datapath.c 46.9 KB
Newer Older
1
/*
2
 * Copyright (c) 2007-2013 Nicira, Inc.
3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46
 *
 * This program is free software; you can redistribute it and/or
 * modify it under the terms of version 2 of the GNU General Public
 * License as published by the Free Software Foundation.
 *
 * This program is distributed in the hope that it will be useful, but
 * WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
 * General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with this program; if not, write to the Free Software
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
 * 02110-1301, USA
 */

#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt

#include <linux/init.h>
#include <linux/module.h>
#include <linux/if_arp.h>
#include <linux/if_vlan.h>
#include <linux/in.h>
#include <linux/ip.h>
#include <linux/jhash.h>
#include <linux/delay.h>
#include <linux/time.h>
#include <linux/etherdevice.h>
#include <linux/genetlink.h>
#include <linux/kernel.h>
#include <linux/kthread.h>
#include <linux/mutex.h>
#include <linux/percpu.h>
#include <linux/rcupdate.h>
#include <linux/tcp.h>
#include <linux/udp.h>
#include <linux/ethtool.h>
#include <linux/wait.h>
#include <asm/div64.h>
#include <linux/highmem.h>
#include <linux/netfilter_bridge.h>
#include <linux/netfilter_ipv4.h>
#include <linux/inetdevice.h>
#include <linux/list.h>
47
#include <linux/lockdep.h>
48 49 50 51 52
#include <linux/openvswitch.h>
#include <linux/rculist.h>
#include <linux/dmi.h>
#include <linux/workqueue.h>
#include <net/genetlink.h>
53 54
#include <net/net_namespace.h>
#include <net/netns/generic.h>
55 56 57

#include "datapath.h"
#include "flow.h"
58
#include "flow_table.h"
59
#include "flow_netlink.h"
60
#include "vport-internal_dev.h"
61
#include "vport-netdev.h"
62

63 64
int ovs_net_id __read_mostly;

65
static void ovs_notify(struct genl_family *family,
66
		       struct sk_buff *skb, struct genl_info *info)
67
{
68
	genl_notify(family, skb, genl_info_net(info), info->snd_portid,
69
		    0, info->nlhdr, GFP_KERNEL);
70 71
}

72 73 74
/**
 * DOC: Locking:
 *
75 76 77 78
 * All writes e.g. Writes to device state (add/remove datapath, port, set
 * operations on vports, etc.), Writes to other state (flow table
 * modifications, set miscellaneous datapath parameters, etc.) are protected
 * by ovs_lock.
79 80 81 82 83 84
 *
 * Reads are protected by RCU.
 *
 * There are a few special cases (mostly stats) that have their own
 * synchronization but they nest under all of above and don't interact with
 * each other.
85 86
 *
 * The RTNL lock nests inside ovs_mutex.
87 88
 */

89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110
static DEFINE_MUTEX(ovs_mutex);

void ovs_lock(void)
{
	mutex_lock(&ovs_mutex);
}

void ovs_unlock(void)
{
	mutex_unlock(&ovs_mutex);
}

#ifdef CONFIG_LOCKDEP
int lockdep_ovsl_is_held(void)
{
	if (debug_locks)
		return lockdep_is_held(&ovs_mutex);
	else
		return 1;
}
#endif

111
static struct vport *new_vport(const struct vport_parms *);
112
static int queue_gso_packets(struct datapath *dp, struct sk_buff *,
113
			     const struct dp_upcall_info *);
114
static int queue_userspace_packet(struct datapath *dp, struct sk_buff *,
115 116
				  const struct dp_upcall_info *);

117
/* Must be called with rcu_read_lock or ovs_mutex. */
118
static struct datapath *get_dp(struct net *net, int dp_ifindex)
119 120 121 122 123
{
	struct datapath *dp = NULL;
	struct net_device *dev;

	rcu_read_lock();
124
	dev = dev_get_by_index_rcu(net, dp_ifindex);
125 126 127 128 129 130 131 132 133 134
	if (dev) {
		struct vport *vport = ovs_internal_dev_get_vport(dev);
		if (vport)
			dp = vport->dp;
	}
	rcu_read_unlock();

	return dp;
}

135
/* Must be called with rcu_read_lock or ovs_mutex. */
S
Stephen Hemminger 已提交
136
static const char *ovs_dp_name(const struct datapath *dp)
137
{
138
	struct vport *vport = ovs_vport_ovsl_rcu(dp, OVSP_LOCAL);
139 140 141 142 143 144 145 146 147 148
	return vport->ops->get_name(vport);
}

static int get_dpifindex(struct datapath *dp)
{
	struct vport *local;
	int ifindex;

	rcu_read_lock();

149
	local = ovs_vport_rcu(dp, OVSP_LOCAL);
150
	if (local)
151
		ifindex = netdev_vport_priv(local)->dev->ifindex;
152 153 154 155 156 157 158 159 160 161 162 163 164
	else
		ifindex = 0;

	rcu_read_unlock();

	return ifindex;
}

static void destroy_dp_rcu(struct rcu_head *rcu)
{
	struct datapath *dp = container_of(rcu, struct datapath, rcu);

	free_percpu(dp->stats_percpu);
165
	release_net(ovs_dp_get_net(dp));
166
	kfree(dp->ports);
167 168 169
	kfree(dp);
}

170 171 172 173 174 175 176 177 178 179 180 181
static struct hlist_head *vport_hash_bucket(const struct datapath *dp,
					    u16 port_no)
{
	return &dp->ports[port_no & (DP_VPORT_HASH_BUCKETS - 1)];
}

struct vport *ovs_lookup_vport(const struct datapath *dp, u16 port_no)
{
	struct vport *vport;
	struct hlist_head *head;

	head = vport_hash_bucket(dp, port_no);
182
	hlist_for_each_entry_rcu(vport, head, dp_hash_node) {
183 184 185 186 187 188
		if (vport->port_no == port_no)
			return vport;
	}
	return NULL;
}

189
/* Called with ovs_mutex. */
190 191 192 193 194 195 196
static struct vport *new_vport(const struct vport_parms *parms)
{
	struct vport *vport;

	vport = ovs_vport_add(parms);
	if (!IS_ERR(vport)) {
		struct datapath *dp = parms->dp;
197
		struct hlist_head *head = vport_hash_bucket(dp, vport->port_no);
198

199
		hlist_add_head_rcu(&vport->dp_hash_node, head);
200 201 202 203 204 205
	}
	return vport;
}

void ovs_dp_detach_port(struct vport *p)
{
206
	ASSERT_OVSL();
207 208

	/* First drop references to device. */
209
	hlist_del_rcu(&p->dp_hash_node);
210 211 212 213 214 215 216 217 218 219 220 221 222

	/* Then destroy it. */
	ovs_vport_del(p);
}

/* Must be called with rcu_read_lock. */
void ovs_dp_process_received_packet(struct vport *p, struct sk_buff *skb)
{
	struct datapath *dp = p->dp;
	struct sw_flow *flow;
	struct dp_stats_percpu *stats;
	struct sw_flow_key key;
	u64 *stats_counter;
223
	u32 n_mask_hit;
224 225
	int error;

226
	stats = this_cpu_ptr(dp->stats_percpu);
227 228

	/* Extract flow from 'skb' into 'key'. */
229
	error = ovs_flow_extract(skb, p->port_no, &key);
230 231 232 233 234 235
	if (unlikely(error)) {
		kfree_skb(skb);
		return;
	}

	/* Look up flow. */
236
	flow = ovs_flow_tbl_lookup_stats(&dp->table, &key, &n_mask_hit);
237 238 239 240 241 242
	if (unlikely(!flow)) {
		struct dp_upcall_info upcall;

		upcall.cmd = OVS_PACKET_CMD_MISS;
		upcall.key = &key;
		upcall.userdata = NULL;
243
		upcall.portid = p->upcall_portid;
244 245 246 247 248 249 250
		ovs_dp_upcall(dp, skb, &upcall);
		consume_skb(skb);
		stats_counter = &stats->n_missed;
		goto out;
	}

	OVS_CB(skb)->flow = flow;
251
	OVS_CB(skb)->pkt_key = &key;
252

253
	ovs_flow_stats_update(OVS_CB(skb)->flow, skb);
254
	ovs_execute_actions(dp, skb);
255
	stats_counter = &stats->n_hit;
256 257 258 259 260

out:
	/* Update datapath statistics. */
	u64_stats_update_begin(&stats->sync);
	(*stats_counter)++;
261
	stats->n_mask_hit += n_mask_hit;
262 263 264 265 266 267 268 269
	u64_stats_update_end(&stats->sync);
}

static struct genl_family dp_packet_genl_family = {
	.id = GENL_ID_GENERATE,
	.hdrsize = sizeof(struct ovs_header),
	.name = OVS_PACKET_FAMILY,
	.version = OVS_PACKET_VERSION,
270
	.maxattr = OVS_PACKET_ATTR_MAX,
271 272
	.netnsok = true,
	.parallel_ops = true,
273 274 275
};

int ovs_dp_upcall(struct datapath *dp, struct sk_buff *skb,
276
		  const struct dp_upcall_info *upcall_info)
277 278 279 280
{
	struct dp_stats_percpu *stats;
	int err;

