *dp, struct sk_buff *skb)
412{
413 if (!(dp->user_features & OVS_DP_F_UNALIGNED)) {
414 size_t plen = NLA_ALIGN(skb->len) - skb->len;
415
416 if (plen > 0)
417 memset(skb_put(skb, plen), 0, plen);
418 }
419}
420
8055a89c 421static int queue_userspace_packet(struct datapath *dp, struct sk_buff *skb,
e8eedb85 422 const struct sw_flow_key *key,
ccb1352e
JG		 423 const struct dp_upcall_info *upcall_info)
424{
425 struct ovs_header *upcall;
426 struct sk_buff *nskb = NULL;
4ee45ea0 427 struct sk_buff *user_skb = NULL; /* to be queued to userspace */
ccb1352e 428 struct nlattr *nla;
795449d8 429 struct genl_info info = {
8055a89c 430 .dst_sk = ovs_dp_get_net(dp)->genl_sock,
795449d8
TG		 431 .snd_portid = upcall_info->portid,
432 };
433 size_t len;
bda56f14 434 unsigned int hlen;
8055a89c
TG		 435 int err, dp_ifindex;
436
437 dp_ifindex = get_dpifindex(dp);
438 if (!dp_ifindex)
439 return -ENODEV;
ccb1352e 440
df8a39de 441 if (skb_vlan_tag_present(skb)) {
ccb1352e
JG		 442 nskb = skb_clone(skb, GFP_ATOMIC);
443 if (!nskb)
444 return -ENOMEM;
445
5968250c 446 nskb = __vlan_hwaccel_push_inside(nskb);
8aa51d64 447 if (!nskb)
ccb1352e
JG		 448 return -ENOMEM;
449
ccb1352e
JG		 450 skb = nskb;
451 }
452
453 if (nla_attr_size(skb->len) > USHRT_MAX) {
454 err = -EFBIG;
455 goto out;
456 }
457
bda56f14
TG		 458 /* Complete checksum if needed */
459 if (skb->ip_summed == CHECKSUM_PARTIAL &&
460 (err = skb_checksum_help(skb)))
461 goto out;
462
463 /* Older versions of OVS user space enforce alignment of the last
464 * Netlink attribute to NLA_ALIGNTO which would require extensive
465 * padding logic. Only perform zerocopy if padding is not required.
466 */
467 if (dp->user_features & OVS_DP_F_UNALIGNED)
468 hlen = skb_zerocopy_headlen(skb);
469 else
470 hlen = skb->len;
471
8f0aad6f 472 len = upcall_msg_size(upcall_info, hlen);
795449d8 473 user_skb = genlmsg_new_unicast(len, &info, GFP_ATOMIC);
ccb1352e
JG		 474 if (!user_skb) {
475 err = -ENOMEM;
476 goto out;
477 }
478
479 upcall = genlmsg_put(user_skb, 0, 0, &dp_packet_genl_family,
480 0, upcall_info->cmd);
481 upcall->dp_ifindex = dp_ifindex;
482
5b4237bb 483 err = ovs_nla_put_key(key, key, OVS_PACKET_ATTR_KEY, false, user_skb);
f53e3831 484 BUG_ON(err);
ccb1352e
JG		 485
486 if (upcall_info->userdata)
4490108b
BP		 487 __nla_put(user_skb, OVS_PACKET_ATTR_USERDATA,
488 nla_len(upcall_info->userdata),
489 nla_data(upcall_info->userdata));
ccb1352e 490
8f0aad6f
WZ		 491 if (upcall_info->egress_tun_info) {
492 nla = nla_nest_start(user_skb, OVS_PACKET_ATTR_EGRESS_TUN_KEY);
493 err = ovs_nla_put_egress_tunnel_key(user_skb,
4c222798
PS		 494 upcall_info->egress_tun_info,
495 upcall_info->egress_tun_opts);
8f0aad6f
WZ		 496 BUG_ON(err);
497 nla_nest_end(user_skb, nla);
498 }
499
ccea7445
NM		 500 if (upcall_info->actions_len) {
501 nla = nla_nest_start(user_skb, OVS_PACKET_ATTR_ACTIONS);
502 err = ovs_nla_put_actions(upcall_info->actions,
503 upcall_info->actions_len,
504 user_skb);
505 if (!err)
506 nla_nest_end(user_skb, nla);
507 else
508 nla_nest_cancel(user_skb, nla);
509 }
510
7f8a436e
JS		 511 /* Add OVS_PACKET_ATTR_MRU */
512 if (upcall_info->mru) {
513 if (nla_put_u16(user_skb, OVS_PACKET_ATTR_MRU,
514 upcall_info->mru)) {
515 err = -ENOBUFS;
516 goto out;
517 }
518 pad_packet(dp, user_skb);
519 }
520
bda56f14
TG		 521 /* Only reserve room for attribute header, packet data is added
522 * in skb_zerocopy() */
523 if (!(nla = nla_reserve(user_skb, OVS_PACKET_ATTR_PACKET, 0))) {
524 err = -ENOBUFS;
525 goto out;
526 }
527 nla->nla_len = nla_attr_size(skb->len);
ccb1352e 528
36d5fe6a
ZK		 529 err = skb_zerocopy(user_skb, skb, skb->len, hlen);
530 if (err)
531 goto out;
ccb1352e 532
aea0bb4f 533 /* Pad OVS_PACKET_ATTR_PACKET if linear copy was performed */
7f8a436e 534 pad_packet(dp, user_skb);
aea0bb4f 535
bda56f14 536 ((struct nlmsghdr *) user_skb->data)->nlmsg_len = user_skb->len;
ccb1352e 537
bda56f14 538 err = genlmsg_unicast(ovs_dp_get_net(dp), user_skb, upcall_info->portid);
4ee45ea0 539 user_skb = NULL;
ccb1352e 540out:
36d5fe6a
ZK		 541 if (err)
542 skb_tx_error(skb);
4ee45ea0 543 kfree_skb(user_skb);
ccb1352e
JG		 544 kfree_skb(nskb);
545 return err;
546}
547
ccb1352e
JG		 548static int ovs_packet_cmd_execute(struct sk_buff *skb, struct genl_info *info)
549{
550 struct ovs_header *ovs_header = info->userhdr;
7f8a436e 551 struct net *net = sock_net(skb->sk);
ccb1352e
JG		 552 struct nlattr **a = info->attrs;
553 struct sw_flow_actions *acts;
554 struct sk_buff *packet;
555 struct sw_flow *flow;
d98612b8 556 struct sw_flow_actions *sf_acts;
ccb1352e
JG		 557 struct datapath *dp;
558 struct ethhdr *eth;
83c8df26 559 struct vport *input_vport;
7f8a436e 560 u16 mru = 0;
ccb1352e
JG		 561 int len;
562 int err;
1ba39804 563 bool log = !a[OVS_PACKET_ATTR_PROBE];
ccb1352e
JG		 564
565 err = -EINVAL;
566 if (!a[OVS_PACKET_ATTR_PACKET] || !a[OVS_PACKET_ATTR_KEY] ||
dded45fc 567 !a[OVS_PACKET_ATTR_ACTIONS])
ccb1352e
JG		 568 goto err;
569
570 len = nla_len(a[OVS_PACKET_ATTR_PACKET]);
571 packet = __dev_alloc_skb(NET_IP_ALIGN + len, GFP_KERNEL);
572 err = -ENOMEM;
573 if (!packet)
574 goto err;
575 skb_reserve(packet, NET_IP_ALIGN);
576
32686a9d 577 nla_memcpy(__skb_put(packet, len), a[OVS_PACKET_ATTR_PACKET], len);
ccb1352e
JG		 578
579 skb_reset_mac_header(packet);
580 eth = eth_hdr(packet);
581
582 /* Normally, setting the skb 'protocol' field would be handled by a
583 * call to eth_type_trans(), but it assumes there's a sending
584 * device, which we may not have. */
6713fc9b 585 if (eth_proto_is_802_3(eth->h_proto))
ccb1352e
JG		 586 packet->protocol = eth->h_proto;
587 else
588 packet->protocol = htons(ETH_P_802_2);
589
7f8a436e
JS		 590 /* Set packet's mru */
591 if (a[OVS_PACKET_ATTR_MRU]) {
592 mru = nla_get_u16(a[OVS_PACKET_ATTR_MRU]);
593 packet->ignore_df = 1;
594 }
595 OVS_CB(packet)->mru = mru;
596
ccb1352e 597 /* Build an sw_flow for sending this packet. */
23dabf88 598 flow = ovs_flow_alloc();
ccb1352e
JG		 599 err = PTR_ERR(flow);
600 if (IS_ERR(flow))
601 goto err_kfree_skb;
602
c2ac6673
JS		 603 err = ovs_flow_key_extract_userspace(net, a[OVS_PACKET_ATTR_KEY],
604 packet, &flow->key, log);
ccb1352e
JG		 605 if (err)
606 goto err_flow_free;
607
7f8a436e 608 err = ovs_nla_copy_actions(net, a[OVS_PACKET_ATTR_ACTIONS],
05da5898 609 &flow->key, &acts, log);
74f84a57
PS		 610 if (err)
611 goto err_flow_free;
ccb1352e 612
f5796684 613 rcu_assign_pointer(flow->sf_acts, acts);
ccb1352e 614 packet->priority = flow->key.phy.priority;
39c7caeb 615 packet->mark = flow->key.phy.skb_mark;
ccb1352e
JG		 616
617 rcu_read_lock();
7f8a436e 618 dp = get_dp_rcu(net, ovs_header->dp_ifindex);
ccb1352e
JG		 619 err = -ENODEV;
620 if (!dp)
621 goto err_unlock;
622
83c8df26
PS		 623 input_vport = ovs_vport_rcu(dp, flow->key.phy.in_port);
624 if (!input_vport)
625 input_vport = ovs_vport_rcu(dp, OVSP_LOCAL);
626
627 if (!input_vport)
628 goto err_unlock;
629
7f8a436e 630 packet->dev = input_vport->dev;
83c8df26 631 OVS_CB(packet)->input_vport = input_vport;
d98612b8 632 sf_acts = rcu_dereference(flow->sf_acts);
83c8df26 633
ccb1352e 634 local_bh_disable();
d98612b8 635 err = ovs_execute_actions(dp, packet, sf_acts, &flow->key);
ccb1352e
JG		 636 local_bh_enable();
637 rcu_read_unlock();
638
03f0d916 639 ovs_flow_free(flow, false);
ccb1352e
JG		 640 return err;
641
642err_unlock:
643 rcu_read_unlock();
644err_flow_free:
03f0d916 645 ovs_flow_free(flow, false);
ccb1352e
JG		 646err_kfree_skb:
647 kfree_skb(packet);
648err:
649 return err;
650}
651
652static const struct nla_policy packet_policy[OVS_PACKET_ATTR_MAX + 1] = {
dded45fc 653 [OVS_PACKET_ATTR_PACKET] = { .len = ETH_HLEN },
ccb1352e
JG		 654 [OVS_PACKET_ATTR_KEY] = { .type = NLA_NESTED },
655 [OVS_PACKET_ATTR_ACTIONS] = { .type = NLA_NESTED },
1ba39804 656 [OVS_PACKET_ATTR_PROBE] = { .type = NLA_FLAG },
7f8a436e 657 [OVS_PACKET_ATTR_MRU] = { .type = NLA_U16 },
ccb1352e
JG		 658};
659
4534de83 660static const struct genl_ops dp_packet_genl_ops[] = {
ccb1352e
JG		 661 { .cmd = OVS_PACKET_CMD_EXECUTE,
662 .flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
663 .policy = packet_policy,
664 .doit = ovs_packet_cmd_execute
665 }
666};
667
0c200ef9
PS		 668static struct genl_family dp_packet_genl_family = {
669 .id = GENL_ID_GENERATE,
670 .hdrsize = sizeof(struct ovs_header),
671 .name = OVS_PACKET_FAMILY,
672 .version = OVS_PACKET_VERSION,
673 .maxattr = OVS_PACKET_ATTR_MAX,
674 .netnsok = true,
675 .parallel_ops = true,
676 .ops = dp_packet_genl_ops,
677 .n_ops = ARRAY_SIZE(dp_packet_genl_ops),
678};
679
