datapath.c 45.9 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_netlink.h"
59
#include "vport-internal_dev.h"
60
#include "vport-netdev.h"
61

62 63
int ovs_net_id __read_mostly;

64 65
static void ovs_notify(struct genl_family *family,
		       struct sk_buff *skb, struct genl_info *info,
66 67
		       struct genl_multicast_group *grp)
{
68
	genl_notify(family, skb, genl_info_net(info), info->snd_portid,
69 70 71
		    grp->id, info->nlhdr, GFP_KERNEL);
}

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 net *, int dp_ifindex, struct sk_buff *,
113
			     const struct dp_upcall_info *);
114 115
static int queue_userspace_packet(struct net *, int dp_ifindex,
				  struct sk_buff *,
116 117
				  const struct dp_upcall_info *);

118
/* Must be called with rcu_read_lock or ovs_mutex. */
119
static struct datapath *get_dp(struct net *net, int dp_ifindex)
120 121 122 123 124
{
	struct datapath *dp = NULL;
	struct net_device *dev;

	rcu_read_lock();
125
	dev = dev_get_by_index_rcu(net, dp_ifindex);
126 127 128 129 130 131 132 133 134 135
	if (dev) {
		struct vport *vport = ovs_internal_dev_get_vport(dev);
		if (vport)
			dp = vport->dp;
	}
	rcu_read_unlock();

	return dp;
}

136
/* Must be called with rcu_read_lock or ovs_mutex. */
137 138
const char *ovs_dp_name(const struct datapath *dp)
{
139
	struct vport *vport = ovs_vport_ovsl_rcu(dp, OVSP_LOCAL);
140 141 142 143 144 145 146 147 148 149
	return vport->ops->get_name(vport);
}

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

	rcu_read_lock();

150
	local = ovs_vport_rcu(dp, OVSP_LOCAL);
151
	if (local)
152
		ifindex = netdev_vport_priv(local)->dev->ifindex;
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);

165
	ovs_flow_tbl_destroy(&dp->table);
166
	free_percpu(dp->stats_percpu);
167
	release_net(ovs_dp_get_net(dp));
168
	kfree(dp->ports);
169 170 171
	kfree(dp);
}

172 173 174 175 176 177 178 179 180 181 182 183
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);
184
	hlist_for_each_entry_rcu(vport, head, dp_hash_node) {
185 186 187 188 189 190
		if (vport->port_no == port_no)
			return vport;
	}
	return NULL;
}

191
/* Called with ovs_mutex. */
192 193 194 195 196 197 198
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;
199
		struct hlist_head *head = vport_hash_bucket(dp, vport->port_no);
200

201
		hlist_add_head_rcu(&vport->dp_hash_node, head);
202 203 204 205 206 207
	}
	return vport;
}

void ovs_dp_detach_port(struct vport *p)
{
208
	ASSERT_OVSL();
209 210

	/* First drop references to device. */
211
	hlist_del_rcu(&p->dp_hash_node);
212 213 214 215 216 217 218 219 220 221 222 223 224

	/* 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;
225
	u32 n_mask_hit;
226 227
	int error;

228
	stats = this_cpu_ptr(dp->stats_percpu);
229 230

	/* Extract flow from 'skb' into 'key'. */
231
	error = ovs_flow_extract(skb, p->port_no, &key);
232 233 234 235 236 237
	if (unlikely(error)) {
		kfree_skb(skb);
		return;
	}

	/* Look up flow. */
238
	flow = ovs_flow_tbl_lookup(&dp->table, &key, &n_mask_hit);
239 240 241 242 243 244
	if (unlikely(!flow)) {
		struct dp_upcall_info upcall;

		upcall.cmd = OVS_PACKET_CMD_MISS;
		upcall.key = &key;
		upcall.userdata = NULL;
245
		upcall.portid = p->upcall_portid;
246 247 248 249 250 251 252
		ovs_dp_upcall(dp, skb, &upcall);
		consume_skb(skb);
		stats_counter = &stats->n_missed;
		goto out;
	}