281
	if (upcall_info->portid == 0) {
282 283 284 285 286
		err = -ENOTCONN;
		goto err;
	}

	if (!skb_is_gso(skb))
287
		err = queue_userspace_packet(dp, skb, upcall_info);
288
	else
289
		err = queue_gso_packets(dp, skb, upcall_info);
290 291 292 293 294 295
	if (err)
		goto err;

	return 0;

err:
296
	stats = this_cpu_ptr(dp->stats_percpu);
297 298 299 300 301 302 303 304

	u64_stats_update_begin(&stats->sync);
	stats->n_lost++;
	u64_stats_update_end(&stats->sync);

	return err;
}

305
static int queue_gso_packets(struct datapath *dp, struct sk_buff *skb,
306 307
			     const struct dp_upcall_info *upcall_info)
{
308
	unsigned short gso_type = skb_shinfo(skb)->gso_type;
309 310 311 312 313
	struct dp_upcall_info later_info;
	struct sw_flow_key later_key;
	struct sk_buff *segs, *nskb;
	int err;

314
	segs = __skb_gso_segment(skb, NETIF_F_SG, false);
315 316
	if (IS_ERR(segs))
		return PTR_ERR(segs);
317 318 319 320

	/* Queue all of the segments. */
	skb = segs;
	do {
321
		err = queue_userspace_packet(dp, skb, upcall_info);
322 323 324
		if (err)
			break;

325
		if (skb == segs && gso_type & SKB_GSO_UDP) {
326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350
			/* The initial flow key extracted by ovs_flow_extract()
			 * in this case is for a first fragment, so we need to
			 * properly mark later fragments.
			 */
			later_key = *upcall_info->key;
			later_key.ip.frag = OVS_FRAG_TYPE_LATER;

			later_info = *upcall_info;
			later_info.key = &later_key;
			upcall_info = &later_info;
		}
	} while ((skb = skb->next));

	/* Free all of the segments. */
	skb = segs;
	do {
		nskb = skb->next;
		if (err)
			kfree_skb(skb);
		else
			consume_skb(skb);
	} while ((skb = nskb));
	return err;
}

351 352 353
static size_t key_attr_size(void)
{
	return    nla_total_size(4)   /* OVS_KEY_ATTR_PRIORITY */
354 355 356 357 358 359 360 361
		+ nla_total_size(0)   /* OVS_KEY_ATTR_TUNNEL */
		  + nla_total_size(8)   /* OVS_TUNNEL_KEY_ATTR_ID */
		  + nla_total_size(4)   /* OVS_TUNNEL_KEY_ATTR_IPV4_SRC */
		  + nla_total_size(4)   /* OVS_TUNNEL_KEY_ATTR_IPV4_DST */
		  + nla_total_size(1)   /* OVS_TUNNEL_KEY_ATTR_TOS */
		  + nla_total_size(1)   /* OVS_TUNNEL_KEY_ATTR_TTL */
		  + nla_total_size(0)   /* OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT */
		  + nla_total_size(0)   /* OVS_TUNNEL_KEY_ATTR_CSUM */
362 363 364 365 366 367 368 369 370 371 372 373
		+ nla_total_size(4)   /* OVS_KEY_ATTR_IN_PORT */
		+ nla_total_size(4)   /* OVS_KEY_ATTR_SKB_MARK */
		+ nla_total_size(12)  /* OVS_KEY_ATTR_ETHERNET */
		+ nla_total_size(2)   /* OVS_KEY_ATTR_ETHERTYPE */
		+ nla_total_size(4)   /* OVS_KEY_ATTR_8021Q */
		+ nla_total_size(0)   /* OVS_KEY_ATTR_ENCAP */
		+ nla_total_size(2)   /* OVS_KEY_ATTR_ETHERTYPE */
		+ nla_total_size(40)  /* OVS_KEY_ATTR_IPV6 */
		+ nla_total_size(2)   /* OVS_KEY_ATTR_ICMPV6 */
		+ nla_total_size(28); /* OVS_KEY_ATTR_ND */
}

374 375
static size_t upcall_msg_size(const struct nlattr *userdata,
			      unsigned int hdrlen)
376 377
{
	size_t size = NLMSG_ALIGN(sizeof(struct ovs_header))
378
		+ nla_total_size(hdrlen) /* OVS_PACKET_ATTR_PACKET */
379 380 381 382 383 384 385 386 387
		+ nla_total_size(key_attr_size()); /* OVS_PACKET_ATTR_KEY */

	/* OVS_PACKET_ATTR_USERDATA */
	if (userdata)
		size += NLA_ALIGN(userdata->nla_len);

	return size;
}

388
static int queue_userspace_packet(struct datapath *dp, struct sk_buff *skb,
389 390 391 392 393 394
				  const struct dp_upcall_info *upcall_info)
{
	struct ovs_header *upcall;
	struct sk_buff *nskb = NULL;
	struct sk_buff *user_skb; /* to be queued to userspace */
	struct nlattr *nla;
395
	struct genl_info info = {
396
		.dst_sk = ovs_dp_get_net(dp)->genl_sock,
397 398 399
		.snd_portid = upcall_info->portid,
	};
	size_t len;
400
	unsigned int hlen;
401 402 403 404 405
	int err, dp_ifindex;

	dp_ifindex = get_dpifindex(dp);
	if (!dp_ifindex)
		return -ENODEV;
406 407 408 409 410 411

	if (vlan_tx_tag_present(skb)) {
		nskb = skb_clone(skb, GFP_ATOMIC);
		if (!nskb)
			return -ENOMEM;

412
		nskb = __vlan_put_tag(nskb, nskb->vlan_proto, vlan_tx_tag_get(nskb));
413
		if (!nskb)
414 415 416 417 418 419 420 421 422 423 424
			return -ENOMEM;

		nskb->vlan_tci = 0;
		skb = nskb;
	}

	if (nla_attr_size(skb->len) > USHRT_MAX) {
		err = -EFBIG;
		goto out;
	}

425 426 427 428 429 430 431 432 433 434 435 436 437 438 439
	/* Complete checksum if needed */
	if (skb->ip_summed == CHECKSUM_PARTIAL &&
	    (err = skb_checksum_help(skb)))
		goto out;

	/* Older versions of OVS user space enforce alignment of the last
	 * Netlink attribute to NLA_ALIGNTO which would require extensive
	 * padding logic. Only perform zerocopy if padding is not required.
	 */
	if (dp->user_features & OVS_DP_F_UNALIGNED)
		hlen = skb_zerocopy_headlen(skb);
	else
		hlen = skb->len;

	len = upcall_msg_size(upcall_info->userdata, hlen);
440
	user_skb = genlmsg_new_unicast(len, &info, GFP_ATOMIC);
441 442 443 444 445 446 447 448 449 450
	if (!user_skb) {
		err = -ENOMEM;
		goto out;
	}

	upcall = genlmsg_put(user_skb, 0, 0, &dp_packet_genl_family,
			     0, upcall_info->cmd);
	upcall->dp_ifindex = dp_ifindex;

	nla = nla_nest_start(user_skb, OVS_PACKET_ATTR_KEY);
451
	ovs_nla_put_flow(upcall_info->key, upcall_info->key, user_skb);
452 453 454
	nla_nest_end(user_skb, nla);

	if (upcall_info->userdata)
455 456 457
		__nla_put(user_skb, OVS_PACKET_ATTR_USERDATA,
			  nla_len(upcall_info->userdata),
			  nla_data(upcall_info->userdata));
458

459 460 461 462 463 464 465
	/* Only reserve room for attribute header, packet data is added
	 * in skb_zerocopy() */
	if (!(nla = nla_reserve(user_skb, OVS_PACKET_ATTR_PACKET, 0))) {
		err = -ENOBUFS;
		goto out;
	}
	nla->nla_len = nla_attr_size(skb->len);
466

467
	skb_zerocopy(user_skb, skb, skb->len, hlen);
468

469 470 471 472 473 474 475 476
	/* Pad OVS_PACKET_ATTR_PACKET if linear copy was performed */
	if (!(dp->user_features & OVS_DP_F_UNALIGNED)) {
		size_t plen = NLA_ALIGN(user_skb->len) - user_skb->len;

		if (plen > 0)
			memset(skb_put(user_skb, plen), 0, plen);
	}

477
	((struct nlmsghdr *) user_skb->data)->nlmsg_len = user_skb->len;
478

479
	err = genlmsg_unicast(ovs_dp_get_net(dp), user_skb, upcall_info->portid);
480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498
out:
	kfree_skb(nskb);
	return err;
}

static int ovs_packet_cmd_execute(struct sk_buff *skb, struct genl_info *info)
{
	struct ovs_header *ovs_header = info->userhdr;
	struct nlattr **a = info->attrs;
	struct sw_flow_actions *acts;
	struct sk_buff *packet;
	struct sw_flow *flow;
	struct datapath *dp;
	struct ethhdr *eth;
	int len;
	int err;

	err = -EINVAL;
	if (!a[OVS_PACKET_ATTR_PACKET] || !a[OVS_PACKET_ATTR_KEY] ||
499
	    !a[OVS_PACKET_ATTR_ACTIONS])
500 501 502 503 504 505 506 507 508
		goto err;

	len = nla_len(a[OVS_PACKET_ATTR_PACKET]);
	packet = __dev_alloc_skb(NET_IP_ALIGN + len, GFP_KERNEL);
	err = -ENOMEM;
	if (!packet)
		goto err;
	skb_reserve(packet, NET_IP_ALIGN);

509
	nla_memcpy(__skb_put(packet, len), a[OVS_PACKET_ATTR_PACKET], len);
510 511 512 513 514 515 516

	skb_reset_mac_header(packet);
	eth = eth_hdr(packet);