12eb18f7 680static void get_dp_stats(const struct datapath *dp, struct ovs_dp_stats *stats,
1bd7116f 681 struct ovs_dp_megaflow_stats *mega_stats)
ccb1352e
JG		 682{
683 int i;
ccb1352e 684
1bd7116f
AZ		 685 memset(mega_stats, 0, sizeof(*mega_stats));
686
b637e498 687 stats->n_flows = ovs_flow_tbl_count(&dp->table);
1bd7116f 688 mega_stats->n_masks = ovs_flow_tbl_num_masks(&dp->table);
ccb1352e
JG		 689
690 stats->n_hit = stats->n_missed = stats->n_lost = 0;
1bd7116f 691
ccb1352e
JG		 692 for_each_possible_cpu(i) {
693 const struct dp_stats_percpu *percpu_stats;
694 struct dp_stats_percpu local_stats;
695 unsigned int start;
696
697 percpu_stats = per_cpu_ptr(dp->stats_percpu, i);
698
699 do {
57a7744e 700 start = u64_stats_fetch_begin_irq(&percpu_stats->syncp);
ccb1352e 701 local_stats = *percpu_stats;
57a7744e 702 } while (u64_stats_fetch_retry_irq(&percpu_stats->syncp, start));
ccb1352e
JG		 703
704 stats->n_hit += local_stats.n_hit;
705 stats->n_missed += local_stats.n_missed;
706 stats->n_lost += local_stats.n_lost;
1bd7116f 707 mega_stats->n_mask_hit += local_stats.n_mask_hit;
ccb1352e
JG		 708 }
709}
710
74ed7ab9
JS		 711static bool should_fill_key(const struct sw_flow_id *sfid, uint32_t ufid_flags)
712{
713 return ovs_identifier_is_ufid(sfid) &&
714 !(ufid_flags & OVS_UFID_F_OMIT_KEY);
715}
716
717static bool should_fill_mask(uint32_t ufid_flags)
718{
719 return !(ufid_flags & OVS_UFID_F_OMIT_MASK);
720}
721
722static bool should_fill_actions(uint32_t ufid_flags)
c3ff8cfe 723{
74ed7ab9
JS		 724 return !(ufid_flags & OVS_UFID_F_OMIT_ACTIONS);
725}
726
727static size_t ovs_flow_cmd_msg_size(const struct sw_flow_actions *acts,
728 const struct sw_flow_id *sfid,
729 uint32_t ufid_flags)
730{
731 size_t len = NLMSG_ALIGN(sizeof(struct ovs_header));
732
733 /* OVS_FLOW_ATTR_UFID */
734 if (sfid && ovs_identifier_is_ufid(sfid))
735 len += nla_total_size(sfid->ufid_len);
736
737 /* OVS_FLOW_ATTR_KEY */
738 if (!sfid || should_fill_key(sfid, ufid_flags))
739 len += nla_total_size(ovs_key_attr_size());
740
741 /* OVS_FLOW_ATTR_MASK */
742 if (should_fill_mask(ufid_flags))
743 len += nla_total_size(ovs_key_attr_size());
744
745 /* OVS_FLOW_ATTR_ACTIONS */
746 if (should_fill_actions(ufid_flags))
8e2fed1c 747 len += nla_total_size(acts->orig_len);
74ed7ab9
JS		 748
749 return len
c3ff8cfe
TG		 750 + nla_total_size(sizeof(struct ovs_flow_stats)) /* OVS_FLOW_ATTR_STATS */
751 + nla_total_size(1) /* OVS_FLOW_ATTR_TCP_FLAGS */
74ed7ab9 752 + nla_total_size(8); /* OVS_FLOW_ATTR_USED */
c3ff8cfe
TG		 753}
754
ca7105f2
JS		 755/* Called with ovs_mutex or RCU read lock. */
756static int ovs_flow_cmd_fill_stats(const struct sw_flow *flow,
757 struct sk_buff *skb)
758{
759 struct ovs_flow_stats stats;
760 __be16 tcp_flags;
761 unsigned long used;
ccb1352e 762
e298e505 763 ovs_flow_stats_get(flow, &stats, &used, &tcp_flags);
0e9796b4 764
028d6a67
DM		 765 if (used &&
766 nla_put_u64(skb, OVS_FLOW_ATTR_USED, ovs_flow_used_time(used)))
ca7105f2 767 return -EMSGSIZE;
ccb1352e 768
028d6a67 769 if (stats.n_packets &&
e298e505 770 nla_put(skb, OVS_FLOW_ATTR_STATS, sizeof(struct ovs_flow_stats), &stats))
ca7105f2 771 return -EMSGSIZE;
ccb1352e 772
e298e505
PS		 773 if ((u8)ntohs(tcp_flags) &&
774 nla_put_u8(skb, OVS_FLOW_ATTR_TCP_FLAGS, (u8)ntohs(tcp_flags)))
ca7105f2
JS		 775 return -EMSGSIZE;
776
777 return 0;
778}
779
780/* Called with ovs_mutex or RCU read lock. */
781static int ovs_flow_cmd_fill_actions(const struct sw_flow *flow,
782 struct sk_buff *skb, int skb_orig_len)
783{
784 struct nlattr *start;
785 int err;
ccb1352e
JG		 786
787 /* If OVS_FLOW_ATTR_ACTIONS doesn't fit, skip dumping the actions if
788 * this is the first flow to be dumped into 'skb'. This is unusual for
789 * Netlink but individual action lists can be longer than
790 * NLMSG_GOODSIZE and thus entirely undumpable if we didn't do this.
791 * The userspace caller can always fetch the actions separately if it
792 * really wants them. (Most userspace callers in fact don't care.)
793 *
794 * This can only fail for dump operations because the skb is always
795 * properly sized for single flows.
796 */
74f84a57
PS		 797 start = nla_nest_start(skb, OVS_FLOW_ATTR_ACTIONS);
798 if (start) {
d57170b1
PS		 799 const struct sw_flow_actions *sf_acts;
800
663efa36 801 sf_acts = rcu_dereference_ovsl(flow->sf_acts);
e6445719
PS		 802 err = ovs_nla_put_actions(sf_acts->actions,
803 sf_acts->actions_len, skb);
0e9796b4 804
74f84a57
PS		 805 if (!err)
806 nla_nest_end(skb, start);
807 else {
808 if (skb_orig_len)
ca7105f2 809 return err;
74f84a57
PS		 810
811 nla_nest_cancel(skb, start);
812 }
ca7105f2
JS		 813 } else if (skb_orig_len) {
814 return -EMSGSIZE;
815 }
816
817 return 0;
818}
819
820/* Called with ovs_mutex or RCU read lock. */
821static int ovs_flow_cmd_fill_info(const struct sw_flow *flow, int dp_ifindex,
822 struct sk_buff *skb, u32 portid,
74ed7ab9 823 u32 seq, u32 flags, u8 cmd, u32 ufid_flags)
ca7105f2
JS		 824{
825 const int skb_orig_len = skb->len;
826 struct ovs_header *ovs_header;
827 int err;
828
829 ovs_header = genlmsg_put(skb, portid, seq, &dp_flow_genl_family,
830 flags, cmd);
831 if (!ovs_header)
832 return -EMSGSIZE;
833
834 ovs_header->dp_ifindex = dp_ifindex;
835
74ed7ab9 836 err = ovs_nla_put_identifier(flow, skb);
5b4237bb
JS		 837 if (err)
838 goto error;
839
74ed7ab9
JS		 840 if (should_fill_key(&flow->id, ufid_flags)) {
841 err = ovs_nla_put_masked_key(flow, skb);
842 if (err)
843 goto error;
844 }
845
846 if (should_fill_mask(ufid_flags)) {
847 err = ovs_nla_put_mask(flow, skb);
848 if (err)
849 goto error;
850 }
ca7105f2
JS		 851
852 err = ovs_flow_cmd_fill_stats(flow, skb);
853 if (err)
854 goto error;
855
74ed7ab9
JS		 856 if (should_fill_actions(ufid_flags)) {
857 err = ovs_flow_cmd_fill_actions(flow, skb, skb_orig_len);
858 if (err)
859 goto error;
860 }
ccb1352e 861
053c095a
JB		 862 genlmsg_end(skb, ovs_header);
863 return 0;
ccb1352e 864
ccb1352e
JG		 865error:
866 genlmsg_cancel(skb, ovs_header);
867 return err;
868}
869
0e9796b4
JR		 870/* May not be called with RCU read lock. */
871static struct sk_buff *ovs_flow_cmd_alloc_info(const struct sw_flow_actions *acts,
74ed7ab9 872 const struct sw_flow_id *sfid,
fb5d1e9e 873 struct genl_info *info,
74ed7ab9
JS		 874 bool always,
875 uint32_t ufid_flags)
ccb1352e 876{
fb5d1e9e 877 struct sk_buff *skb;
74ed7ab9 878 size_t len;
ccb1352e 879
9b67aa4a 880 if (!always && !ovs_must_notify(&dp_flow_genl_family, info, 0))
fb5d1e9e
JR		 881 return NULL;
882
74ed7ab9
JS		 883 len = ovs_flow_cmd_msg_size(acts, sfid, ufid_flags);
884 skb = genlmsg_new_unicast(len, info, GFP_KERNEL);
fb5d1e9e
JR		 885 if (!skb)
886 return ERR_PTR(-ENOMEM);
887
888 return skb;
ccb1352e
JG		 889}
890
0e9796b4
JR		 891/* Called with ovs_mutex. */
892static struct sk_buff *ovs_flow_cmd_build_info(const struct sw_flow *flow,
893 int dp_ifindex,
894 struct genl_info *info, u8 cmd,
74ed7ab9 895 bool always, u32 ufid_flags)
ccb1352e
JG		 896{
897 struct sk_buff *skb;
898 int retval;
899
74ed7ab9
JS		 900 skb = ovs_flow_cmd_alloc_info(ovsl_dereference(flow->sf_acts),
901 &flow->id, info, always, ufid_flags);
d0e992aa 902 if (IS_ERR_OR_NULL(skb))
fb5d1e9e 903 return skb;
ccb1352e 904
0e9796b4
JR		 905 retval = ovs_flow_cmd_fill_info(flow, dp_ifindex, skb,
906 info->snd_portid, info->snd_seq, 0,
74ed7ab9 907 cmd, ufid_flags);
ccb1352e
JG		 908 BUG_ON(retval < 0);
909 return skb;
910}
911
37bdc87b 912static int ovs_flow_cmd_new(struct sk_buff *skb, struct genl_info *info)
ccb1352e 913{
7f8a436e 914 struct net *net = sock_net(skb->sk);
ccb1352e
JG		 915 struct nlattr **a = info->attrs;
916 struct ovs_header *ovs_header = info->userhdr;
74ed7ab9 917 struct sw_flow *flow = NULL, *new_flow;
03f0d916 918 struct sw_flow_mask mask;
ccb1352e
JG		 919 struct sk_buff *reply;
920 struct datapath *dp;
74ed7ab9 921 struct sw_flow_key key;
37bdc87b 922 struct sw_flow_actions *acts;
03f0d916 923 struct sw_flow_match match;
74ed7ab9 924 u32 ufid_flags = ovs_nla_get_ufid_flags(a[OVS_FLOW_ATTR_UFID_FLAGS]);
ccb1352e 925 int error;
05da5898 926 bool log = !a[OVS_FLOW_ATTR_PROBE];
ccb1352e 927
893f139b 928 /* Must have key and actions. */
ccb1352e 929 error = -EINVAL;
426cda5c 930 if (!a[OVS_FLOW_ATTR_KEY]) {
05da5898 931 OVS_NLERR(log, "Flow key attr not present in new flow.");
ccb1352e 932 goto error;
426cda5c
JG		 933 }
934 if (!a[OVS_FLOW_ATTR_ACTIONS]) {
05da5898 935 OVS_NLERR(log, "Flow actions attr not present in new flow.");
893f139b 936 goto error;
426cda5c 937 }
03f0d916 938
893f139b
JR		 939 /* Most of the time we need to allocate a new flow, do it before
940 * locking.