	OVS_CB(skb)->flow = flow;
253
	OVS_CB(skb)->pkt_key = &key;
254 255 256 257 258 259 260 261 262

	stats_counter = &stats->n_hit;
	ovs_flow_used(OVS_CB(skb)->flow, skb);
	ovs_execute_actions(dp, skb);

out:
	/* Update datapath statistics. */
	u64_stats_update_begin(&stats->sync);
	(*stats_counter)++;
263
	stats->n_mask_hit += n_mask_hit;
264 265 266 267 268 269 270 271
	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,
272
	.maxattr = OVS_PACKET_ATTR_MAX,
273 274
	.netnsok = true,
	.parallel_ops = true,
275 276 277
};

int ovs_dp_upcall(struct datapath *dp, struct sk_buff *skb,
278
		  const struct dp_upcall_info *upcall_info)
279 280 281 282 283
{
	struct dp_stats_percpu *stats;
	int dp_ifindex;
	int err;

284
	if (upcall_info->portid == 0) {
285 286 287 288 289 290 291 292 293 294 295
		err = -ENOTCONN;
		goto err;
	}

	dp_ifindex = get_dpifindex(dp);
	if (!dp_ifindex) {
		err = -ENODEV;
		goto err;
	}

	if (!skb_is_gso(skb))
296
		err = queue_userspace_packet(ovs_dp_get_net(dp), dp_ifindex, skb, upcall_info);
297
	else
298
		err = queue_gso_packets(ovs_dp_get_net(dp), dp_ifindex, skb, upcall_info);
299 300 301 302 303 304
	if (err)
		goto err;

	return 0;

err:
305
	stats = this_cpu_ptr(dp->stats_percpu);
306 307 308 309 310 311 312 313

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

	return err;
}

314 315
static int queue_gso_packets(struct net *net, int dp_ifindex,
			     struct sk_buff *skb,
316 317
			     const struct dp_upcall_info *upcall_info)
{
318
	unsigned short gso_type = skb_shinfo(skb)->gso_type;
319 320 321 322 323
	struct dp_upcall_info later_info;
	struct sw_flow_key later_key;
	struct sk_buff *segs, *nskb;
	int err;

324
	segs = __skb_gso_segment(skb, NETIF_F_SG | NETIF_F_HW_CSUM, false);
325 326
	if (IS_ERR(segs))
		return PTR_ERR(segs);
327 328 329 330

	/* Queue all of the segments. */
	skb = segs;
	do {
331
		err = queue_userspace_packet(net, dp_ifindex, skb, upcall_info);
332 333 334
		if (err)
			break;

335
		if (skb == segs && gso_type & SKB_GSO_UDP) {
336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360
			/* 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;
}

361 362 363
static size_t key_attr_size(void)
{
	return    nla_total_size(4)   /* OVS_KEY_ATTR_PRIORITY */
364 365 366 367 368 369 370 371
		+ 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 */
372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397
		+ 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 */
}

static size_t upcall_msg_size(const struct sk_buff *skb,
			      const struct nlattr *userdata)
{
	size_t size = NLMSG_ALIGN(sizeof(struct ovs_header))
		+ nla_total_size(skb->len) /* OVS_PACKET_ATTR_PACKET */
		+ nla_total_size(key_attr_size()); /* OVS_PACKET_ATTR_KEY */

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

	return size;
}

398 399
static int queue_userspace_packet(struct net *net, int dp_ifindex,
				  struct sk_buff *skb,
400 401 402 403 404 405 406 407 408 409 410 411 412
				  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;
	int err;

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

413
		nskb = __vlan_put_tag(nskb, nskb->vlan_proto, vlan_tx_tag_get(nskb));
414
		if (!nskb)
415 416 417 418 419 420 421 422 423 424 425
			return -ENOMEM;

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

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

426
	user_skb = genlmsg_new(upcall_msg_size(skb, upcall_info->userdata), GFP_ATOMIC);
427 428 429 430 431 432 433 434 435 436
	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);
437
	ovs_nla_put_flow(upcall_info->key, upcall_info->key, user_skb);
438 439 440
	nla_nest_end(user_skb, nla);

	if (upcall_info->userdata)
441 442 443
		__nla_put(user_skb, OVS_PACKET_ATTR_USERDATA,
			  nla_len(upcall_info->userdata),
			  nla_data(upcall_info->userdata));
444 445 446 447 448

	nla = __nla_reserve(user_skb, OVS_PACKET_ATTR_PACKET, skb->len);

	skb_copy_and_csum_dev(skb, nla_data(nla));

449
	genlmsg_end(user_skb, upcall);
450
	err = genlmsg_unicast(net, user_skb, upcall_info->portid);
451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478

out:
	kfree_skb(nskb);
	return err;
}

static void clear_stats(struct sw_flow *flow)
{
	flow->used = 0;
	flow->tcp_flags = 0;
	flow->packet_count = 0;
	flow->byte_count = 0;
}