	/* Normally, setting the skb 'protocol' field would be handled by a
	 * call to eth_type_trans(), but it assumes there's a sending
	 * device, which we may not have. */
S
Simon Horman 已提交
517
	if (ntohs(eth->h_proto) >= ETH_P_802_3_MIN)
518 519 520 521 522
		packet->protocol = eth->h_proto;
	else
		packet->protocol = htons(ETH_P_802_2);

	/* Build an sw_flow for sending this packet. */
523
	flow = ovs_flow_alloc(false);
524 525 526 527
	err = PTR_ERR(flow);
	if (IS_ERR(flow))
		goto err_kfree_skb;

528
	err = ovs_flow_extract(packet, -1, &flow->key);
529 530 531
	if (err)
		goto err_flow_free;

532
	err = ovs_nla_get_flow_metadata(flow, a[OVS_PACKET_ATTR_KEY]);
533 534
	if (err)
		goto err_flow_free;
535
	acts = ovs_nla_alloc_flow_actions(nla_len(a[OVS_PACKET_ATTR_ACTIONS]));
536 537 538
	err = PTR_ERR(acts);
	if (IS_ERR(acts))
		goto err_flow_free;
539

540 541
	err = ovs_nla_copy_actions(a[OVS_PACKET_ATTR_ACTIONS],
				   &flow->key, 0, &acts);
542
	rcu_assign_pointer(flow->sf_acts, acts);
543 544
	if (err)
		goto err_flow_free;
545 546

	OVS_CB(packet)->flow = flow;
547
	OVS_CB(packet)->pkt_key = &flow->key;
548
	packet->priority = flow->key.phy.priority;
549
	packet->mark = flow->key.phy.skb_mark;
550 551

	rcu_read_lock();
552
	dp = get_dp(sock_net(skb->sk), ovs_header->dp_ifindex);
553 554 555 556 557 558 559 560 561
	err = -ENODEV;
	if (!dp)
		goto err_unlock;

	local_bh_disable();
	err = ovs_execute_actions(dp, packet);
	local_bh_enable();
	rcu_read_unlock();

562
	ovs_flow_free(flow, false);
563 564 565 566 567
	return err;

err_unlock:
	rcu_read_unlock();
err_flow_free:
568
	ovs_flow_free(flow, false);
569 570 571 572 573 574 575
err_kfree_skb:
	kfree_skb(packet);
err:
	return err;
}

static const struct nla_policy packet_policy[OVS_PACKET_ATTR_MAX + 1] = {
576
	[OVS_PACKET_ATTR_PACKET] = { .len = ETH_HLEN },
577 578 579 580
	[OVS_PACKET_ATTR_KEY] = { .type = NLA_NESTED },
	[OVS_PACKET_ATTR_ACTIONS] = { .type = NLA_NESTED },
};

581
static const struct genl_ops dp_packet_genl_ops[] = {
582 583 584 585 586 587 588
	{ .cmd = OVS_PACKET_CMD_EXECUTE,
	  .flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
	  .policy = packet_policy,
	  .doit = ovs_packet_cmd_execute
	}
};

589 590
static void get_dp_stats(struct datapath *dp, struct ovs_dp_stats *stats,
			 struct ovs_dp_megaflow_stats *mega_stats)
591 592 593
{
	int i;

594 595
	memset(mega_stats, 0, sizeof(*mega_stats));

596
	stats->n_flows = ovs_flow_tbl_count(&dp->table);
597
	mega_stats->n_masks = ovs_flow_tbl_num_masks(&dp->table);
598 599

	stats->n_hit = stats->n_missed = stats->n_lost = 0;
600

601 602 603 604 605 606 607 608 609 610 611 612 613 614 615
	for_each_possible_cpu(i) {
		const struct dp_stats_percpu *percpu_stats;
		struct dp_stats_percpu local_stats;
		unsigned int start;

		percpu_stats = per_cpu_ptr(dp->stats_percpu, i);

		do {
			start = u64_stats_fetch_begin_bh(&percpu_stats->sync);
			local_stats = *percpu_stats;
		} while (u64_stats_fetch_retry_bh(&percpu_stats->sync, start));

		stats->n_hit += local_stats.n_hit;
		stats->n_missed += local_stats.n_missed;
		stats->n_lost += local_stats.n_lost;
616
		mega_stats->n_mask_hit += local_stats.n_mask_hit;
617 618 619 620 621 622 623 624 625 626 627 628 629 630
	}
}

static const struct nla_policy flow_policy[OVS_FLOW_ATTR_MAX + 1] = {
	[OVS_FLOW_ATTR_KEY] = { .type = NLA_NESTED },
	[OVS_FLOW_ATTR_ACTIONS] = { .type = NLA_NESTED },
	[OVS_FLOW_ATTR_CLEAR] = { .type = NLA_FLAG },
};

static struct genl_family dp_flow_genl_family = {
	.id = GENL_ID_GENERATE,
	.hdrsize = sizeof(struct ovs_header),
	.name = OVS_FLOW_FAMILY,
	.version = OVS_FLOW_VERSION,
631
	.maxattr = OVS_FLOW_ATTR_MAX,
632 633
	.netnsok = true,
	.parallel_ops = true,
634 635 636 637 638 639
};

static struct genl_multicast_group ovs_dp_flow_multicast_group = {
	.name = OVS_FLOW_MCGROUP
};

640 641 642 643
static size_t ovs_flow_cmd_msg_size(const struct sw_flow_actions *acts)
{
	return NLMSG_ALIGN(sizeof(struct ovs_header))
		+ nla_total_size(key_attr_size()) /* OVS_FLOW_ATTR_KEY */
644
		+ nla_total_size(key_attr_size()) /* OVS_FLOW_ATTR_MASK */
645 646 647 648 649 650
		+ nla_total_size(sizeof(struct ovs_flow_stats)) /* OVS_FLOW_ATTR_STATS */
		+ nla_total_size(1) /* OVS_FLOW_ATTR_TCP_FLAGS */
		+ nla_total_size(8) /* OVS_FLOW_ATTR_USED */
		+ nla_total_size(acts->actions_len); /* OVS_FLOW_ATTR_ACTIONS */
}

651
/* Called with ovs_mutex. */
652
static int ovs_flow_cmd_fill_info(struct sw_flow *flow, struct datapath *dp,
653
				  struct sk_buff *skb, u32 portid,
654 655 656
				  u32 seq, u32 flags, u8 cmd)
{
	const int skb_orig_len = skb->len;
657
	struct nlattr *start;
658
	struct ovs_flow_stats stats;
659 660
	__be16 tcp_flags;
	unsigned long used;
661 662 663 664
	struct ovs_header *ovs_header;
	struct nlattr *nla;
	int err;

665
	ovs_header = genlmsg_put(skb, portid, seq, &dp_flow_genl_family, flags, cmd);
666 667 668 669 670
	if (!ovs_header)
		return -EMSGSIZE;

	ovs_header->dp_ifindex = get_dpifindex(dp);

671
	/* Fill flow key. */
672 673 674
	nla = nla_nest_start(skb, OVS_FLOW_ATTR_KEY);
	if (!nla)
		goto nla_put_failure;
675

676
	err = ovs_nla_put_flow(&flow->unmasked_key, &flow->unmasked_key, skb);
677 678 679 680 681 682 683 684
	if (err)
		goto error;
	nla_nest_end(skb, nla);

	nla = nla_nest_start(skb, OVS_FLOW_ATTR_MASK);
	if (!nla)
		goto nla_put_failure;

685
	err = ovs_nla_put_flow(&flow->key, &flow->mask->key, skb);
686 687
	if (err)
		goto error;
688

689 690
	nla_nest_end(skb, nla);

691
	ovs_flow_stats_get(flow, &stats, &used, &tcp_flags);
692 693 694
	if (used &&
	    nla_put_u64(skb, OVS_FLOW_ATTR_USED, ovs_flow_used_time(used)))
		goto nla_put_failure;
695

696
	if (stats.n_packets &&
697
	    nla_put(skb, OVS_FLOW_ATTR_STATS, sizeof(struct ovs_flow_stats), &stats))
698
		goto nla_put_failure;
699

700 701
	if ((u8)ntohs(tcp_flags) &&
	     nla_put_u8(skb, OVS_FLOW_ATTR_TCP_FLAGS, (u8)ntohs(tcp_flags)))
702
		goto nla_put_failure;
703 704 705 706 707 708 709 710 711 712 713

	/* If OVS_FLOW_ATTR_ACTIONS doesn't fit, skip dumping the actions if
	 * this is the first flow to be dumped into 'skb'.  This is unusual for
	 * Netlink but individual action lists can be longer than
	 * NLMSG_GOODSIZE and thus entirely undumpable if we didn't do this.
	 * The userspace caller can always fetch the actions separately if it
	 * really wants them.  (Most userspace callers in fact don't care.)
	 *
	 * This can only fail for dump operations because the skb is always
	 * properly sized for single flows.
	 */
714 715
	start = nla_nest_start(skb, OVS_FLOW_ATTR_ACTIONS);
	if (start) {
716 717
		const struct sw_flow_actions *sf_acts;

718
		sf_acts = rcu_dereference_ovsl(flow->sf_acts);
719

720 721
		err = ovs_nla_put_actions(sf_acts->actions,
					  sf_acts->actions_len, skb);
722 723 724 725 726 727 728 729 730 731
		if (!err)
			nla_nest_end(skb, start);
		else {
			if (skb_orig_len)
				goto error;

			nla_nest_cancel(skb, start);
		}
	} else if (skb_orig_len)
		goto nla_put_failure;
732 733 734 735 736 737 738 739 740 741

	return genlmsg_end(skb, ovs_header);

nla_put_failure:
	err = -EMSGSIZE;
error:
	genlmsg_cancel(skb, ovs_header);
	return err;
}