941 */
942 new_flow = ovs_flow_alloc();
943 if (IS_ERR(new_flow)) {
944 error = PTR_ERR(new_flow);
945 goto error;
946 }
947
948 /* Extract key. */
74ed7ab9 949 ovs_match_init(&match, &key, &mask);
c2ac6673 950 error = ovs_nla_get_match(net, &match, a[OVS_FLOW_ATTR_KEY],
05da5898 951 a[OVS_FLOW_ATTR_MASK], log);
ccb1352e 952 if (error)
893f139b 953 goto err_kfree_flow;
ccb1352e 954
74ed7ab9
JS		 955 ovs_flow_mask_key(&new_flow->key, &key, &mask);
956
957 /* Extract flow identifier. */
958 error = ovs_nla_get_identifier(&new_flow->id, a[OVS_FLOW_ATTR_UFID],
959 &key, log);
960 if (error)
961 goto err_kfree_flow;
74f84a57 962
893f139b 963 /* Validate actions. */
7f8a436e
JS		 964 error = ovs_nla_copy_actions(net, a[OVS_FLOW_ATTR_ACTIONS],
965 &new_flow->key, &acts, log);
37bdc87b 966 if (error) {
05da5898 967 OVS_NLERR(log, "Flow actions may not be safe on all matching packets.");
2fdb957d 968 goto err_kfree_flow;
893f139b
JR		 969 }
970
74ed7ab9
JS		 971 reply = ovs_flow_cmd_alloc_info(acts, &new_flow->id, info, false,
972 ufid_flags);
893f139b
JR		 973 if (IS_ERR(reply)) {
974 error = PTR_ERR(reply);
975 goto err_kfree_acts;
ccb1352e
JG		 976 }
977
8e4e1713 978 ovs_lock();
7f8a436e 979 dp = get_dp(net, ovs_header->dp_ifindex);
893f139b
JR		 980 if (unlikely(!dp)) {
981 error = -ENODEV;
8e4e1713 982 goto err_unlock_ovs;
893f139b 983 }
74ed7ab9 984
03f0d916 985 /* Check if this is a duplicate flow */
74ed7ab9
JS		 986 if (ovs_identifier_is_ufid(&new_flow->id))
987 flow = ovs_flow_tbl_lookup_ufid(&dp->table, &new_flow->id);
988 if (!flow)
989 flow = ovs_flow_tbl_lookup(&dp->table, &key);
893f139b
JR		 990 if (likely(!flow)) {
991 rcu_assign_pointer(new_flow->sf_acts, acts);
ccb1352e
JG		 992
993 /* Put flow in bucket. */
893f139b
JR		 994 error = ovs_flow_tbl_insert(&dp->table, new_flow, &mask);
995 if (unlikely(error)) {
618ed0c8 996 acts = NULL;
893f139b
JR		 997 goto err_unlock_ovs;
998 }
999
1000 if (unlikely(reply)) {
1001 error = ovs_flow_cmd_fill_info(new_flow,
1002 ovs_header->dp_ifindex,
1003 reply, info->snd_portid,
1004 info->snd_seq, 0,
74ed7ab9
JS		 1005 OVS_FLOW_CMD_NEW,
1006 ufid_flags);
893f139b 1007 BUG_ON(error < 0);
618ed0c8 1008 }
893f139b 1009 ovs_unlock();
ccb1352e 1010 } else {
37bdc87b
JR		 1011 struct sw_flow_actions *old_acts;
1012
ccb1352e
JG		 1013 /* Bail out if we're not allowed to modify an existing flow.
1014 * We accept NLM_F_CREATE in place of the intended NLM_F_EXCL
1015 * because Generic Netlink treats the latter as a dump
1016 * request. We also accept NLM_F_EXCL in case that bug ever
1017 * gets fixed.
1018 */
893f139b
JR		 1019 if (unlikely(info->nlhdr->nlmsg_flags & (NLM_F_CREATE
1020 | NLM_F_EXCL))) {
1021 error = -EEXIST;
8e4e1713 1022 goto err_unlock_ovs;
893f139b 1023 }
74ed7ab9
JS		 1024 /* The flow identifier has to be the same for flow updates.
1025 * Look for any overlapping flow.
1026 */
1027 if (unlikely(!ovs_flow_cmp(flow, &match))) {
1028 if (ovs_identifier_is_key(&flow->id))
1029 flow = ovs_flow_tbl_lookup_exact(&dp->table,
1030 &match);
1031 else /* UFID matches but key is different */
1032 flow = NULL;
4a46b24e
AW		 1033 if (!flow) {
1034 error = -ENOENT;
1035 goto err_unlock_ovs;
1036 }
893f139b 1037 }
37bdc87b
JR		 1038 /* Update actions. */
1039 old_acts = ovsl_dereference(flow->sf_acts);
1040 rcu_assign_pointer(flow->sf_acts, acts);
37bdc87b 1041
893f139b
JR		 1042 if (unlikely(reply)) {
1043 error = ovs_flow_cmd_fill_info(flow,
1044 ovs_header->dp_ifindex,
1045 reply, info->snd_portid,
1046 info->snd_seq, 0,
74ed7ab9
JS		 1047 OVS_FLOW_CMD_NEW,
1048 ufid_flags);
893f139b
JR		 1049 BUG_ON(error < 0);
1050 }
1051 ovs_unlock();
37bdc87b 1052
34ae932a 1053 ovs_nla_free_flow_actions_rcu(old_acts);
893f139b 1054 ovs_flow_free(new_flow, false);
37bdc87b 1055 }
893f139b
JR		 1056
1057 if (reply)
1058 ovs_notify(&dp_flow_genl_family, reply, info);
37bdc87b
JR		 1059 return 0;
1060
37bdc87b
JR		 1061err_unlock_ovs:
1062 ovs_unlock();
893f139b
JR		 1063 kfree_skb(reply);
1064err_kfree_acts:
34ae932a 1065 ovs_nla_free_flow_actions(acts);
893f139b
JR		 1066err_kfree_flow:
1067 ovs_flow_free(new_flow, false);
37bdc87b
JR		 1068error:
1069 return error;
1070}
ccb1352e 1071
2fdb957d 1072/* Factor out action copy to avoid "Wframe-larger-than=1024" warning. */
7f8a436e
JS		 1073static struct sw_flow_actions *get_flow_actions(struct net *net,
1074 const struct nlattr *a,
6b205b2c 1075 const struct sw_flow_key *key,
05da5898
JR		 1076 const struct sw_flow_mask *mask,
1077 bool log)
6b205b2c
JG		 1078{
1079 struct sw_flow_actions *acts;
1080 struct sw_flow_key masked_key;
1081 int error;
1082
6b205b2c 1083 ovs_flow_mask_key(&masked_key, key, mask);
7f8a436e 1084 error = ovs_nla_copy_actions(net, a, &masked_key, &acts, log);
6b205b2c 1085 if (error) {
05da5898
JR		 1086 OVS_NLERR(log,
1087 "Actions may not be safe on all matching packets");
6b205b2c
JG		 1088 return ERR_PTR(error);
1089 }
1090
1091 return acts;
1092}
1093
37bdc87b
JR		 1094static int ovs_flow_cmd_set(struct sk_buff *skb, struct genl_info *info)
1095{
7f8a436e 1096 struct net *net = sock_net(skb->sk);
37bdc87b
JR		 1097 struct nlattr **a = info->attrs;
1098 struct ovs_header *ovs_header = info->userhdr;
6b205b2c 1099 struct sw_flow_key key;
37bdc87b
JR		 1100 struct sw_flow *flow;
1101 struct sw_flow_mask mask;
1102 struct sk_buff *reply = NULL;
1103 struct datapath *dp;
893f139b 1104 struct sw_flow_actions *old_acts = NULL, *acts = NULL;
37bdc87b 1105 struct sw_flow_match match;
74ed7ab9
JS		 1106 struct sw_flow_id sfid;
1107 u32 ufid_flags = ovs_nla_get_ufid_flags(a[OVS_FLOW_ATTR_UFID_FLAGS]);
37bdc87b 1108 int error;
05da5898 1109 bool log = !a[OVS_FLOW_ATTR_PROBE];
74ed7ab9 1110 bool ufid_present;
37bdc87b
JR		 1111
1112 /* Extract key. */
1113 error = -EINVAL;
426cda5c 1114 if (!a[OVS_FLOW_ATTR_KEY]) {
05da5898 1115 OVS_NLERR(log, "Flow key attribute not present in set flow.");
37bdc87b 1116 goto error;
426cda5c 1117 }
37bdc87b 1118
74ed7ab9 1119 ufid_present = ovs_nla_get_ufid(&sfid, a[OVS_FLOW_ATTR_UFID], log);
37bdc87b 1120 ovs_match_init(&match, &key, &mask);
c2ac6673 1121 error = ovs_nla_get_match(net, &match, a[OVS_FLOW_ATTR_KEY],
05da5898 1122 a[OVS_FLOW_ATTR_MASK], log);
37bdc87b
JR		 1123 if (error)
1124 goto error;
1125
1126 /* Validate actions. */
1127 if (a[OVS_FLOW_ATTR_ACTIONS]) {
7f8a436e
JS		 1128 acts = get_flow_actions(net, a[OVS_FLOW_ATTR_ACTIONS], &key,
1129 &mask, log);
6b205b2c
JG		 1130 if (IS_ERR(acts)) {
1131 error = PTR_ERR(acts);
37bdc87b 1132 goto error;
893f139b 1133 }
893f139b 1134
2fdb957d 1135 /* Can allocate before locking if have acts. */
74ed7ab9
JS		 1136 reply = ovs_flow_cmd_alloc_info(acts, &sfid, info, false,
1137 ufid_flags);
893f139b
JR		 1138 if (IS_ERR(reply)) {
1139 error = PTR_ERR(reply);
1140 goto err_kfree_acts;
be52c9e9 1141 }
37bdc87b 1142 }
0e9796b4 1143
37bdc87b 1144 ovs_lock();
7f8a436e 1145 dp = get_dp(net, ovs_header->dp_ifindex);
893f139b
JR		 1146 if (unlikely(!dp)) {
1147 error = -ENODEV;
37bdc87b 1148 goto err_unlock_ovs;
893f139b 1149 }
37bdc87b 1150 /* Check that the flow exists. */
74ed7ab9
JS		 1151 if (ufid_present)
1152 flow = ovs_flow_tbl_lookup_ufid(&dp->table, &sfid);
1153 else
1154 flow = ovs_flow_tbl_lookup_exact(&dp->table, &match);
893f139b
JR		 1155 if (unlikely(!flow)) {
1156 error = -ENOENT;
37bdc87b 1157 goto err_unlock_ovs;
893f139b 1158 }
4a46b24e 1159