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] ||
479
	    !a[OVS_PACKET_ATTR_ACTIONS])
480 481 482 483 484 485 486 487 488
		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);

489
	nla_memcpy(__skb_put(packet, len), a[OVS_PACKET_ATTR_PACKET], len);
490 491 492 493 494 495 496

	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. */
Simon Horman's avatar
Simon Horman committed
497
	if (ntohs(eth->h_proto) >= ETH_P_802_3_MIN)
498 499 500 501 502 503 504 505 506 507
		packet->protocol = eth->h_proto;
	else
		packet->protocol = htons(ETH_P_802_2);

	/* Build an sw_flow for sending this packet. */
	flow = ovs_flow_alloc();
	err = PTR_ERR(flow);
	if (IS_ERR(flow))
		goto err_kfree_skb;

508
	err = ovs_flow_extract(packet, -1, &flow->key);
509 510 511
	if (err)
		goto err_flow_free;

512
	err = ovs_nla_get_flow_metadata(flow, a[OVS_PACKET_ATTR_KEY]);
513 514
	if (err)
		goto err_flow_free;
515
	acts = ovs_nla_alloc_flow_actions(nla_len(a[OVS_PACKET_ATTR_ACTIONS]));
516 517 518
	err = PTR_ERR(acts);
	if (IS_ERR(acts))
		goto err_flow_free;
519

520 521
	err = ovs_nla_copy_actions(a[OVS_PACKET_ATTR_ACTIONS],
				   &flow->key, 0, &acts);
522
	rcu_assign_pointer(flow->sf_acts, acts);
523 524
	if (err)
		goto err_flow_free;
525 526

	OVS_CB(packet)->flow = flow;
527
	OVS_CB(packet)->pkt_key = &flow->key;
528
	packet->priority = flow->key.phy.priority;
529
	packet->mark = flow->key.phy.skb_mark;
530 531

	rcu_read_lock();
532
	dp = get_dp(sock_net(skb->sk), ovs_header->dp_ifindex);
533 534 535 536 537 538 539 540 541
	err = -ENODEV;
	if (!dp)
		goto err_unlock;

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

542
	ovs_flow_free(flow, false);
543 544 545 546 547
	return err;

err_unlock:
	rcu_read_unlock();
err_flow_free:
548
	ovs_flow_free(flow, false);
549 550 551 552 553 554 555
err_kfree_skb:
	kfree_skb(packet);
err:
	return err;
}

static const struct nla_policy packet_policy[OVS_PACKET_ATTR_MAX + 1] = {
556
	[OVS_PACKET_ATTR_PACKET] = { .len = ETH_HLEN },
557 558 559 560
	[OVS_PACKET_ATTR_KEY] = { .type = NLA_NESTED },
	[OVS_PACKET_ATTR_ACTIONS] = { .type = NLA_NESTED },
};

561
static const struct genl_ops dp_packet_genl_ops[] = {
562 563 564 565 566 567 568
	{ .cmd = OVS_PACKET_CMD_EXECUTE,
	  .flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
	  .policy = packet_policy,
	  .doit = ovs_packet_cmd_execute
	}
};

569 570
static void get_dp_stats(struct datapath *dp, struct ovs_dp_stats *stats,
			 struct ovs_dp_megaflow_stats *mega_stats)
571 572 573
{
	int i;

574 575
	memset(mega_stats, 0, sizeof(*mega_stats));

576
	stats->n_flows = ovs_flow_tbl_count(&dp->table);
577
	mega_stats->n_masks = ovs_flow_tbl_num_masks(&dp->table);
578 579

	stats->n_hit = stats->n_missed = stats->n_lost = 0;
580

581 582 583 584 585 586 587 588 589 590 591 592 593 594 595
	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;
596
		mega_stats->n_mask_hit += local_stats.n_mask_hit;
597 598 599 600 601 602 603 604 605 606 607 608 609 610
	}
}

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,
611
	.maxattr = OVS_FLOW_ATTR_MAX,
612 613
	.netnsok = true,
	.parallel_ops = true,
614 615 616 617 618 619
};

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

620 621 622 623
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 */
624
		+ nla_total_size(key_attr_size()) /* OVS_FLOW_ATTR_MASK */
625 626 627 628 629 630
		+ 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 */
}