742 743
static struct sk_buff *ovs_flow_cmd_alloc_info(struct sw_flow *flow,
					       struct genl_info *info)
744
{
745
	size_t len;
746

747
	len = ovs_flow_cmd_msg_size(ovsl_dereference(flow->sf_acts));
748

749
	return genlmsg_new_unicast(len, info, GFP_KERNEL);
750 751 752 753
}

static struct sk_buff *ovs_flow_cmd_build_info(struct sw_flow *flow,
					       struct datapath *dp,
754 755
					       struct genl_info *info,
					       u8 cmd)
756 757 758 759
{
	struct sk_buff *skb;
	int retval;

760
	skb = ovs_flow_cmd_alloc_info(flow, info);
761 762 763
	if (!skb)
		return ERR_PTR(-ENOMEM);

764 765
	retval = ovs_flow_cmd_fill_info(flow, dp, skb, info->snd_portid,
					info->snd_seq, 0, cmd);
766 767 768 769 770 771 772 773
	BUG_ON(retval < 0);
	return skb;
}

static int ovs_flow_cmd_new_or_set(struct sk_buff *skb, struct genl_info *info)
{
	struct nlattr **a = info->attrs;
	struct ovs_header *ovs_header = info->userhdr;
774 775 776
	struct sw_flow_key key, masked_key;
	struct sw_flow *flow = NULL;
	struct sw_flow_mask mask;
777 778
	struct sk_buff *reply;
	struct datapath *dp;
779
	struct sw_flow_actions *acts = NULL;
780
	struct sw_flow_match match;
781
	bool exact_5tuple;
782 783 784 785 786 787
	int error;

	/* Extract key. */
	error = -EINVAL;
	if (!a[OVS_FLOW_ATTR_KEY])
		goto error;
788 789

	ovs_match_init(&match, &key, &mask);
790
	error = ovs_nla_get_match(&match, &exact_5tuple,
791
				  a[OVS_FLOW_ATTR_KEY], a[OVS_FLOW_ATTR_MASK]);
792 793 794 795 796
	if (error)
		goto error;

	/* Validate actions. */
	if (a[OVS_FLOW_ATTR_ACTIONS]) {
797
		acts = ovs_nla_alloc_flow_actions(nla_len(a[OVS_FLOW_ATTR_ACTIONS]));
798 799
		error = PTR_ERR(acts);
		if (IS_ERR(acts))
800
			goto error;
801

802 803 804
		ovs_flow_mask_key(&masked_key, &key, &mask);
		error = ovs_nla_copy_actions(a[OVS_FLOW_ATTR_ACTIONS],
					     &masked_key, 0, &acts);
805 806
		if (error) {
			OVS_NLERR("Flow actions may not be safe on all matching packets.\n");
807
			goto err_kfree;
808
		}
809 810 811 812 813
	} else if (info->genlhdr->cmd == OVS_FLOW_CMD_NEW) {
		error = -EINVAL;
		goto error;
	}

814
	ovs_lock();
815
	dp = get_dp(sock_net(skb->sk), ovs_header->dp_ifindex);
816 817
	error = -ENODEV;
	if (!dp)
818
		goto err_unlock_ovs;
819

820
	/* Check if this is a duplicate flow */
821
	flow = ovs_flow_tbl_lookup(&dp->table, &key);
822 823 824 825
	if (!flow) {
		/* Bail out if we're not allowed to create a new flow. */
		error = -ENOENT;
		if (info->genlhdr->cmd == OVS_FLOW_CMD_SET)
826
			goto err_unlock_ovs;
827 828

		/* Allocate flow. */
829
		flow = ovs_flow_alloc(!exact_5tuple);
830 831
		if (IS_ERR(flow)) {
			error = PTR_ERR(flow);
832
			goto err_unlock_ovs;
833 834
		}

835 836
		flow->key = masked_key;
		flow->unmasked_key = key;
837 838 839
		rcu_assign_pointer(flow->sf_acts, acts);

		/* Put flow in bucket. */
840 841 842 843 844
		error = ovs_flow_tbl_insert(&dp->table, flow, &mask);
		if (error) {
			acts = NULL;
			goto err_flow_free;
		}
845

846
		reply = ovs_flow_cmd_build_info(flow, dp, info, OVS_FLOW_CMD_NEW);
847 848 849 850 851 852 853 854 855 856 857 858 859
	} else {
		/* We found a matching flow. */
		struct sw_flow_actions *old_acts;

		/* Bail out if we're not allowed to modify an existing flow.
		 * We accept NLM_F_CREATE in place of the intended NLM_F_EXCL
		 * because Generic Netlink treats the latter as a dump
		 * request.  We also accept NLM_F_EXCL in case that bug ever
		 * gets fixed.
		 */
		error = -EEXIST;
		if (info->genlhdr->cmd == OVS_FLOW_CMD_NEW &&
		    info->nlhdr->nlmsg_flags & (NLM_F_CREATE | NLM_F_EXCL))
860
			goto err_unlock_ovs;
861

862
		/* The unmasked key has to be the same for flow updates. */
863
		if (!ovs_flow_cmp_unmasked_key(flow, &match))
864 865
			goto err_unlock_ovs;

866
		/* Update actions. */
867
		old_acts = ovsl_dereference(flow->sf_acts);
868
		rcu_assign_pointer(flow->sf_acts, acts);
869
		ovs_nla_free_flow_actions(old_acts);
870

871
		reply = ovs_flow_cmd_build_info(flow, dp, info, OVS_FLOW_CMD_NEW);
872 873

		/* Clear stats. */
874 875
		if (a[OVS_FLOW_ATTR_CLEAR])
			ovs_flow_stats_clear(flow);
876
	}
877
	ovs_unlock();
878 879

	if (!IS_ERR(reply))
880
		ovs_notify(&dp_flow_genl_family, reply, info);
881
	else
882
		genl_set_err(&dp_flow_genl_family, sock_net(skb->sk), 0,
883
			     0, PTR_ERR(reply));
884 885
	return 0;

886 887
err_flow_free:
	ovs_flow_free(flow, false);
888 889
err_unlock_ovs:
	ovs_unlock();
890 891
err_kfree:
	kfree(acts);
892 893 894 895 896 897 898 899 900 901 902 903
error:
	return error;
}

static int ovs_flow_cmd_get(struct sk_buff *skb, struct genl_info *info)
{
	struct nlattr **a = info->attrs;
	struct ovs_header *ovs_header = info->userhdr;
	struct sw_flow_key key;
	struct sk_buff *reply;
	struct sw_flow *flow;
	struct datapath *dp;
904
	struct sw_flow_match match;
905 906
	int err;

907 908
	if (!a[OVS_FLOW_ATTR_KEY]) {
		OVS_NLERR("Flow get message rejected, Key attribute missing.\n");
909
		return -EINVAL;
910 911 912
	}

	ovs_match_init(&match, &key, NULL);
913
	err = ovs_nla_get_match(&match, NULL, a[OVS_FLOW_ATTR_KEY], NULL);
914 915 916
	if (err)
		return err;

917
	ovs_lock();
918
	dp = get_dp(sock_net(skb->sk), ovs_header->dp_ifindex);
919 920 921 922
	if (!dp) {
		err = -ENODEV;
		goto unlock;
	}
923

924
	flow = ovs_flow_tbl_lookup(&dp->table, &key);
925
	if (!flow || !ovs_flow_cmp_unmasked_key(flow, &match)) {
926 927 928
		err = -ENOENT;
		goto unlock;
	}
929

930
	reply = ovs_flow_cmd_build_info(flow, dp, info, OVS_FLOW_CMD_NEW);
931 932 933 934
	if (IS_ERR(reply)) {
		err = PTR_ERR(reply);
		goto unlock;
	}
935

936
	ovs_unlock();
937
	return genlmsg_reply(reply, info);
938 939 940
unlock:
	ovs_unlock();
	return err;
941 942 943 944 945 946 947 948 949 950
}

static int ovs_flow_cmd_del(struct sk_buff *skb, struct genl_info *info)
{
	struct nlattr **a = info->attrs;
	struct ovs_header *ovs_header = info->userhdr;
	struct sw_flow_key key;
	struct sk_buff *reply;
	struct sw_flow *flow;
	struct datapath *dp;
951
	struct sw_flow_match match;
952 953
	int err;

954
	ovs_lock();
955
	dp = get_dp(sock_net(skb->sk), ovs_header->dp_ifindex);
956 957 958 959
	if (!dp) {
		err = -ENODEV;
		goto unlock;
	}
960

961
	if (!a[OVS_FLOW_ATTR_KEY]) {
962
		err = ovs_flow_tbl_flush(&dp->table);
963 964
		goto unlock;
	}
965 966

	ovs_match_init(&match, &key, NULL);
967
	err = ovs_nla_get_match(&match, NULL, a[OVS_FLOW_ATTR_KEY], NULL);
968
	if (err)
969
		goto unlock;
970

971
	flow = ovs_flow_tbl_lookup(&dp->table, &key);
972
	if (!flow || !ovs_flow_cmp_unmasked_key(flow, &match)) {
973 974 975
		err = -ENOENT;
		goto unlock;
	}
976

977
	reply = ovs_flow_cmd_alloc_info(flow, info);
978 979 980 981
	if (!reply) {
		err = -ENOMEM;
		goto unlock;
	}
982