37bdc87b 1160 /* Update actions, if present. */
893f139b 1161 if (likely(acts)) {
37bdc87b
JR		 1162 old_acts = ovsl_dereference(flow->sf_acts);
1163 rcu_assign_pointer(flow->sf_acts, acts);
893f139b
JR		 1164
1165 if (unlikely(reply)) {
1166 error = ovs_flow_cmd_fill_info(flow,
1167 ovs_header->dp_ifindex,
1168 reply, info->snd_portid,
1169 info->snd_seq, 0,
74ed7ab9
JS		 1170 OVS_FLOW_CMD_NEW,
1171 ufid_flags);
893f139b
JR		 1172 BUG_ON(error < 0);
1173 }
1174 } else {
1175 /* Could not alloc without acts before locking. */
1176 reply = ovs_flow_cmd_build_info(flow, ovs_header->dp_ifindex,
74ed7ab9
JS		 1177 info, OVS_FLOW_CMD_NEW, false,
1178 ufid_flags);
1179
893f139b
JR		 1180 if (unlikely(IS_ERR(reply))) {
1181 error = PTR_ERR(reply);
1182 goto err_unlock_ovs;
1183 }
ccb1352e 1184 }
37bdc87b 1185
37bdc87b
JR		 1186 /* Clear stats. */
1187 if (a[OVS_FLOW_ATTR_CLEAR])
1188 ovs_flow_stats_clear(flow);
8e4e1713 1189 ovs_unlock();
ccb1352e 1190
893f139b
JR		 1191 if (reply)
1192 ovs_notify(&dp_flow_genl_family, reply, info);
1193 if (old_acts)
34ae932a 1194 ovs_nla_free_flow_actions_rcu(old_acts);
fb5d1e9e 1195
ccb1352e
JG		 1196 return 0;
1197
8e4e1713
PS		 1198err_unlock_ovs:
1199 ovs_unlock();
893f139b
JR		 1200 kfree_skb(reply);
1201err_kfree_acts:
34ae932a 1202 ovs_nla_free_flow_actions(acts);
ccb1352e
JG		 1203error:
1204 return error;
1205}
1206
1207static int ovs_flow_cmd_get(struct sk_buff *skb, struct genl_info *info)
1208{
1209 struct nlattr **a = info->attrs;
1210 struct ovs_header *ovs_header = info->userhdr;
c2ac6673 1211 struct net *net = sock_net(skb->sk);
ccb1352e
JG		 1212 struct sw_flow_key key;
1213 struct sk_buff *reply;
1214 struct sw_flow *flow;
1215 struct datapath *dp;
03f0d916 1216 struct sw_flow_match match;
74ed7ab9
JS		 1217 struct sw_flow_id ufid;
1218 u32 ufid_flags = ovs_nla_get_ufid_flags(a[OVS_FLOW_ATTR_UFID_FLAGS]);
1219 int err = 0;
05da5898 1220 bool log = !a[OVS_FLOW_ATTR_PROBE];
74ed7ab9 1221 bool ufid_present;
ccb1352e 1222
74ed7ab9
JS		 1223 ufid_present = ovs_nla_get_ufid(&ufid, a[OVS_FLOW_ATTR_UFID], log);
1224 if (a[OVS_FLOW_ATTR_KEY]) {
1225 ovs_match_init(&match, &key, NULL);
c2ac6673 1226 err = ovs_nla_get_match(net, &match, a[OVS_FLOW_ATTR_KEY], NULL,
74ed7ab9
JS		 1227 log);
1228 } else if (!ufid_present) {
05da5898
JR		 1229 OVS_NLERR(log,
1230 "Flow get message rejected, Key attribute missing.");
74ed7ab9 1231 err = -EINVAL;
03f0d916 1232 }
ccb1352e
JG		 1233 if (err)
1234 return err;
1235
8e4e1713 1236 ovs_lock();
46df7b81 1237 dp = get_dp(sock_net(skb->sk), ovs_header->dp_ifindex);
8e4e1713
PS		 1238 if (!dp) {
1239 err = -ENODEV;
1240 goto unlock;
1241 }
ccb1352e 1242
74ed7ab9
JS		 1243 if (ufid_present)
1244 flow = ovs_flow_tbl_lookup_ufid(&dp->table, &ufid);
1245 else
1246 flow = ovs_flow_tbl_lookup_exact(&dp->table, &match);
4a46b24e 1247 if (!flow) {
8e4e1713
PS		 1248 err = -ENOENT;
1249 goto unlock;
1250 }
ccb1352e 1251
0e9796b4 1252 reply = ovs_flow_cmd_build_info(flow, ovs_header->dp_ifindex, info,
74ed7ab9 1253 OVS_FLOW_CMD_NEW, true, ufid_flags);
8e4e1713
PS		 1254 if (IS_ERR(reply)) {
1255 err = PTR_ERR(reply);
1256 goto unlock;
1257 }
ccb1352e 1258
8e4e1713 1259 ovs_unlock();
ccb1352e 1260 return genlmsg_reply(reply, info);
8e4e1713
PS		 1261unlock:
1262 ovs_unlock();
1263 return err;
ccb1352e
JG		 1264}
1265
1266static int ovs_flow_cmd_del(struct sk_buff *skb, struct genl_info *info)
1267{
1268 struct nlattr **a = info->attrs;
1269 struct ovs_header *ovs_header = info->userhdr;
c2ac6673 1270 struct net *net = sock_net(skb->sk);
ccb1352e
JG		 1271 struct sw_flow_key key;
1272 struct sk_buff *reply;
74ed7ab9 1273 struct sw_flow *flow = NULL;
ccb1352e 1274 struct datapath *dp;
03f0d916 1275 struct sw_flow_match match;
74ed7ab9
JS		 1276 struct sw_flow_id ufid;
1277 u32 ufid_flags = ovs_nla_get_ufid_flags(a[OVS_FLOW_ATTR_UFID_FLAGS]);
ccb1352e 1278 int err;
05da5898 1279 bool log = !a[OVS_FLOW_ATTR_PROBE];
74ed7ab9 1280 bool ufid_present;
ccb1352e 1281
74ed7ab9
JS		 1282 ufid_present = ovs_nla_get_ufid(&ufid, a[OVS_FLOW_ATTR_UFID], log);
1283 if (a[OVS_FLOW_ATTR_KEY]) {
aed06778 1284 ovs_match_init(&match, &key, NULL);
c2ac6673
JS		 1285 err = ovs_nla_get_match(net, &match, a[OVS_FLOW_ATTR_KEY],
1286 NULL, log);
aed06778
JR		 1287 if (unlikely(err))
1288 return err;
1289 }
1290
8e4e1713 1291 ovs_lock();
46df7b81 1292 dp = get_dp(sock_net(skb->sk), ovs_header->dp_ifindex);
aed06778 1293 if (unlikely(!dp)) {
8e4e1713
PS		 1294 err = -ENODEV;
1295 goto unlock;
1296 }
46df7b81 1297
74ed7ab9 1298 if (unlikely(!a[OVS_FLOW_ATTR_KEY] && !ufid_present)) {
b637e498 1299 err = ovs_flow_tbl_flush(&dp->table);
8e4e1713
PS		 1300 goto unlock;
1301 }
03f0d916 1302
74ed7ab9
JS		 1303 if (ufid_present)
1304 flow = ovs_flow_tbl_lookup_ufid(&dp->table, &ufid);
1305 else
1306 flow = ovs_flow_tbl_lookup_exact(&dp->table, &match);
4a46b24e 1307 if (unlikely(!flow)) {
8e4e1713
PS		 1308 err = -ENOENT;
1309 goto unlock;
1310 }
ccb1352e 1311
b637e498 1312 ovs_flow_tbl_remove(&dp->table, flow);
aed06778 1313 ovs_unlock();
ccb1352e 1314
aed06778 1315 reply = ovs_flow_cmd_alloc_info((const struct sw_flow_actions __force *) flow->sf_acts,
74ed7ab9 1316 &flow->id, info, false, ufid_flags);
aed06778
JR		 1317 if (likely(reply)) {
1318 if (likely(!IS_ERR(reply))) {
1319 rcu_read_lock(); /*To keep RCU checker happy. */
1320 err = ovs_flow_cmd_fill_info(flow, ovs_header->dp_ifindex,
1321 reply, info->snd_portid,
1322 info->snd_seq, 0,
74ed7ab9
JS		 1323 OVS_FLOW_CMD_DEL,
1324 ufid_flags);
aed06778
JR		 1325 rcu_read_unlock();
1326 BUG_ON(err < 0);
1327
1328 ovs_notify(&dp_flow_genl_family, reply, info);
1329 } else {
1330 netlink_set_err(sock_net(skb->sk)->genl_sock, 0, 0, PTR_ERR(reply));
1331 }
fb5d1e9e 1332 }
ccb1352e 1333
aed06778 1334 ovs_flow_free(flow, true);
ccb1352e 1335 return 0;
8e4e1713
PS		 1336unlock:
1337 ovs_unlock();
1338 return err;
ccb1352e
JG		 1339}
1340
1341static int ovs_flow_cmd_dump(struct sk_buff *skb, struct netlink_callback *cb)
1342{
74ed7ab9 1343 struct nlattr *a[__OVS_FLOW_ATTR_MAX];
ccb1352e 1344 struct ovs_header *ovs_header = genlmsg_data(nlmsg_data(cb->nlh));
b637e498 1345 struct table_instance *ti;
ccb1352e 1346 struct datapath *dp;
74ed7ab9
JS		 1347 u32 ufid_flags;
1348 int err;
1349
1350 err = genlmsg_parse(cb->nlh, &dp_flow_genl_family, a,
1351 OVS_FLOW_ATTR_MAX, flow_policy);
1352 if (err)
1353 return err;
1354 ufid_flags = ovs_nla_get_ufid_flags(a[OVS_FLOW_ATTR_UFID_FLAGS]);
ccb1352e 1355
d57170b1 1356 rcu_read_lock();
cc3a5ae6 1357 dp = get_dp_rcu(sock_net(skb->sk), ovs_header->dp_ifindex);
8e4e1713 1358 if (!dp) {
d57170b1 1359 rcu_read_unlock();
ccb1352e 1360 return -ENODEV;
8e4e1713 1361 }
ccb1352e 1362
b637e498 1363 ti = rcu_dereference(dp->table.ti);
ccb1352e
JG		 1364 for (;;) {
1365 struct sw_flow *flow;
1366 u32 bucket, obj;
1367
1368 bucket = cb->args[0];
1369 obj = cb->args[1];
b637e498 1370 flow = ovs_flow_tbl_dump_next(ti, &bucket, &obj);
ccb1352e
JG		 1371 if (!flow)
1372 break;
1373
0e9796b4 1374 if (ovs_flow_cmd_fill_info(flow, ovs_header->dp_ifindex, skb,
15e47304 1375 NETLINK_CB(cb->skb).portid,
ccb1352e 1376 cb->nlh->nlmsg_seq, NLM_F_MULTI,
74ed7ab9 1377 OVS_FLOW_CMD_NEW, ufid_flags) < 0)
ccb1352e
JG		 1378 break;
1379
1380 cb->args[0] = bucket;
1381 cb->args[1] = obj;
1382 }
d57170b1 1383 rcu_read_unlock();
ccb1352e
JG		 1384 return skb->len;