631
/* Called with ovs_mutex. */
632
static int ovs_flow_cmd_fill_info(struct sw_flow *flow, struct datapath *dp,
633
				  struct sk_buff *skb, u32 portid,
634 635 636
				  u32 seq, u32 flags, u8 cmd)
{
	const int skb_orig_len = skb->len;
637
	struct nlattr *start;
638 639 640 641 642 643 644
	struct ovs_flow_stats stats;
	struct ovs_header *ovs_header;
	struct nlattr *nla;
	unsigned long used;
	u8 tcp_flags;
	int err;

645
	ovs_header = genlmsg_put(skb, portid, seq, &dp_flow_genl_family, flags, cmd);
646 647 648 649 650
	if (!ovs_header)
		return -EMSGSIZE;

	ovs_header->dp_ifindex = get_dpifindex(dp);

651
	/* Fill flow key. */
652 653 654
	nla = nla_nest_start(skb, OVS_FLOW_ATTR_KEY);
	if (!nla)
		goto nla_put_failure;
655

656
	err = ovs_nla_put_flow(&flow->unmasked_key, &flow->unmasked_key, skb);
657 658 659 660 661 662 663 664
	if (err)
		goto error;
	nla_nest_end(skb, nla);

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

665
	err = ovs_nla_put_flow(&flow->key, &flow->mask->key, skb);
666 667
	if (err)
		goto error;
668

669 670 671 672 673 674
	nla_nest_end(skb, nla);

	spin_lock_bh(&flow->lock);
	used = flow->used;
	stats.n_packets = flow->packet_count;
	stats.n_bytes = flow->byte_count;
675
	tcp_flags = (u8)ntohs(flow->tcp_flags);
676 677
	spin_unlock_bh(&flow->lock);

678 679 680
	if (used &&
	    nla_put_u64(skb, OVS_FLOW_ATTR_USED, ovs_flow_used_time(used)))
		goto nla_put_failure;
681

682 683 684 685
	if (stats.n_packets &&
	    nla_put(skb, OVS_FLOW_ATTR_STATS,
		    sizeof(struct ovs_flow_stats), &stats))
		goto nla_put_failure;
686

687 688 689
	if (tcp_flags &&
	    nla_put_u8(skb, OVS_FLOW_ATTR_TCP_FLAGS, tcp_flags))
		goto nla_put_failure;
690 691 692 693 694 695 696 697 698 699 700

	/* 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.
	 */
701 702
	start = nla_nest_start(skb, OVS_FLOW_ATTR_ACTIONS);
	if (start) {
703 704 705 706 707
		const struct sw_flow_actions *sf_acts;

		sf_acts = rcu_dereference_check(flow->sf_acts,
						lockdep_ovsl_is_held());

708 709
		err = ovs_nla_put_actions(sf_acts->actions,
					  sf_acts->actions_len, skb);
710 711 712 713 714 715 716 717 718 719
		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;
720 721 722 723 724 725 726 727 728 729 730 731 732 733

	return genlmsg_end(skb, ovs_header);

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

static struct sk_buff *ovs_flow_cmd_alloc_info(struct sw_flow *flow)
{
	const struct sw_flow_actions *sf_acts;

734
	sf_acts = ovsl_dereference(flow->sf_acts);
735

736
	return genlmsg_new(ovs_flow_cmd_msg_size(sf_acts), GFP_KERNEL);
737 738 739 740
}

static struct sk_buff *ovs_flow_cmd_build_info(struct sw_flow *flow,
					       struct datapath *dp,
741
					       u32 portid, u32 seq, u8 cmd)
742 743 744 745 746 747 748 749
{
	struct sk_buff *skb;
	int retval;

	skb = ovs_flow_cmd_alloc_info(flow);
	if (!skb)
		return ERR_PTR(-ENOMEM);

750
	retval = ovs_flow_cmd_fill_info(flow, dp, skb, portid, seq, 0, cmd);
751 752 753 754
	BUG_ON(retval < 0);
	return skb;
}

755 756 757 758 759 760 761 762
static struct sw_flow *__ovs_flow_tbl_lookup(struct flow_table *tbl,
					      const struct sw_flow_key *key)
{
	u32 __always_unused n_mask_hit;

	return ovs_flow_tbl_lookup(tbl, key, &n_mask_hit);
}

763 764 765 766
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;
767 768 769
	struct sw_flow_key key, masked_key;
	struct sw_flow *flow = NULL;
	struct sw_flow_mask mask;
770 771
	struct sk_buff *reply;
	struct datapath *dp;
772
	struct sw_flow_actions *acts = NULL;
773
	struct sw_flow_match match;
774 775 776 777 778 779
	int error;