983
	ovs_flow_tbl_remove(&dp->table, flow);
984

985
	err = ovs_flow_cmd_fill_info(flow, dp, reply, info->snd_portid,
986 987 988
				     info->snd_seq, 0, OVS_FLOW_CMD_DEL);
	BUG_ON(err < 0);

989
	ovs_flow_free(flow, true);
990
	ovs_unlock();
991

992
	ovs_notify(&dp_flow_genl_family, reply, info);
993
	return 0;
994 995 996
unlock:
	ovs_unlock();
	return err;
997 998 999 1000 1001
}

static int ovs_flow_cmd_dump(struct sk_buff *skb, struct netlink_callback *cb)
{
	struct ovs_header *ovs_header = genlmsg_data(nlmsg_data(cb->nlh));
1002
	struct table_instance *ti;
1003 1004
	struct datapath *dp;

1005
	rcu_read_lock();
1006
	dp = get_dp(sock_net(skb->sk), ovs_header->dp_ifindex);
1007
	if (!dp) {
1008
		rcu_read_unlock();
1009
		return -ENODEV;
1010
	}
1011

1012
	ti = rcu_dereference(dp->table.ti);
1013 1014 1015 1016 1017 1018
	for (;;) {
		struct sw_flow *flow;
		u32 bucket, obj;

		bucket = cb->args[0];
		obj = cb->args[1];
1019
		flow = ovs_flow_tbl_dump_next(ti, &bucket, &obj);
1020 1021 1022 1023
		if (!flow)
			break;

		if (ovs_flow_cmd_fill_info(flow, dp, skb,
1024
					   NETLINK_CB(cb->skb).portid,
1025 1026 1027 1028 1029 1030 1031
					   cb->nlh->nlmsg_seq, NLM_F_MULTI,
					   OVS_FLOW_CMD_NEW) < 0)
			break;

		cb->args[0] = bucket;
		cb->args[1] = obj;
	}
1032
	rcu_read_unlock();
1033 1034 1035
	return skb->len;
}

1036
static const struct genl_ops dp_flow_genl_ops[] = {
1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062
	{ .cmd = OVS_FLOW_CMD_NEW,
	  .flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
	  .policy = flow_policy,
	  .doit = ovs_flow_cmd_new_or_set
	},
	{ .cmd = OVS_FLOW_CMD_DEL,
	  .flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
	  .policy = flow_policy,
	  .doit = ovs_flow_cmd_del
	},
	{ .cmd = OVS_FLOW_CMD_GET,
	  .flags = 0,		    /* OK for unprivileged users. */
	  .policy = flow_policy,
	  .doit = ovs_flow_cmd_get,
	  .dumpit = ovs_flow_cmd_dump
	},
	{ .cmd = OVS_FLOW_CMD_SET,
	  .flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
	  .policy = flow_policy,
	  .doit = ovs_flow_cmd_new_or_set,
	},
};

static const struct nla_policy datapath_policy[OVS_DP_ATTR_MAX + 1] = {
	[OVS_DP_ATTR_NAME] = { .type = NLA_NUL_STRING, .len = IFNAMSIZ - 1 },
	[OVS_DP_ATTR_UPCALL_PID] = { .type = NLA_U32 },
1063
	[OVS_DP_ATTR_USER_FEATURES] = { .type = NLA_U32 },
1064 1065 1066 1067 1068 1069 1070
};

static struct genl_family dp_datapath_genl_family = {
	.id = GENL_ID_GENERATE,
	.hdrsize = sizeof(struct ovs_header),
	.name = OVS_DATAPATH_FAMILY,
	.version = OVS_DATAPATH_VERSION,
1071
	.maxattr = OVS_DP_ATTR_MAX,
1072 1073
	.netnsok = true,
	.parallel_ops = true,
1074 1075 1076 1077 1078 1079
};

static struct genl_multicast_group ovs_dp_datapath_multicast_group = {
	.name = OVS_DATAPATH_MCGROUP
};

1080 1081 1082 1083 1084 1085
static size_t ovs_dp_cmd_msg_size(void)
{
	size_t msgsize = NLMSG_ALIGN(sizeof(struct ovs_header));

	msgsize += nla_total_size(IFNAMSIZ);
	msgsize += nla_total_size(sizeof(struct ovs_dp_stats));
1086
	msgsize += nla_total_size(sizeof(struct ovs_dp_megaflow_stats));
1087
	msgsize += nla_total_size(sizeof(u32)); /* OVS_DP_ATTR_USER_FEATURES */
1088 1089 1090 1091

	return msgsize;
}

1092
static int ovs_dp_cmd_fill_info(struct datapath *dp, struct sk_buff *skb,
1093
				u32 portid, u32 seq, u32 flags, u8 cmd)
1094 1095 1096
{
	struct ovs_header *ovs_header;
	struct ovs_dp_stats dp_stats;
1097
	struct ovs_dp_megaflow_stats dp_megaflow_stats;
1098 1099
	int err;

1100
	ovs_header = genlmsg_put(skb, portid, seq, &dp_datapath_genl_family,
1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112
				   flags, cmd);
	if (!ovs_header)
		goto error;

	ovs_header->dp_ifindex = get_dpifindex(dp);

	rcu_read_lock();
	err = nla_put_string(skb, OVS_DP_ATTR_NAME, ovs_dp_name(dp));
	rcu_read_unlock();
	if (err)
		goto nla_put_failure;

1113 1114 1115 1116 1117 1118 1119 1120
	get_dp_stats(dp, &dp_stats, &dp_megaflow_stats);
	if (nla_put(skb, OVS_DP_ATTR_STATS, sizeof(struct ovs_dp_stats),
			&dp_stats))
		goto nla_put_failure;

	if (nla_put(skb, OVS_DP_ATTR_MEGAFLOW_STATS,
			sizeof(struct ovs_dp_megaflow_stats),
			&dp_megaflow_stats))
1121
		goto nla_put_failure;
1122

1123 1124 1125
	if (nla_put_u32(skb, OVS_DP_ATTR_USER_FEATURES, dp->user_features))
		goto nla_put_failure;

1126 1127 1128 1129 1130 1131 1132 1133
	return genlmsg_end(skb, ovs_header);

nla_put_failure:
	genlmsg_cancel(skb, ovs_header);
error:
	return -EMSGSIZE;
}

1134 1135
static struct sk_buff *ovs_dp_cmd_build_info(struct datapath *dp,
					     struct genl_info *info, u8 cmd)
1136 1137 1138 1139
{
	struct sk_buff *skb;
	int retval;

1140
	skb = genlmsg_new_unicast(ovs_dp_cmd_msg_size(), info, GFP_KERNEL);
1141 1142 1143
	if (!skb)
		return ERR_PTR(-ENOMEM);

1144
	retval = ovs_dp_cmd_fill_info(dp, skb, info->snd_portid, info->snd_seq, 0, cmd);
1145 1146 1147 1148 1149 1150 1151
	if (retval < 0) {
		kfree_skb(skb);
		return ERR_PTR(retval);
	}
	return skb;
}

1152
/* Called with ovs_mutex. */
1153 1154
static struct datapath *lookup_datapath(struct net *net,
					struct ovs_header *ovs_header,
1155 1156 1157 1158 1159
					struct nlattr *a[OVS_DP_ATTR_MAX + 1])
{
	struct datapath *dp;

	if (!a[OVS_DP_ATTR_NAME])
1160
		dp = get_dp(net, ovs_header->dp_ifindex);
1161 1162 1163 1164
	else {
		struct vport *vport;

		rcu_read_lock();
1165
		vport = ovs_vport_locate(net, nla_data(a[OVS_DP_ATTR_NAME]));
1166 1167 1168 1169 1170 1171
		dp = vport && vport->port_no == OVSP_LOCAL ? vport->dp : NULL;
		rcu_read_unlock();
	}
	return dp ? dp : ERR_PTR(-ENODEV);
}

1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183
static void ovs_dp_reset_user_features(struct sk_buff *skb, struct genl_info *info)
{
	struct datapath *dp;

	dp = lookup_datapath(sock_net(skb->sk), info->userhdr, info->attrs);
	if (!dp)
		return;

	WARN(dp->user_features, "Dropping previously announced user features\n");
	dp->user_features = 0;
}

1184 1185 1186 1187 1188 1189
static void ovs_dp_change(struct datapath *dp, struct nlattr **a)
{
	if (a[OVS_DP_ATTR_USER_FEATURES])
		dp->user_features = nla_get_u32(a[OVS_DP_ATTR_USER_FEATURES]);
}

1190 1191 1192 1193 1194 1195 1196
static int ovs_dp_cmd_new(struct sk_buff *skb, struct genl_info *info)
{
	struct nlattr **a = info->attrs;
	struct vport_parms parms;
	struct sk_buff *reply;
	struct datapath *dp;
	struct vport *vport;
1197
	struct ovs_net *ovs_net;
1198
	int err, i;
1199 1200 1201 1202 1203

	err = -EINVAL;
	if (!a[OVS_DP_ATTR_NAME] || !a[OVS_DP_ATTR_UPCALL_PID])
		goto err;

1204
	ovs_lock();
1205 1206 1207 1208

	err = -ENOMEM;
	dp = kzalloc(sizeof(*dp), GFP_KERNEL);
	if (dp == NULL)
1209
		goto err_unlock_ovs;
1210 1211

	ovs_dp_set_net(dp, hold_net(sock_net(skb->sk)));
1212 1213

	/* Allocate table. */
1214 1215
	err = ovs_flow_tbl_init(&dp->table);
	if (err)
1216 1217 1218 1219 1220 1221 1222 1223
		goto err_free_dp;

	dp->stats_percpu = alloc_percpu(struct dp_stats_percpu);
	if (!dp->stats_percpu) {
		err = -ENOMEM;
		goto err_destroy_table;
	}