1385}
1386
0c200ef9
PS		 1387static const struct nla_policy flow_policy[OVS_FLOW_ATTR_MAX + 1] = {
1388 [OVS_FLOW_ATTR_KEY] = { .type = NLA_NESTED },
05da5898 1389 [OVS_FLOW_ATTR_MASK] = { .type = NLA_NESTED },
0c200ef9
PS		 1390 [OVS_FLOW_ATTR_ACTIONS] = { .type = NLA_NESTED },
1391 [OVS_FLOW_ATTR_CLEAR] = { .type = NLA_FLAG },
05da5898 1392 [OVS_FLOW_ATTR_PROBE] = { .type = NLA_FLAG },
74ed7ab9
JS		 1393 [OVS_FLOW_ATTR_UFID] = { .type = NLA_UNSPEC, .len = 1 },
1394 [OVS_FLOW_ATTR_UFID_FLAGS] = { .type = NLA_U32 },
0c200ef9
PS		 1395};
1396
48e48a70 1397static const struct genl_ops dp_flow_genl_ops[] = {
ccb1352e
JG		 1398 { .cmd = OVS_FLOW_CMD_NEW,
1399 .flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
1400 .policy = flow_policy,
37bdc87b 1401 .doit = ovs_flow_cmd_new
ccb1352e
JG		 1402 },
1403 { .cmd = OVS_FLOW_CMD_DEL,
1404 .flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
1405 .policy = flow_policy,
1406 .doit = ovs_flow_cmd_del
1407 },
1408 { .cmd = OVS_FLOW_CMD_GET,
1409 .flags = 0, /* OK for unprivileged users. */
1410 .policy = flow_policy,
1411 .doit = ovs_flow_cmd_get,
1412 .dumpit = ovs_flow_cmd_dump
1413 },
1414 { .cmd = OVS_FLOW_CMD_SET,
1415 .flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
1416 .policy = flow_policy,
37bdc87b 1417 .doit = ovs_flow_cmd_set,
ccb1352e
JG		 1418 },
1419};
1420
0c200ef9 1421static struct genl_family dp_flow_genl_family = {
ccb1352e
JG		 1422 .id = GENL_ID_GENERATE,
1423 .hdrsize = sizeof(struct ovs_header),
0c200ef9
PS		 1424 .name = OVS_FLOW_FAMILY,
1425 .version = OVS_FLOW_VERSION,
1426 .maxattr = OVS_FLOW_ATTR_MAX,
3a4e0d6a
PS		 1427 .netnsok = true,
1428 .parallel_ops = true,
0c200ef9
PS		 1429 .ops = dp_flow_genl_ops,
1430 .n_ops = ARRAY_SIZE(dp_flow_genl_ops),
1431 .mcgrps = &ovs_dp_flow_multicast_group,
1432 .n_mcgrps = 1,
ccb1352e
JG		 1433};
1434
c3ff8cfe
TG		 1435static size_t ovs_dp_cmd_msg_size(void)
1436{
1437 size_t msgsize = NLMSG_ALIGN(sizeof(struct ovs_header));
1438
1439 msgsize += nla_total_size(IFNAMSIZ);
1440 msgsize += nla_total_size(sizeof(struct ovs_dp_stats));
1bd7116f 1441 msgsize += nla_total_size(sizeof(struct ovs_dp_megaflow_stats));
45fb9c35 1442 msgsize += nla_total_size(sizeof(u32)); /* OVS_DP_ATTR_USER_FEATURES */
c3ff8cfe
TG		 1443
1444 return msgsize;
1445}
1446
8ec609d8 1447/* Called with ovs_mutex. */
ccb1352e 1448static int ovs_dp_cmd_fill_info(struct datapath *dp, struct sk_buff *skb,
15e47304 1449 u32 portid, u32 seq, u32 flags, u8 cmd)
ccb1352e
JG		 1450{
1451 struct ovs_header *ovs_header;
1452 struct ovs_dp_stats dp_stats;
1bd7116f 1453 struct ovs_dp_megaflow_stats dp_megaflow_stats;
ccb1352e
JG		 1454 int err;
1455
15e47304 1456 ovs_header = genlmsg_put(skb, portid, seq, &dp_datapath_genl_family,
ccb1352e
JG		 1457 flags, cmd);
1458 if (!ovs_header)
1459 goto error;
1460
1461 ovs_header->dp_ifindex = get_dpifindex(dp);
1462
ccb1352e 1463 err = nla_put_string(skb, OVS_DP_ATTR_NAME, ovs_dp_name(dp));
ccb1352e
JG		 1464 if (err)
1465 goto nla_put_failure;
1466
1bd7116f
AZ		 1467 get_dp_stats(dp, &dp_stats, &dp_megaflow_stats);
1468 if (nla_put(skb, OVS_DP_ATTR_STATS, sizeof(struct ovs_dp_stats),
1469 &dp_stats))
1470 goto nla_put_failure;
1471
1472 if (nla_put(skb, OVS_DP_ATTR_MEGAFLOW_STATS,
1473 sizeof(struct ovs_dp_megaflow_stats),
1474 &dp_megaflow_stats))
028d6a67 1475 goto nla_put_failure;
ccb1352e 1476
43d4be9c
TG		 1477 if (nla_put_u32(skb, OVS_DP_ATTR_USER_FEATURES, dp->user_features))
1478 goto nla_put_failure;
1479
053c095a
JB		 1480 genlmsg_end(skb, ovs_header);
1481 return 0;
ccb1352e
JG		 1482
1483nla_put_failure:
1484 genlmsg_cancel(skb, ovs_header);
1485error:
1486 return -EMSGSIZE;
1487}
1488
6093ae9a 1489static struct sk_buff *ovs_dp_cmd_alloc_info(struct genl_info *info)
ccb1352e 1490{
6093ae9a 1491 return genlmsg_new_unicast(ovs_dp_cmd_msg_size(), info, GFP_KERNEL);
ccb1352e
JG		 1492}
1493
bb6f9a70 1494/* Called with rcu_read_lock or ovs_mutex. */
46df7b81 1495static struct datapath *lookup_datapath(struct net *net,
12eb18f7 1496 const struct ovs_header *ovs_header,
ccb1352e
JG		 1497 struct nlattr *a[OVS_DP_ATTR_MAX + 1])
1498{
1499 struct datapath *dp;
1500
1501 if (!a[OVS_DP_ATTR_NAME])
46df7b81 1502 dp = get_dp(net, ovs_header->dp_ifindex);
ccb1352e
JG		 1503 else {
1504 struct vport *vport;
1505
46df7b81 1506 vport = ovs_vport_locate(net, nla_data(a[OVS_DP_ATTR_NAME]));
ccb1352e 1507 dp = vport && vport->port_no == OVSP_LOCAL ? vport->dp : NULL;
ccb1352e
JG		 1508 }
1509 return dp ? dp : ERR_PTR(-ENODEV);
1510}
1511
44da5ae5
TG		 1512static void ovs_dp_reset_user_features(struct sk_buff *skb, struct genl_info *info)
1513{
1514 struct datapath *dp;
1515
1516 dp = lookup_datapath(sock_net(skb->sk), info->userhdr, info->attrs);
3c7eacfc 1517 if (IS_ERR(dp))
44da5ae5
TG		 1518 return;
1519
1520 WARN(dp->user_features, "Dropping previously announced user features\n");
1521 dp->user_features = 0;
1522}
1523
12eb18f7 1524static void ovs_dp_change(struct datapath *dp, struct nlattr *a[])
43d4be9c
TG		 1525{
1526 if (a[OVS_DP_ATTR_USER_FEATURES])
1527 dp->user_features = nla_get_u32(a[OVS_DP_ATTR_USER_FEATURES]);
1528}
1529
ccb1352e
JG		 1530static int ovs_dp_cmd_new(struct sk_buff *skb, struct genl_info *info)
1531{
1532 struct nlattr **a = info->attrs;
1533 struct vport_parms parms;
1534 struct sk_buff *reply;
1535 struct datapath *dp;
1536 struct vport *vport;
46df7b81 1537 struct ovs_net *ovs_net;
15eac2a7 1538 int err, i;
ccb1352e
JG		 1539
1540 err = -EINVAL;
1541 if (!a[OVS_DP_ATTR_NAME] || !a[OVS_DP_ATTR_UPCALL_PID])
1542 goto err;
1543
6093ae9a
JR		 1544 reply = ovs_dp_cmd_alloc_info(info);
1545 if (!reply)
1546 return -ENOMEM;
ccb1352e
JG		 1547
1548 err = -ENOMEM;
1549 dp = kzalloc(sizeof(*dp), GFP_KERNEL);
1550 if (dp == NULL)
6093ae9a 1551 goto err_free_reply;
46df7b81 1552
efd7ef1c 1553 ovs_dp_set_net(dp, sock_net(skb->sk));
ccb1352e
JG		 1554
1555 /* Allocate table. */
b637e498
PS		 1556 err = ovs_flow_tbl_init(&dp->table);
1557 if (err)
ccb1352e
JG		 1558 goto err_free_dp;
1559
1c213bd2 1560 dp->stats_percpu = netdev_alloc_pcpu_stats(struct dp_stats_percpu);
ccb1352e
JG		 1561 if (!dp->stats_percpu) {
1562 err = -ENOMEM;
1563 goto err_destroy_table;
1564 }
1565
15eac2a7 1566 dp->ports = kmalloc(DP_VPORT_HASH_BUCKETS * sizeof(struct hlist_head),
e6445719 1567 GFP_KERNEL);
15eac2a7
PS		 1568 if (!dp->ports) {
1569 err = -ENOMEM;
1570 goto err_destroy_percpu;
1571 }
1572
1573 for (i = 0; i < DP_VPORT_HASH_BUCKETS; i++)
1574 INIT_HLIST_HEAD(&dp->ports[i]);
1575
ccb1352e
JG		 1576 /* Set up our datapath device. */
1577 parms.name = nla_data(a[OVS_DP_ATTR_NAME]);
1578 parms.type = OVS_VPORT_TYPE_INTERNAL;
1579 parms.options = NULL;
1580 parms.dp = dp;
1581 parms.port_no = OVSP_LOCAL;
5cd667b0 1582 parms.upcall_portids = a[OVS_DP_ATTR_UPCALL_PID];
ccb1352e 1583
43d4be9c
TG		 1584 ovs_dp_change(dp, a);
1585
6093ae9a
JR		 1586 /* So far only local changes have been made, now need the lock. */
1587 ovs_lock();
1588
ccb1352e
JG		 1589 vport = new_vport(&parms);
1590 if (IS_ERR(vport)) {
1591 err = PTR_ERR(vport);
1592 if (err == -EBUSY)
1593 err = -EEXIST;
1594
44da5ae5
TG		 1595 if (err == -EEXIST) {
1596 /* An outdated user space instance that does not understand
1597 * the concept of user_features has attempted to create a new
1598 * datapath and is likely to reuse it. Drop all user features.