1224 1225 1226 1227 1228 1229
	for_each_possible_cpu(i) {
		struct dp_stats_percpu *dpath_stats;
		dpath_stats = per_cpu_ptr(dp->stats_percpu, i);
		u64_stats_init(&dpath_stats->sync);
	}

1230
	dp->ports = kmalloc(DP_VPORT_HASH_BUCKETS * sizeof(struct hlist_head),
1231
			    GFP_KERNEL);
1232 1233 1234 1235 1236 1237 1238 1239
	if (!dp->ports) {
		err = -ENOMEM;
		goto err_destroy_percpu;
	}

	for (i = 0; i < DP_VPORT_HASH_BUCKETS; i++)
		INIT_HLIST_HEAD(&dp->ports[i]);

1240 1241 1242 1243 1244 1245
	/* Set up our datapath device. */
	parms.name = nla_data(a[OVS_DP_ATTR_NAME]);
	parms.type = OVS_VPORT_TYPE_INTERNAL;
	parms.options = NULL;
	parms.dp = dp;
	parms.port_no = OVSP_LOCAL;
1246
	parms.upcall_portid = nla_get_u32(a[OVS_DP_ATTR_UPCALL_PID]);
1247

1248 1249
	ovs_dp_change(dp, a);

1250 1251 1252 1253 1254 1255
	vport = new_vport(&parms);
	if (IS_ERR(vport)) {
		err = PTR_ERR(vport);
		if (err == -EBUSY)
			err = -EEXIST;

1256 1257 1258 1259 1260 1261 1262 1263 1264
		if (err == -EEXIST) {
			/* An outdated user space instance that does not understand
			 * the concept of user_features has attempted to create a new
			 * datapath and is likely to reuse it. Drop all user features.
			 */
			if (info->genlhdr->version < OVS_DP_VER_FEATURES)
				ovs_dp_reset_user_features(skb, info);
		}

1265
		goto err_destroy_ports_array;
1266 1267
	}

1268
	reply = ovs_dp_cmd_build_info(dp, info, OVS_DP_CMD_NEW);
1269 1270 1271 1272
	err = PTR_ERR(reply);
	if (IS_ERR(reply))
		goto err_destroy_local_port;

1273
	ovs_net = net_generic(ovs_dp_get_net(dp), ovs_net_id);
1274
	list_add_tail_rcu(&dp->list_node, &ovs_net->dps);
1275 1276

	ovs_unlock();
1277

1278
	ovs_notify(&dp_datapath_genl_family, reply, info);
1279 1280 1281
	return 0;

err_destroy_local_port:
1282
	ovs_dp_detach_port(ovs_vport_ovsl(dp, OVSP_LOCAL));
1283 1284
err_destroy_ports_array:
	kfree(dp->ports);
1285 1286 1287
err_destroy_percpu:
	free_percpu(dp->stats_percpu);
err_destroy_table:
1288
	ovs_flow_tbl_destroy(&dp->table, false);
1289
err_free_dp:
1290
	release_net(ovs_dp_get_net(dp));
1291
	kfree(dp);
1292 1293
err_unlock_ovs:
	ovs_unlock();
1294 1295 1296 1297
err:
	return err;
}

1298
/* Called with ovs_mutex. */
1299
static void __dp_destroy(struct datapath *dp)
1300
{
1301
	int i;
1302

1303 1304
	for (i = 0; i < DP_VPORT_HASH_BUCKETS; i++) {
		struct vport *vport;
1305
		struct hlist_node *n;
1306

1307
		hlist_for_each_entry_safe(vport, n, &dp->ports[i], dp_hash_node)
1308 1309 1310
			if (vport->port_no != OVSP_LOCAL)
				ovs_dp_detach_port(vport);
	}
1311

1312
	list_del_rcu(&dp->list_node);
1313

1314
	/* OVSP_LOCAL is datapath internal port. We need to make sure that
1315
	 * all ports in datapath are destroyed first before freeing datapath.
1316
	 */
1317
	ovs_dp_detach_port(ovs_vport_ovsl(dp, OVSP_LOCAL));
1318

1319 1320 1321
	/* RCU destroy the flow table */
	ovs_flow_tbl_destroy(&dp->table, true);

1322
	call_rcu(&dp->rcu, destroy_dp_rcu);
1323 1324 1325 1326 1327 1328 1329 1330
}

static int ovs_dp_cmd_del(struct sk_buff *skb, struct genl_info *info)
{
	struct sk_buff *reply;
	struct datapath *dp;
	int err;

1331
	ovs_lock();
1332 1333 1334
	dp = lookup_datapath(sock_net(skb->sk), info->userhdr, info->attrs);
	err = PTR_ERR(dp);
	if (IS_ERR(dp))
1335
		goto unlock;
1336

1337
	reply = ovs_dp_cmd_build_info(dp, info, OVS_DP_CMD_DEL);
1338 1339
	err = PTR_ERR(reply);
	if (IS_ERR(reply))
1340
		goto unlock;
1341 1342

	__dp_destroy(dp);
1343
	ovs_unlock();
1344

1345
	ovs_notify(&dp_datapath_genl_family, reply, info);
1346 1347

	return 0;
1348 1349 1350
unlock:
	ovs_unlock();
	return err;
1351 1352 1353 1354 1355 1356 1357 1358
}

static int ovs_dp_cmd_set(struct sk_buff *skb, struct genl_info *info)
{
	struct sk_buff *reply;
	struct datapath *dp;
	int err;

1359
	ovs_lock();
1360
	dp = lookup_datapath(sock_net(skb->sk), info->userhdr, info->attrs);
1361
	err = PTR_ERR(dp);
1362
	if (IS_ERR(dp))
1363
		goto unlock;
1364

1365 1366
	ovs_dp_change(dp, info->attrs);

1367
	reply = ovs_dp_cmd_build_info(dp, info, OVS_DP_CMD_NEW);
1368 1369
	if (IS_ERR(reply)) {
		err = PTR_ERR(reply);
1370
		genl_set_err(&dp_datapath_genl_family, sock_net(skb->sk), 0,
1371
			     0, err);
1372 1373
		err = 0;
		goto unlock;
1374 1375
	}

1376
	ovs_unlock();
1377
	ovs_notify(&dp_datapath_genl_family, reply, info);
1378 1379

	return 0;
1380 1381 1382
unlock:
	ovs_unlock();
	return err;
1383 1384 1385 1386 1387 1388
}

static int ovs_dp_cmd_get(struct sk_buff *skb, struct genl_info *info)
{
	struct sk_buff *reply;
	struct datapath *dp;
1389
	int err;
1390

1391
	ovs_lock();
1392
	dp = lookup_datapath(sock_net(skb->sk), info->userhdr, info->attrs);
1393 1394 1395 1396
	if (IS_ERR(dp)) {
		err = PTR_ERR(dp);
		goto unlock;
	}
1397

1398
	reply = ovs_dp_cmd_build_info(dp, info, OVS_DP_CMD_NEW);
1399 1400 1401 1402
	if (IS_ERR(reply)) {
		err = PTR_ERR(reply);
		goto unlock;
	}
1403

1404
	ovs_unlock();
1405
	return genlmsg_reply(reply, info);
1406 1407 1408 1409

unlock:
	ovs_unlock();
	return err;
1410 1411 1412 1413
}

static int ovs_dp_cmd_dump(struct sk_buff *skb, struct netlink_callback *cb)
{
1414
	struct ovs_net *ovs_net = net_generic(sock_net(skb->sk), ovs_net_id);
1415 1416 1417 1418
	struct datapath *dp;
	int skip = cb->args[0];
	int i = 0;

1419 1420
	rcu_read_lock();
	list_for_each_entry_rcu(dp, &ovs_net->dps, list_node) {
1421
		if (i >= skip &&
1422
		    ovs_dp_cmd_fill_info(dp, skb, NETLINK_CB(cb->skb).portid,
1423 1424 1425 1426 1427
					 cb->nlh->nlmsg_seq, NLM_F_MULTI,
					 OVS_DP_CMD_NEW) < 0)
			break;
		i++;
	}
1428
	rcu_read_unlock();
1429 1430 1431 1432 1433 1434

	cb->args[0] = i;

	return skb->len;
}

1435
static const struct genl_ops dp_datapath_genl_ops[] = {
1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467
	{ .cmd = OVS_DP_CMD_NEW,
	  .flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
	  .policy = datapath_policy,
	  .doit = ovs_dp_cmd_new
	},
	{ .cmd = OVS_DP_CMD_DEL,
	  .flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
	  .policy = datapath_policy,
	  .doit = ovs_dp_cmd_del
	},
	{ .cmd = OVS_DP_CMD_GET,
	  .flags = 0,		    /* OK for unprivileged users. */
	  .policy = datapath_policy,
	  .doit = ovs_dp_cmd_get,
	  .dumpit = ovs_dp_cmd_dump
	},
	{ .cmd = OVS_DP_CMD_SET,
	  .flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
	  .policy = datapath_policy,
	  .doit = ovs_dp_cmd_set,
	},
};

static const struct nla_policy vport_policy[OVS_VPORT_ATTR_MAX + 1] = {
	[OVS_VPORT_ATTR_NAME] = { .type = NLA_NUL_STRING, .len = IFNAMSIZ - 1 },
	[OVS_VPORT_ATTR_STATS] = { .len = sizeof(struct ovs_vport_stats) },
	[OVS_VPORT_ATTR_PORT_NO] = { .type = NLA_U32 },
	[OVS_VPORT_ATTR_TYPE] = { .type = NLA_U32 },
	[OVS_VPORT_ATTR_UPCALL_PID] = { .type = NLA_U32 },
	[OVS_VPORT_ATTR_OPTIONS] = { .type = NLA_NESTED },
};