1599 */
1600 if (info->genlhdr->version < OVS_DP_VER_FEATURES)
1601 ovs_dp_reset_user_features(skb, info);
1602 }
1603
15eac2a7 1604 goto err_destroy_ports_array;
ccb1352e
JG		 1605 }
1606
6093ae9a
JR		 1607 err = ovs_dp_cmd_fill_info(dp, reply, info->snd_portid,
1608 info->snd_seq, 0, OVS_DP_CMD_NEW);
1609 BUG_ON(err < 0);
ccb1352e 1610
46df7b81 1611 ovs_net = net_generic(ovs_dp_get_net(dp), ovs_net_id);
59a35d60 1612 list_add_tail_rcu(&dp->list_node, &ovs_net->dps);
8e4e1713
PS		 1613
1614 ovs_unlock();
ccb1352e 1615
2a94fe48 1616 ovs_notify(&dp_datapath_genl_family, reply, info);
ccb1352e
JG		 1617 return 0;
1618
15eac2a7 1619err_destroy_ports_array:
6093ae9a 1620 ovs_unlock();
15eac2a7 1621 kfree(dp->ports);
ccb1352e
JG		 1622err_destroy_percpu:
1623 free_percpu(dp->stats_percpu);
1624err_destroy_table:
9b996e54 1625 ovs_flow_tbl_destroy(&dp->table);
ccb1352e
JG		 1626err_free_dp:
1627 kfree(dp);
6093ae9a
JR		 1628err_free_reply:
1629 kfree_skb(reply);
ccb1352e
JG		 1630err:
1631 return err;
1632}
1633
8e4e1713 1634/* Called with ovs_mutex. */
46df7b81 1635static void __dp_destroy(struct datapath *dp)
ccb1352e 1636{
15eac2a7 1637 int i;
ccb1352e 1638
15eac2a7
PS		 1639 for (i = 0; i < DP_VPORT_HASH_BUCKETS; i++) {
1640 struct vport *vport;
b67bfe0d 1641 struct hlist_node *n;
15eac2a7 1642
b67bfe0d 1643 hlist_for_each_entry_safe(vport, n, &dp->ports[i], dp_hash_node)
15eac2a7
PS		 1644 if (vport->port_no != OVSP_LOCAL)
1645 ovs_dp_detach_port(vport);
1646 }
ccb1352e 1647
59a35d60 1648 list_del_rcu(&dp->list_node);
ccb1352e 1649
8e4e1713 1650 /* OVSP_LOCAL is datapath internal port. We need to make sure that
e80857cc 1651 * all ports in datapath are destroyed first before freeing datapath.
ccb1352e 1652 */
8e4e1713 1653 ovs_dp_detach_port(ovs_vport_ovsl(dp, OVSP_LOCAL));
ccb1352e 1654
e80857cc 1655 /* RCU destroy the flow table */
ccb1352e 1656 call_rcu(&dp->rcu, destroy_dp_rcu);
46df7b81
PS		 1657}
1658
1659static int ovs_dp_cmd_del(struct sk_buff *skb, struct genl_info *info)
1660{
1661 struct sk_buff *reply;
1662 struct datapath *dp;
1663 int err;
1664
6093ae9a
JR		 1665 reply = ovs_dp_cmd_alloc_info(info);
1666 if (!reply)
1667 return -ENOMEM;
1668
8e4e1713 1669 ovs_lock();
46df7b81
PS		 1670 dp = lookup_datapath(sock_net(skb->sk), info->userhdr, info->attrs);
1671 err = PTR_ERR(dp);
1672 if (IS_ERR(dp))
6093ae9a 1673 goto err_unlock_free;
46df7b81 1674
6093ae9a
JR		 1675 err = ovs_dp_cmd_fill_info(dp, reply, info->snd_portid,
1676 info->snd_seq, 0, OVS_DP_CMD_DEL);
1677 BUG_ON(err < 0);
46df7b81
PS		 1678
1679 __dp_destroy(dp);
8e4e1713 1680 ovs_unlock();
ccb1352e 1681
2a94fe48 1682 ovs_notify(&dp_datapath_genl_family, reply, info);
ccb1352e
JG		 1683
1684 return 0;
6093ae9a
JR		 1685
1686err_unlock_free:
8e4e1713 1687 ovs_unlock();
6093ae9a 1688 kfree_skb(reply);
8e4e1713 1689 return err;
ccb1352e
JG		 1690}
1691
1692static int ovs_dp_cmd_set(struct sk_buff *skb, struct genl_info *info)
1693{
1694 struct sk_buff *reply;
1695 struct datapath *dp;
1696 int err;
1697
6093ae9a
JR		 1698 reply = ovs_dp_cmd_alloc_info(info);
1699 if (!reply)
1700 return -ENOMEM;
1701
8e4e1713 1702 ovs_lock();
46df7b81 1703 dp = lookup_datapath(sock_net(skb->sk), info->userhdr, info->attrs);
8e4e1713 1704 err = PTR_ERR(dp);
ccb1352e 1705 if (IS_ERR(dp))
6093ae9a 1706 goto err_unlock_free;
ccb1352e 1707
43d4be9c
TG		 1708 ovs_dp_change(dp, info->attrs);
1709
6093ae9a
JR		 1710 err = ovs_dp_cmd_fill_info(dp, reply, info->snd_portid,
1711 info->snd_seq, 0, OVS_DP_CMD_NEW);
1712 BUG_ON(err < 0);
ccb1352e 1713
8e4e1713 1714 ovs_unlock();
2a94fe48 1715 ovs_notify(&dp_datapath_genl_family, reply, info);
ccb1352e
JG		 1716
1717 return 0;
6093ae9a
JR		 1718
1719err_unlock_free:
8e4e1713 1720 ovs_unlock();
6093ae9a 1721 kfree_skb(reply);
8e4e1713 1722 return err;
ccb1352e
JG		 1723}
1724
1725static int ovs_dp_cmd_get(struct sk_buff *skb, struct genl_info *info)
1726{
1727 struct sk_buff *reply;
1728 struct datapath *dp;
8e4e1713 1729 int err;
ccb1352e 1730
6093ae9a
JR		 1731 reply = ovs_dp_cmd_alloc_info(info);
1732 if (!reply)
1733 return -ENOMEM;
1734
8ec609d8 1735 ovs_lock();
46df7b81 1736 dp = lookup_datapath(sock_net(skb->sk), info->userhdr, info->attrs);
8e4e1713
PS		 1737 if (IS_ERR(dp)) {
1738 err = PTR_ERR(dp);
6093ae9a 1739 goto err_unlock_free;
8e4e1713 1740 }
6093ae9a
JR		 1741 err = ovs_dp_cmd_fill_info(dp, reply, info->snd_portid,
1742 info->snd_seq, 0, OVS_DP_CMD_NEW);
1743 BUG_ON(err < 0);
8ec609d8 1744 ovs_unlock();
ccb1352e
JG		 1745
1746 return genlmsg_reply(reply, info);
8e4e1713 1747
6093ae9a 1748err_unlock_free:
8ec609d8 1749 ovs_unlock();
6093ae9a 1750 kfree_skb(reply);
8e4e1713 1751 return err;
ccb1352e
JG		 1752}
1753
1754static int ovs_dp_cmd_dump(struct sk_buff *skb, struct netlink_callback *cb)
1755{
46df7b81 1756 struct ovs_net *ovs_net = net_generic(sock_net(skb->sk), ovs_net_id);
ccb1352e
JG		 1757 struct datapath *dp;
1758 int skip = cb->args[0];
1759 int i = 0;
1760
8ec609d8
PS		 1761 ovs_lock();
1762 list_for_each_entry(dp, &ovs_net->dps, list_node) {
77676fdb 1763 if (i >= skip &&
15e47304 1764 ovs_dp_cmd_fill_info(dp, skb, NETLINK_CB(cb->skb).portid,
ccb1352e
JG		 1765 cb->nlh->nlmsg_seq, NLM_F_MULTI,
1766 OVS_DP_CMD_NEW) < 0)
1767 break;
1768 i++;
1769 }
8ec609d8 1770 ovs_unlock();
ccb1352e
JG		 1771
1772 cb->args[0] = i;
1773
1774 return skb->len;
1775}
1776
0c200ef9
PS		 1777static const struct nla_policy datapath_policy[OVS_DP_ATTR_MAX + 1] = {
1778 [OVS_DP_ATTR_NAME] = { .type = NLA_NUL_STRING, .len = IFNAMSIZ - 1 },
1779 [OVS_DP_ATTR_UPCALL_PID] = { .type = NLA_U32 },
1780 [OVS_DP_ATTR_USER_FEATURES] = { .type = NLA_U32 },
1781};
1782
48e48a70 1783static const struct genl_ops dp_datapath_genl_ops[] = {
ccb1352e
JG		 1784 { .cmd = OVS_DP_CMD_NEW,
1785 .flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
1786 .policy = datapath_policy,
1787 .doit = ovs_dp_cmd_new
1788 },
1789 { .cmd = OVS_DP_CMD_DEL,
1790 .flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
1791 .policy = datapath_policy,
1792 .doit = ovs_dp_cmd_del
1793 },
1794 { .cmd = OVS_DP_CMD_GET,
1795 .flags = 0, /* OK for unprivileged users. */
1796 .policy = datapath_policy,
1797 .doit = ovs_dp_cmd_get,
1798 .dumpit = ovs_dp_cmd_dump
1799 },
1800 { .cmd = OVS_DP_CMD_SET,
1801 .flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
1802 .policy = datapath_policy,
1803 .doit = ovs_dp_cmd_set,
1804 },
1805};
1806
0c200ef9 1807static struct genl_family dp_datapath_genl_family = {
ccb1352e
JG		 1808 .id = GENL_ID_GENERATE,
1809 .hdrsize = sizeof(struct ovs_header),
0c200ef9
PS		 1810 .name = OVS_DATAPATH_FAMILY,
1811 .version = OVS_DATAPATH_VERSION,
1812 .maxattr = OVS_DP_ATTR_MAX,
3a4e0d6a
PS		 1813 .netnsok = true,
1814 .parallel_ops = true,
0c200ef9
PS		 1815 .ops = dp_datapath_genl_ops,
1816 .n_ops = ARRAY_SIZE(dp_datapath_genl_ops),
1817 .mcgrps = &ovs_dp_datapath_multicast_group,
1818 .n_mcgrps = 1,
ccb1352e
JG		 1819};
1820
8e4e1713 1821/* Called with ovs_mutex or RCU read lock. */
ccb1352e 1822static int ovs_vport_cmd_fill_info(struct vport *vport, struct sk_buff *skb,
15e47304 1823 u32 portid, u32 seq, u32 flags, u8 cmd)
ccb1352e
JG		 1824{
1825 struct ovs_header *ovs_header;
1826 struct ovs_vport_stats vport_stats;
1827 int err;
1828
15e47304 1829 ovs_header = genlmsg_put(skb, portid, seq, &dp_vport_genl_family,
ccb1352e
JG		 1830 flags, cmd);
1831 if (!ovs_header)
1832 return -EMSGSIZE;
1833
1834 ovs_header->dp_ifindex = get_dpifindex(vport->dp);
1835
028d6a67
DM		 1836 if (nla_put_u32(skb, OVS_VPORT_ATTR_PORT_NO, vport->port_no) ||
1837 nla_put_u32(skb, OVS_VPORT_ATTR_TYPE, vport->ops->type) ||
5cd667b0 1838 nla_put_string(skb, OVS_VPORT_ATTR_NAME,
c9db965c 1839 ovs_vport_name(vport)))
028d6a67 1840 goto nla_put_failure;
ccb1352e
JG		 1841
1842 ovs_vport_get_stats(vport, &vport_stats);
028d6a67
DM		 1843 if (nla_put(skb, OVS_VPORT_ATTR_STATS, sizeof(struct ovs_vport_stats),
1844 &vport_stats))
1845 goto nla_put_failure;
ccb1352e 1846
5cd667b0
AW		 1847 if (ovs_vport_get_upcall_portids(vport, skb))
1848 goto nla_put_failure;
1849
ccb1352e
JG		 1850 err = ovs_vport_get_options(vport, skb);
1851 if (err == -EMSGSIZE)
1852 goto error;
1853
053c095a
JB		 1854 genlmsg_end(skb, ovs_header);
1855 return 0;
ccb1352e
JG		 1856
1857nla_put_failure:
1858 err = -EMSGSIZE;
1859error:
1860 genlmsg_cancel(skb, ovs_header);
1861 return err;
1862}
1863
6093ae9a
JR		 1864static struct sk_buff *ovs_vport_cmd_alloc_info(void)
1865{
1866 return nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
1867}
1868
1869/* Called with ovs_mutex, only via ovs_dp_notify_wq(). */
15e47304 1870struct sk_buff *ovs_vport_cmd_build_info(struct vport *vport, u32 portid,
ccb1352e
JG		 1871 u32 seq, u8 cmd)
1872{
1873 struct sk_buff *skb;
1874 int retval;
1875
1876 skb = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_ATOMIC);
1877 if (!skb)
1878 return ERR_PTR(-ENOMEM);
1879
15e47304 1880 retval = ovs_vport_cmd_fill_info(vport, skb, portid, seq, 0, cmd);
a9341512
JG		 1881 BUG_ON(retval < 0);
1882
ccb1352e
JG		 1883 return skb;
1884}
1885
8e4e1713 1886/* Called with ovs_mutex or RCU read lock. */
46df7b81 1887static struct vport *lookup_vport(struct net *net,
12eb18f7 1888 const struct ovs_header *ovs_header,
ccb1352e
JG		 1889 struct nlattr *a[OVS_VPORT_ATTR_MAX + 1])
1890{
1891 struct datapath *dp;
1892 struct vport *vport;
1893
1894 if (a[OVS_VPORT_ATTR_NAME]) {
46df7b81 1895 vport = ovs_vport_locate(net, nla_data(a[OVS_VPORT_ATTR_NAME]));
ccb1352e
JG		 1896 if (!vport)
1897 return ERR_PTR(-ENODEV);
651a68ea
BP		 1898 if (ovs_header->dp_ifindex &&
1899 ovs_header->dp_ifindex != get_dpifindex(vport->dp))
1900 return ERR_PTR(-ENODEV);
ccb1352e
JG		 1901 return vport;
1902 } else if (a[OVS_VPORT_ATTR_PORT_NO]) {
1903 u32 port_no = nla_get_u32(a[OVS_VPORT_ATTR_PORT_NO]);
1904
1905 if (port_no >= DP_MAX_PORTS)
1906 return ERR_PTR(-EFBIG);
1907
46df7b81 1908 dp = get_dp(net, ovs_header->dp_ifindex);
ccb1352e
JG		 1909 if (!dp)
1910 return ERR_PTR(-ENODEV);
1911
8e4e1713 1912 vport = ovs_vport_ovsl_rcu(dp, port_no);
ccb1352e 1913 if (!vport)
14408dba 1914 return ERR_PTR(-ENODEV);
ccb1352e
JG		 1915 return vport;
1916 } else
1917 return ERR_PTR(-EINVAL);
1918}
1919
1920static int ovs_vport_cmd_new(struct sk_buff *skb, struct genl_info *info)
1921{
1922 struct nlattr **a = info->attrs;
1923 struct ovs_header *ovs_header = info->userhdr;
1924 struct vport_parms parms;
1925 struct sk_buff *reply;
1926 struct vport *vport;
1927 struct datapath *dp;
1928 u32 port_no;
1929 int err;
1930
ccb1352e
JG		 1931 if (!a[OVS_VPORT_ATTR_NAME] || !a[OVS_VPORT_ATTR_TYPE] ||
1932 !a[OVS_VPORT_ATTR_UPCALL_PID])
6093ae9a
JR		 1933 return -EINVAL;
1934
1935 port_no = a[OVS_VPORT_ATTR_PORT_NO]
1936 ? nla_get_u32(a[OVS_VPORT_ATTR_PORT_NO]) : 0;
1937 if (port_no >= DP_MAX_PORTS)
1938 return -EFBIG;
1939
1940 reply = ovs_vport_cmd_alloc_info();
1941 if (!reply)
1942 return -ENOMEM;
ccb1352e 1943
8e4e1713 1944 ovs_lock();
62b9c8d0 1945restart:
46df7b81 1946 dp = get_dp(sock_net(skb->sk), ovs_header->dp_ifindex);
ccb1352e
JG		 1947 err = -ENODEV;
1948 if (!dp)
6093ae9a 1949 goto exit_unlock_free;
ccb1352e 1950
6093ae9a 1951 if (port_no) {
8e4e1713 1952 vport = ovs_vport_ovsl(dp, port_no);
ccb1352e
JG		 1953 err = -EBUSY;
1954 if (vport)
6093ae9a 1955 goto exit_unlock_free;
ccb1352e
JG		 1956 } else {
1957 for (port_no = 1; ; port_no++) {
1958 if (port_no >= DP_MAX_PORTS) {
1959 err = -EFBIG;
6093ae9a 1960 goto exit_unlock_free;
ccb1352e 1961 }
8e4e1713 1962 vport = ovs_vport_ovsl(dp, port_no);
ccb1352e
JG		 1963 if (!vport)
1964 break;
1965 }
1966 }
1967
1968 parms.name = nla_data(a[OVS_VPORT_ATTR_NAME]);
1969 parms.type = nla_get_u32(a[OVS_VPORT_ATTR_TYPE]);
1970 parms.options = a[OVS_VPORT_ATTR_OPTIONS];
1971 parms.dp = dp;
1972 parms.port_no = port_no;
5cd667b0 1973 parms.upcall_portids = a[OVS_VPORT_ATTR_UPCALL_PID];
ccb1352e
JG		 1974
1975 vport = new_vport(&parms);
1976 err = PTR_ERR(vport);
62b9c8d0
TG		 1977 if (IS_ERR(vport)) {
1978 if (err == -EAGAIN)
1979 goto restart;
6093ae9a 1980 goto exit_unlock_free;
62b9c8d0 1981 }
ccb1352e 1982
6093ae9a
JR		 1983 err = ovs_vport_cmd_fill_info(vport, reply, info->snd_portid,
1984 info->snd_seq, 0, OVS_VPORT_CMD_NEW);
1985 BUG_ON(err < 0);
1986 ovs_unlock();
ed661185 1987
2a94fe48 1988 ovs_notify(&dp_vport_genl_family, reply, info);
6093ae9a 1989 return 0;
ccb1352e 1990
6093ae9a 1991exit_unlock_free:
8e4e1713 1992 ovs_unlock();
6093ae9a 1993 kfree_skb(reply);
ccb1352e
JG		 1994 return err;
1995}
1996
1997static int ovs_vport_cmd_set(struct sk_buff *skb, struct genl_info *info)
1998{
1999 struct nlattr **a = info->attrs;
2000 struct sk_buff *reply;
2001 struct vport *vport;
2002 int err;
2003
6093ae9a
JR		 2004 reply = ovs_vport_cmd_alloc_info();
2005 if (!reply)
2006 return -ENOMEM;
2007
8e4e1713 2008 ovs_lock();
46df7b81 2009 vport = lookup_vport(sock_net(skb->sk), info->userhdr, a);
ccb1352e
JG		 2010 err = PTR_ERR(vport);
2011 if (IS_ERR(vport))
6093ae9a 2012 goto exit_unlock_free;
ccb1352e 2013
ccb1352e 2014 if (a[OVS_VPORT_ATTR_TYPE] &&
f44f3408 2015 nla_get_u32(a[OVS_VPORT_ATTR_TYPE]) != vport->ops->type) {
ccb1352e 2016 err = -EINVAL;
6093ae9a 2017 goto exit_unlock_free;
a9341512
JG		 2018 }
2019
f44f3408 2020 if (a[OVS_VPORT_ATTR_OPTIONS]) {
ccb1352e 2021 err = ovs_vport_set_options(vport, a[OVS_VPORT_ATTR_OPTIONS]);
f44f3408 2022 if (err)
6093ae9a 2023 goto exit_unlock_free;
f44f3408 2024 }
a9341512 2025
5cd667b0
AW		 2026
2027 if (a[OVS_VPORT_ATTR_UPCALL_PID]) {
2028 struct nlattr *ids = a[OVS_VPORT_ATTR_UPCALL_PID];
2029
2030 err = ovs_vport_set_upcall_portids(vport, ids);
2031 if (err)
2032 goto exit_unlock_free;
2033 }
ccb1352e 2034
a9341512
JG		 2035 err = ovs_vport_cmd_fill_info(vport, reply, info->snd_portid,
2036 info->snd_seq, 0, OVS_VPORT_CMD_NEW);
2037 BUG_ON(err < 0);
ccb1352e 2038
8e4e1713 2039 ovs_unlock();
2a94fe48 2040 ovs_notify(&dp_vport_genl_family, reply, info);
8e4e1713 2041 return 0;
ccb1352e 2042
6093ae9a 2043exit_unlock_free:
8e4e1713 2044 ovs_unlock();
6093ae9a 2045 kfree_skb(reply);
ccb1352e
JG		 2046 return err;
2047}
2048
2049static int ovs_vport_cmd_del(struct sk_buff *skb, struct genl_info *info)
2050{
2051 struct nlattr **a = info->attrs;
2052 struct sk_buff *reply;
2053 struct vport *vport;
2054 int err;
2055
6093ae9a
JR		 2056 reply = ovs_vport_cmd_alloc_info();
2057 if (!reply)
2058 return -ENOMEM;
2059
8e4e1713 2060 ovs_lock();
46df7b81 2061 vport = lookup_vport(sock_net(skb->sk), info->userhdr, a);
ccb1352e
JG		 2062 err = PTR_ERR(vport);
2063 if (IS_ERR(vport))
6093ae9a 2064 goto exit_unlock_free;
ccb1352e
JG		 2065
2066 if (vport->port_no == OVSP_LOCAL) {
2067 err = -EINVAL;
6093ae9a 2068 goto exit_unlock_free;
ccb1352e
JG		 2069 }
2070
6093ae9a
JR		 2071 err = ovs_vport_cmd_fill_info(vport, reply, info->snd_portid,
2072 info->snd_seq, 0, OVS_VPORT_CMD_DEL);
2073 BUG_ON(err < 0);
ccb1352e 2074 ovs_dp_detach_port(vport);
6093ae9a 2075 ovs_unlock();
ccb1352e 2076
2a94fe48 2077 ovs_notify(&dp_vport_genl_family, reply, info);
6093ae9a 2078 return 0;
ccb1352e 2079
6093ae9a 2080exit_unlock_free:
8e4e1713 2081 ovs_unlock();
6093ae9a 2082 kfree_skb(reply);
ccb1352e
JG		 2083 return err;
2084}
2085
2086static int ovs_vport_cmd_get(struct sk_buff *skb, struct genl_info *info)
2087{
2088 struct nlattr **a = info->attrs;
2089 struct ovs_header *ovs_header = info->userhdr;
2090 struct sk_buff *reply;
2091 struct vport *vport;
2092 int err;
2093
6093ae9a
JR		 2094 reply = ovs_vport_cmd_alloc_info();
2095 if (!reply)
2096 return -ENOMEM;
2097
ccb1352e 2098 rcu_read_lock();
46df7b81 2099 vport = lookup_vport(sock_net(skb->sk), ovs_header, a);
ccb1352e
JG		 2100 err = PTR_ERR(vport);
2101 if (IS_ERR(vport))
6093ae9a
JR		 2102 goto exit_unlock_free;
2103 err = ovs_vport_cmd_fill_info(vport, reply, info->snd_portid,
2104 info->snd_seq, 0, OVS_VPORT_CMD_NEW);
2105 BUG_ON(err < 0);
ccb1352e
JG		 2106 rcu_read_unlock();
2107
2108 return genlmsg_reply(reply, info);
2109
6093ae9a 2110exit_unlock_free:
ccb1352e 2111 rcu_read_unlock();
6093ae9a 2112 kfree_skb(reply);
ccb1352e
JG		 2113 return err;
2114}
2115
2116static int ovs_vport_cmd_dump(struct sk_buff *skb, struct netlink_callback *cb)