1468
struct genl_family dp_vport_genl_family = {
1469 1470 1471 1472
	.id = GENL_ID_GENERATE,
	.hdrsize = sizeof(struct ovs_header),
	.name = OVS_VPORT_FAMILY,
	.version = OVS_VPORT_VERSION,
1473
	.maxattr = OVS_VPORT_ATTR_MAX,
1474 1475
	.netnsok = true,
	.parallel_ops = true,
1476 1477
};

S
Stephen Hemminger 已提交
1478
static struct genl_multicast_group ovs_dp_vport_multicast_group = {
1479 1480 1481
	.name = OVS_VPORT_MCGROUP
};

1482
/* Called with ovs_mutex or RCU read lock. */
1483
static int ovs_vport_cmd_fill_info(struct vport *vport, struct sk_buff *skb,
1484
				   u32 portid, u32 seq, u32 flags, u8 cmd)
1485 1486 1487 1488 1489
{
	struct ovs_header *ovs_header;
	struct ovs_vport_stats vport_stats;
	int err;

1490
	ovs_header = genlmsg_put(skb, portid, seq, &dp_vport_genl_family,
1491 1492 1493 1494 1495 1496
				 flags, cmd);
	if (!ovs_header)
		return -EMSGSIZE;

	ovs_header->dp_ifindex = get_dpifindex(vport->dp);

1497 1498 1499
	if (nla_put_u32(skb, OVS_VPORT_ATTR_PORT_NO, vport->port_no) ||
	    nla_put_u32(skb, OVS_VPORT_ATTR_TYPE, vport->ops->type) ||
	    nla_put_string(skb, OVS_VPORT_ATTR_NAME, vport->ops->get_name(vport)) ||
1500
	    nla_put_u32(skb, OVS_VPORT_ATTR_UPCALL_PID, vport->upcall_portid))
1501
		goto nla_put_failure;
1502 1503

	ovs_vport_get_stats(vport, &vport_stats);
1504 1505 1506
	if (nla_put(skb, OVS_VPORT_ATTR_STATS, sizeof(struct ovs_vport_stats),
		    &vport_stats))
		goto nla_put_failure;
1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520

	err = ovs_vport_get_options(vport, skb);
	if (err == -EMSGSIZE)
		goto error;

	return genlmsg_end(skb, ovs_header);

nla_put_failure:
	err = -EMSGSIZE;
error:
	genlmsg_cancel(skb, ovs_header);
	return err;
}

1521
/* Called with ovs_mutex or RCU read lock. */
1522
struct sk_buff *ovs_vport_cmd_build_info(struct vport *vport, u32 portid,
1523 1524 1525 1526 1527 1528 1529 1530 1531
					 u32 seq, u8 cmd)
{
	struct sk_buff *skb;
	int retval;

	skb = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_ATOMIC);
	if (!skb)
		return ERR_PTR(-ENOMEM);

1532
	retval = ovs_vport_cmd_fill_info(vport, skb, portid, seq, 0, cmd);
1533 1534
	BUG_ON(retval < 0);

1535 1536 1537
	return skb;
}

1538
/* Called with ovs_mutex or RCU read lock. */
1539 1540
static struct vport *lookup_vport(struct net *net,
				  struct ovs_header *ovs_header,
1541 1542 1543 1544 1545 1546
				  struct nlattr *a[OVS_VPORT_ATTR_MAX + 1])
{
	struct datapath *dp;
	struct vport *vport;

	if (a[OVS_VPORT_ATTR_NAME]) {
1547
		vport = ovs_vport_locate(net, nla_data(a[OVS_VPORT_ATTR_NAME]));
1548 1549
		if (!vport)
			return ERR_PTR(-ENODEV);
1550 1551 1552
		if (ovs_header->dp_ifindex &&
		    ovs_header->dp_ifindex != get_dpifindex(vport->dp))
			return ERR_PTR(-ENODEV);
1553 1554 1555 1556 1557 1558 1559
		return vport;
	} else if (a[OVS_VPORT_ATTR_PORT_NO]) {
		u32 port_no = nla_get_u32(a[OVS_VPORT_ATTR_PORT_NO]);

		if (port_no >= DP_MAX_PORTS)
			return ERR_PTR(-EFBIG);

1560
		dp = get_dp(net, ovs_header->dp_ifindex);
1561 1562 1563
		if (!dp)
			return ERR_PTR(-ENODEV);

1564
		vport = ovs_vport_ovsl_rcu(dp, port_no);
1565
		if (!vport)
1566
			return ERR_PTR(-ENODEV);
1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587
		return vport;
	} else
		return ERR_PTR(-EINVAL);
}

static int ovs_vport_cmd_new(struct sk_buff *skb, struct genl_info *info)
{
	struct nlattr **a = info->attrs;
	struct ovs_header *ovs_header = info->userhdr;
	struct vport_parms parms;
	struct sk_buff *reply;
	struct vport *vport;
	struct datapath *dp;
	u32 port_no;
	int err;

	err = -EINVAL;
	if (!a[OVS_VPORT_ATTR_NAME] || !a[OVS_VPORT_ATTR_TYPE] ||
	    !a[OVS_VPORT_ATTR_UPCALL_PID])
		goto exit;

1588
	ovs_lock();
1589
	dp = get_dp(sock_net(skb->sk), ovs_header->dp_ifindex);
1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600
	err = -ENODEV;
	if (!dp)
		goto exit_unlock;

	if (a[OVS_VPORT_ATTR_PORT_NO]) {
		port_no = nla_get_u32(a[OVS_VPORT_ATTR_PORT_NO]);

		err = -EFBIG;
		if (port_no >= DP_MAX_PORTS)
			goto exit_unlock;

1601
		vport = ovs_vport_ovsl(dp, port_no);
1602 1603 1604 1605 1606 1607 1608 1609 1610
		err = -EBUSY;
		if (vport)
			goto exit_unlock;
	} else {
		for (port_no = 1; ; port_no++) {
			if (port_no >= DP_MAX_PORTS) {
				err = -EFBIG;
				goto exit_unlock;
			}
1611
			vport = ovs_vport_ovsl(dp, port_no);
1612 1613 1614 1615 1616 1617 1618 1619 1620 1621
			if (!vport)
				break;
		}
	}

	parms.name = nla_data(a[OVS_VPORT_ATTR_NAME]);
	parms.type = nla_get_u32(a[OVS_VPORT_ATTR_TYPE]);
	parms.options = a[OVS_VPORT_ATTR_OPTIONS];
	parms.dp = dp;
	parms.port_no = port_no;
1622
	parms.upcall_portid = nla_get_u32(a[OVS_VPORT_ATTR_UPCALL_PID]);
1623 1624 1625 1626 1627 1628

	vport = new_vport(&parms);
	err = PTR_ERR(vport);
	if (IS_ERR(vport))
		goto exit_unlock;

1629
	err = 0;
1630
	reply = ovs_vport_cmd_build_info(vport, info->snd_portid, info->snd_seq,
1631 1632 1633 1634 1635 1636
					 OVS_VPORT_CMD_NEW);
	if (IS_ERR(reply)) {
		err = PTR_ERR(reply);
		ovs_dp_detach_port(vport);
		goto exit_unlock;
	}
1637

1638
	ovs_notify(&dp_vport_genl_family, reply, info);
1639 1640

exit_unlock:
1641
	ovs_unlock();
1642 1643 1644 1645 1646 1647 1648 1649 1650 1651 1652
exit:
	return err;
}

static int ovs_vport_cmd_set(struct sk_buff *skb, struct genl_info *info)
{
	struct nlattr **a = info->attrs;
	struct sk_buff *reply;
	struct vport *vport;
	int err;

1653
	ovs_lock();
1654
	vport = lookup_vport(sock_net(skb->sk), info->userhdr, a);
1655 1656 1657 1658 1659
	err = PTR_ERR(vport);
	if (IS_ERR(vport))
		goto exit_unlock;

	if (a[OVS_VPORT_ATTR_TYPE] &&
1660
	    nla_get_u32(a[OVS_VPORT_ATTR_TYPE]) != vport->ops->type) {
1661
		err = -EINVAL;
1662 1663
		goto exit_unlock;
	}
1664

1665 1666 1667 1668 1669 1670
	reply = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
	if (!reply) {
		err = -ENOMEM;
		goto exit_unlock;
	}

1671
	if (a[OVS_VPORT_ATTR_OPTIONS]) {
1672
		err = ovs_vport_set_options(vport, a[OVS_VPORT_ATTR_OPTIONS]);
1673 1674 1675
		if (err)
			goto exit_free;
	}
1676

1677
	if (a[OVS_VPORT_ATTR_UPCALL_PID])
1678
		vport->upcall_portid = nla_get_u32(a[OVS_VPORT_ATTR_UPCALL_PID]);
1679

1680 1681 1682
	err = ovs_vport_cmd_fill_info(vport, reply, info->snd_portid,
				      info->snd_seq, 0, OVS_VPORT_CMD_NEW);
	BUG_ON(err < 0);
1683

1684
	ovs_unlock();
1685
	ovs_notify(&dp_vport_genl_family, reply, info);
1686
	return 0;
1687

1688 1689
exit_free:
	kfree_skb(reply);
1690
exit_unlock:
1691
	ovs_unlock();
1692 1693 1694 1695 1696 1697 1698 1699 1700 1701
	return err;
}

static int ovs_vport_cmd_del(struct sk_buff *skb, struct genl_info *info)
{
	struct nlattr **a = info->attrs;
	struct sk_buff *reply;
	struct vport *vport;
	int err;

1702
	ovs_lock();
1703
	vport = lookup_vport(sock_net(skb->sk), info->userhdr, a);
1704 1705 1706 1707 1708 1709 1710 1711 1712
	err = PTR_ERR(vport);
	if (IS_ERR(vport))
		goto exit_unlock;

	if (vport->port_no == OVSP_LOCAL) {
		err = -EINVAL;
		goto exit_unlock;
	}

1713 1714
	reply = ovs_vport_cmd_build_info(vport, info->snd_portid,
					 info->snd_seq, OVS_VPORT_CMD_DEL);
1715 1716 1717 1718
	err = PTR_ERR(reply);
	if (IS_ERR(reply))
		goto exit_unlock;

1719
	err = 0;
1720 1721
	ovs_dp_detach_port(vport);

1722
	ovs_notify(&dp_vport_genl_family, reply, info);
1723 1724

exit_unlock:
1725
	ovs_unlock();
1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737
	return err;
}

static int ovs_vport_cmd_get(struct sk_buff *skb, struct genl_info *info)
{
	struct nlattr **a = info->attrs;
	struct ovs_header *ovs_header = info->userhdr;
	struct sk_buff *reply;
	struct vport *vport;
	int err;

	rcu_read_lock();
1738
	vport = lookup_vport(sock_net(skb->sk), ovs_header, a);
1739 1740 1741 1742
	err = PTR_ERR(vport);
	if (IS_ERR(vport))
		goto exit_unlock;

1743 1744
	reply = ovs_vport_cmd_build_info(vport, info->snd_portid,
					 info->snd_seq, OVS_VPORT_CMD_NEW);
1745 1746 1747 1748 1749 1750 1751 1752 1753 1754 1755 1756 1757 1758 1759 1760 1761
	err = PTR_ERR(reply);
	if (IS_ERR(reply))
		goto exit_unlock;

	rcu_read_unlock();

	return genlmsg_reply(reply, info);

exit_unlock:
	rcu_read_unlock();
	return err;
}

static int ovs_vport_cmd_dump(struct sk_buff *skb, struct netlink_callback *cb)
{
	struct ovs_header *ovs_header = genlmsg_data(nlmsg_data(cb->nlh));
	struct datapath *dp;
1762 1763
	int bucket = cb->args[0], skip = cb->args[1];
	int i, j = 0;
1764

1765
	dp = get_dp(sock_net(skb->sk), ovs_header->dp_ifindex);
1766 1767 1768 1769
	if (!dp)
		return -ENODEV;

	rcu_read_lock();
1770
	for (i = bucket; i < DP_VPORT_HASH_BUCKETS; i++) {
1771
		struct vport *vport;
1772 1773

		j = 0;
1774
		hlist_for_each_entry_rcu(vport, &dp->ports[i], dp_hash_node) {
1775 1776
			if (j >= skip &&
			    ovs_vport_cmd_fill_info(vport, skb,
1777
						    NETLINK_CB(cb->skb).portid,
1778 1779 1780 1781 1782 1783 1784 1785
						    cb->nlh->nlmsg_seq,
						    NLM_F_MULTI,
						    OVS_VPORT_CMD_NEW) < 0)
				goto out;

			j++;
		}
		skip = 0;
1786
	}
1787
out:
1788 1789
	rcu_read_unlock();

1790 1791
	cb->args[0] = i;
	cb->args[1] = j;
1792

1793
	return skb->len;
1794 1795
}

1796
static const struct genl_ops dp_vport_genl_ops[] = {
1797 1798 1799 1800 1801 1802 1803 1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821
	{ .cmd = OVS_VPORT_CMD_NEW,
	  .flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
	  .policy = vport_policy,
	  .doit = ovs_vport_cmd_new
	},
	{ .cmd = OVS_VPORT_CMD_DEL,
	  .flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
	  .policy = vport_policy,
	  .doit = ovs_vport_cmd_del
	},
	{ .cmd = OVS_VPORT_CMD_GET,
	  .flags = 0,		    /* OK for unprivileged users. */
	  .policy = vport_policy,
	  .doit = ovs_vport_cmd_get,
	  .dumpit = ovs_vport_cmd_dump
	},
	{ .cmd = OVS_VPORT_CMD_SET,
	  .flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
	  .policy = vport_policy,
	  .doit = ovs_vport_cmd_set,
	},
};

struct genl_family_and_ops {
	struct genl_family *family;
1822
	const struct genl_ops *ops;
1823
	int n_ops;
1824
	const struct genl_multicast_group *group;
1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859
};

static const struct genl_family_and_ops dp_genl_families[] = {
	{ &dp_datapath_genl_family,
	  dp_datapath_genl_ops, ARRAY_SIZE(dp_datapath_genl_ops),
	  &ovs_dp_datapath_multicast_group },
	{ &dp_vport_genl_family,
	  dp_vport_genl_ops, ARRAY_SIZE(dp_vport_genl_ops),
	  &ovs_dp_vport_multicast_group },
	{ &dp_flow_genl_family,
	  dp_flow_genl_ops, ARRAY_SIZE(dp_flow_genl_ops),
	  &ovs_dp_flow_multicast_group },
	{ &dp_packet_genl_family,
	  dp_packet_genl_ops, ARRAY_SIZE(dp_packet_genl_ops),
	  NULL },
};

static void dp_unregister_genl(int n_families)
{
	int i;

	for (i = 0; i < n_families; i++)
		genl_unregister_family(dp_genl_families[i].family);
}

static int dp_register_genl(void)
{
	int n_registered;
	int err;
	int i;

	n_registered = 0;
	for (i = 0; i < ARRAY_SIZE(dp_genl_families); i++) {
		const struct genl_family_and_ops *f = &dp_genl_families[i];

1860 1861
		f->family->ops = f->ops;
		f->family->n_ops = f->n_ops;
1862 1863
		f->family->mcgrps = f->group;
		f->family->n_mcgrps = f->group ? 1 : 0;
1864
		err = genl_register_family(f->family);
1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876
		if (err)
			goto error;
		n_registered++;
	}

	return 0;

error:
	dp_unregister_genl(n_registered);
	return err;
}

1877 1878 1879 1880 1881
static int __net_init ovs_init_net(struct net *net)
{
	struct ovs_net *ovs_net = net_generic(net, ovs_net_id);

	INIT_LIST_HEAD(&ovs_net->dps);
1882
	INIT_WORK(&ovs_net->dp_notify_work, ovs_dp_notify_wq);
1883 1884 1885 1886 1887 1888
	return 0;
}

static void __net_exit ovs_exit_net(struct net *net)
{
	struct datapath *dp, *dp_next;
1889
	struct ovs_net *ovs_net = net_generic(net, ovs_net_id);
1890

1891
	ovs_lock();
1892 1893
	list_for_each_entry_safe(dp, dp_next, &ovs_net->dps, list_node)
		__dp_destroy(dp);
1894 1895 1896
	ovs_unlock();

	cancel_work_sync(&ovs_net->dp_notify_work);
1897 1898 1899 1900 1901 1902 1903 1904 1905
}

static struct pernet_operations ovs_net_ops = {
	.init = ovs_init_net,
	.exit = ovs_exit_net,
	.id   = &ovs_net_id,
	.size = sizeof(struct ovs_net),
};

1906 1907 1908 1909
static int __init dp_init(void)
{
	int err;

1910
	BUILD_BUG_ON(sizeof(struct ovs_skb_cb) > FIELD_SIZEOF(struct sk_buff, cb));
1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921

	pr_info("Open vSwitch switching datapath\n");

	err = ovs_flow_init();
	if (err)
		goto error;

	err = ovs_vport_init();
	if (err)
		goto error_flow_exit;

1922
	err = register_pernet_device(&ovs_net_ops);
1923 1924 1925
	if (err)
		goto error_vport_exit;

1926 1927 1928 1929
	err = register_netdevice_notifier(&ovs_dp_device_notifier);
	if (err)
		goto error_netns_exit;

1930 1931 1932 1933 1934 1935 1936 1937
	err = dp_register_genl();
	if (err < 0)
		goto error_unreg_notifier;

	return 0;

error_unreg_notifier:
	unregister_netdevice_notifier(&ovs_dp_device_notifier);
1938 1939
error_netns_exit:
	unregister_pernet_device(&ovs_net_ops);
1940 1941 1942 1943 1944 1945 1946 1947 1948 1949 1950 1951
error_vport_exit:
	ovs_vport_exit();
error_flow_exit:
	ovs_flow_exit();
error:
	return err;
}

static void dp_cleanup(void)
{
	dp_unregister_genl(ARRAY_SIZE(dp_genl_families));
	unregister_netdevice_notifier(&ovs_dp_device_notifier);
1952 1953
	unregister_pernet_device(&ovs_net_ops);
	rcu_barrier();
1954 1955 1956 1957 1958 1959 1960 1961 1962
	ovs_vport_exit();
	ovs_flow_exit();
}

module_init(dp_init);
module_exit(dp_cleanup);

MODULE_DESCRIPTION("Open vSwitch switching datapath");
MODULE_LICENSE("GPL");