2117{
2118 struct ovs_header *ovs_header = genlmsg_data(nlmsg_data(cb->nlh));
2119 struct datapath *dp;
15eac2a7
PS		 2120 int bucket = cb->args[0], skip = cb->args[1];
2121 int i, j = 0;
ccb1352e 2122
42ee19e2 2123 rcu_read_lock();
cc3a5ae6 2124 dp = get_dp_rcu(sock_net(skb->sk), ovs_header->dp_ifindex);
42ee19e2
JR		 2125 if (!dp) {
2126 rcu_read_unlock();
ccb1352e 2127 return -ENODEV;
42ee19e2 2128 }
15eac2a7 2129 for (i = bucket; i < DP_VPORT_HASH_BUCKETS; i++) {
ccb1352e 2130 struct vport *vport;
15eac2a7
PS		 2131
2132 j = 0;
b67bfe0d 2133 hlist_for_each_entry_rcu(vport, &dp->ports[i], dp_hash_node) {
15eac2a7
PS		 2134 if (j >= skip &&
2135 ovs_vport_cmd_fill_info(vport, skb,
15e47304 2136 NETLINK_CB(cb->skb).portid,
15eac2a7
PS		 2137 cb->nlh->nlmsg_seq,
2138 NLM_F_MULTI,
2139 OVS_VPORT_CMD_NEW) < 0)
2140 goto out;
2141
2142 j++;
2143 }
2144 skip = 0;
ccb1352e 2145 }
15eac2a7 2146out:
ccb1352e
JG		 2147 rcu_read_unlock();
2148
15eac2a7
PS		 2149 cb->args[0] = i;
2150 cb->args[1] = j;
ccb1352e 2151
15eac2a7 2152 return skb->len;
ccb1352e
JG		 2153}
2154
0c200ef9
PS		 2155static const struct nla_policy vport_policy[OVS_VPORT_ATTR_MAX + 1] = {
2156 [OVS_VPORT_ATTR_NAME] = { .type = NLA_NUL_STRING, .len = IFNAMSIZ - 1 },
2157 [OVS_VPORT_ATTR_STATS] = { .len = sizeof(struct ovs_vport_stats) },
2158 [OVS_VPORT_ATTR_PORT_NO] = { .type = NLA_U32 },
2159 [OVS_VPORT_ATTR_TYPE] = { .type = NLA_U32 },
2160 [OVS_VPORT_ATTR_UPCALL_PID] = { .type = NLA_U32 },
2161 [OVS_VPORT_ATTR_OPTIONS] = { .type = NLA_NESTED },
2162};
2163
48e48a70 2164static const struct genl_ops dp_vport_genl_ops[] = {
ccb1352e
JG		 2165 { .cmd = OVS_VPORT_CMD_NEW,
2166 .flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
2167 .policy = vport_policy,
2168 .doit = ovs_vport_cmd_new
2169 },
2170 { .cmd = OVS_VPORT_CMD_DEL,
2171 .flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
2172 .policy = vport_policy,
2173 .doit = ovs_vport_cmd_del
2174 },
2175 { .cmd = OVS_VPORT_CMD_GET,
2176 .flags = 0, /* OK for unprivileged users. */
2177 .policy = vport_policy,
2178 .doit = ovs_vport_cmd_get,
2179 .dumpit = ovs_vport_cmd_dump
2180 },
2181 { .cmd = OVS_VPORT_CMD_SET,
2182 .flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
2183 .policy = vport_policy,
2184 .doit = ovs_vport_cmd_set,
2185 },
2186};
2187
0c200ef9
PS		 2188struct genl_family dp_vport_genl_family = {
2189 .id = GENL_ID_GENERATE,
2190 .hdrsize = sizeof(struct ovs_header),
2191 .name = OVS_VPORT_FAMILY,
2192 .version = OVS_VPORT_VERSION,
2193 .maxattr = OVS_VPORT_ATTR_MAX,
2194 .netnsok = true,
2195 .parallel_ops = true,
2196 .ops = dp_vport_genl_ops,
2197 .n_ops = ARRAY_SIZE(dp_vport_genl_ops),
2198 .mcgrps = &ovs_dp_vport_multicast_group,
2199 .n_mcgrps = 1,
ccb1352e
JG		 2200};
2201
0c200ef9
PS		 2202static struct genl_family * const dp_genl_families[] = {
2203 &dp_datapath_genl_family,
2204 &dp_vport_genl_family,
2205 &dp_flow_genl_family,
2206 &dp_packet_genl_family,
ccb1352e
JG		 2207};
2208
2209static void dp_unregister_genl(int n_families)
2210{
2211 int i;
2212
2213 for (i = 0; i < n_families; i++)
0c200ef9 2214 genl_unregister_family(dp_genl_families[i]);
ccb1352e
JG		 2215}
2216
2217static int dp_register_genl(void)
2218{
ccb1352e
JG		 2219 int err;
2220 int i;
2221
ccb1352e 2222 for (i = 0; i < ARRAY_SIZE(dp_genl_families); i++) {
ccb1352e 2223
0c200ef9 2224 err = genl_register_family(dp_genl_families[i]);
ccb1352e
JG		 2225 if (err)
2226 goto error;
ccb1352e
JG		 2227 }
2228
2229 return 0;
2230
2231error:
0c200ef9 2232 dp_unregister_genl(i);
ccb1352e
JG		 2233 return err;
2234}
2235
46df7b81
PS		 2236static int __net_init ovs_init_net(struct net *net)
2237{
2238 struct ovs_net *ovs_net = net_generic(net, ovs_net_id);
2239
2240 INIT_LIST_HEAD(&ovs_net->dps);
8e4e1713 2241 INIT_WORK(&ovs_net->dp_notify_work, ovs_dp_notify_wq);
c2ac6673 2242 ovs_ct_init(net);
46df7b81
PS		 2243 return 0;
2244}
2245
7b4577a9
PS		 2246static void __net_exit list_vports_from_net(struct net *net, struct net *dnet,
2247 struct list_head *head)
46df7b81 2248{
8e4e1713 2249 struct ovs_net *ovs_net = net_generic(net, ovs_net_id);
7b4577a9
PS		 2250 struct datapath *dp;
2251
2252 list_for_each_entry(dp, &ovs_net->dps, list_node) {
2253 int i;
2254
2255 for (i = 0; i < DP_VPORT_HASH_BUCKETS; i++) {
2256 struct vport *vport;
2257
2258 hlist_for_each_entry(vport, &dp->ports[i], dp_hash_node) {
7b4577a9
PS		 2259 if (vport->ops->type != OVS_VPORT_TYPE_INTERNAL)
2260 continue;
2261
be4ace6e 2262 if (dev_net(vport->dev) == dnet)
7b4577a9
PS		 2263 list_add(&vport->detach_list, head);
2264 }
2265 }
2266 }
2267}
2268
2269static void __net_exit ovs_exit_net(struct net *dnet)
2270{
2271 struct datapath *dp, *dp_next;
2272 struct ovs_net *ovs_net = net_generic(dnet, ovs_net_id);
2273 struct vport *vport, *vport_next;
2274 struct net *net;
2275 LIST_HEAD(head);
46df7b81 2276
c2ac6673 2277 ovs_ct_exit(dnet);
8e4e1713 2278 ovs_lock();
46df7b81
PS		 2279 list_for_each_entry_safe(dp, dp_next, &ovs_net->dps, list_node)
2280 __dp_destroy(dp);
7b4577a9
PS		 2281
2282 rtnl_lock();
2283 for_each_net(net)
2284 list_vports_from_net(net, dnet, &head);
2285 rtnl_unlock();
2286
2287 /* Detach all vports from given namespace. */
2288 list_for_each_entry_safe(vport, vport_next, &head, detach_list) {
2289 list_del(&vport->detach_list);
2290 ovs_dp_detach_port(vport);
2291 }
2292
8e4e1713
PS		 2293 ovs_unlock();
2294
2295 cancel_work_sync(&ovs_net->dp_notify_work);
46df7b81
PS		 2296}
2297
2298static struct pernet_operations ovs_net_ops = {
2299 .init = ovs_init_net,
2300 .exit = ovs_exit_net,
2301 .id = &ovs_net_id,
2302 .size = sizeof(struct ovs_net),
2303};
2304
ccb1352e
JG		 2305static int __init dp_init(void)
2306{
ccb1352e
JG		 2307 int err;
2308
3523b29b 2309 BUILD_BUG_ON(sizeof(struct ovs_skb_cb) > FIELD_SIZEOF(struct sk_buff, cb));
ccb1352e
JG		 2310
2311 pr_info("Open vSwitch switching datapath\n");
2312
971427f3 2313 err = action_fifos_init();
ccb1352e
JG		 2314 if (err)
2315 goto error;
2316
971427f3
AZ		 2317 err = ovs_internal_dev_rtnl_link_register();
2318 if (err)
2319 goto error_action_fifos_exit;
2320
5b9e7e16
JP		 2321 err = ovs_flow_init();
2322 if (err)
2323 goto error_unreg_rtnl_link;
2324
ccb1352e
JG		 2325 err = ovs_vport_init();
2326 if (err)
2327 goto error_flow_exit;
2328
46df7b81 2329 err = register_pernet_device(&ovs_net_ops);
ccb1352e
JG		 2330 if (err)
2331 goto error_vport_exit;
2332
46df7b81
PS		 2333 err = register_netdevice_notifier(&ovs_dp_device_notifier);
2334 if (err)
2335 goto error_netns_exit;
2336
62b9c8d0
TG		 2337 err = ovs_netdev_init();
2338 if (err)
2339 goto error_unreg_notifier;
2340
ccb1352e
JG		 2341 err = dp_register_genl();
2342 if (err < 0)
62b9c8d0 2343 goto error_unreg_netdev;
ccb1352e 2344
ccb1352e
JG		 2345 return 0;
2346
62b9c8d0
TG		 2347error_unreg_netdev:
2348 ovs_netdev_exit();
ccb1352e
JG		 2349error_unreg_notifier:
2350 unregister_netdevice_notifier(&ovs_dp_device_notifier);
46df7b81
PS		 2351error_netns_exit:
2352 unregister_pernet_device(&ovs_net_ops);
ccb1352e
JG		 2353error_vport_exit:
2354 ovs_vport_exit();
2355error_flow_exit:
2356 ovs_flow_exit();
5b9e7e16
JP		 2357error_unreg_rtnl_link:
2358 ovs_internal_dev_rtnl_link_unregister();
971427f3
AZ		 2359error_action_fifos_exit:
2360 action_fifos_exit();
ccb1352e
JG		 2361error:
2362 return err;
2363}
2364
2365static void dp_cleanup(void)
2366{
ccb1352e 2367 dp_unregister_genl(ARRAY_SIZE(dp_genl_families));
62b9c8d0 2368 ovs_netdev_exit();
ccb1352e 2369 unregister_netdevice_notifier(&ovs_dp_device_notifier);
46df7b81
PS		 2370 unregister_pernet_device(&ovs_net_ops);
2371 rcu_barrier();
ccb1352e
JG		 2372 ovs_vport_exit();
2373 ovs_flow_exit();
5b9e7e16 2374 ovs_internal_dev_rtnl_link_unregister();
971427f3 2375 action_fifos_exit();
ccb1352e
JG		 2376}
2377
2378module_init(dp_init);
2379module_exit(dp_cleanup);
2380
2381MODULE_DESCRIPTION("Open vSwitch switching datapath");
2382MODULE_LICENSE("GPL");
Linux kernel - Deliverables
This page took 0.360972 seconds and 5 git commands to generate.