	/* Extract key. */
	error = -EINVAL;
	if (!a[OVS_FLOW_ATTR_KEY])
		goto error;
780 781

	ovs_match_init(&match, &key, &mask);
782 783
	error = ovs_nla_get_match(&match,
				  a[OVS_FLOW_ATTR_KEY], a[OVS_FLOW_ATTR_MASK]);
784 785 786 787 788
	if (error)
		goto error;

	/* Validate actions. */
	if (a[OVS_FLOW_ATTR_ACTIONS]) {
789
		acts = ovs_nla_alloc_flow_actions(nla_len(a[OVS_FLOW_ATTR_ACTIONS]));
790 791
		error = PTR_ERR(acts);
		if (IS_ERR(acts))
792
			goto error;
793

794 795 796
		ovs_flow_mask_key(&masked_key, &key, &mask);
		error = ovs_nla_copy_actions(a[OVS_FLOW_ATTR_ACTIONS],
					     &masked_key, 0, &acts);
797 798
		if (error) {
			OVS_NLERR("Flow actions may not be safe on all matching packets.\n");
799
			goto err_kfree;
800
		}
801 802 803 804 805
	} else if (info->genlhdr->cmd == OVS_FLOW_CMD_NEW) {
		error = -EINVAL;
		goto error;
	}

806
	ovs_lock();
807
	dp = get_dp(sock_net(skb->sk), ovs_header->dp_ifindex);
808 809
	error = -ENODEV;
	if (!dp)
810
		goto err_unlock_ovs;
811

812
	/* Check if this is a duplicate flow */
813
	flow = __ovs_flow_tbl_lookup(&dp->table, &key);
814 815 816 817
	if (!flow) {
		/* Bail out if we're not allowed to create a new flow. */
		error = -ENOENT;
		if (info->genlhdr->cmd == OVS_FLOW_CMD_SET)
818
			goto err_unlock_ovs;
819 820 821 822 823

		/* Allocate flow. */
		flow = ovs_flow_alloc();
		if (IS_ERR(flow)) {
			error = PTR_ERR(flow);
824
			goto err_unlock_ovs;
825 826 827
		}
		clear_stats(flow);

828 829
		flow->key = masked_key;
		flow->unmasked_key = key;
830 831 832
		rcu_assign_pointer(flow->sf_acts, acts);

		/* Put flow in bucket. */
833 834 835 836 837
		error = ovs_flow_tbl_insert(&dp->table, flow, &mask);
		if (error) {
			acts = NULL;
			goto err_flow_free;
		}
838

839
		reply = ovs_flow_cmd_build_info(flow, dp, info->snd_portid,
840
						info->snd_seq, OVS_FLOW_CMD_NEW);
841 842 843 844 845 846 847 848 849 850 851 852 853
	} 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))
854
			goto err_unlock_ovs;
855

856 857
		/* The unmasked key has to be the same for flow updates. */
		error = -EINVAL;
858
		if (!ovs_flow_cmp_unmasked_key(flow, &match)) {
859 860 861 862
			OVS_NLERR("Flow modification message rejected, unmasked key does not match.\n");
			goto err_unlock_ovs;
		}

863
		/* Update actions. */
864
		old_acts = ovsl_dereference(flow->sf_acts);
865
		rcu_assign_pointer(flow->sf_acts, acts);
866
		ovs_nla_free_flow_actions(old_acts);
867

868
		reply = ovs_flow_cmd_build_info(flow, dp, info->snd_portid,
869 870 871 872 873 874 875 876 877
					       info->snd_seq, OVS_FLOW_CMD_NEW);

		/* Clear stats. */
		if (a[OVS_FLOW_ATTR_CLEAR]) {
			spin_lock_bh(&flow->lock);
			clear_stats(flow);
			spin_unlock_bh(&flow->lock);
		}
	}
878
	ovs_unlock();
879 880

	if (!IS_ERR(reply))
881 882
		ovs_notify(&dp_flow_genl_family, reply, info,
			   &ovs_dp_flow_multicast_group);
883
	else
884
		genl_set_err(&dp_flow_genl_family, sock_net(skb->sk), 0,
885
			     ovs_dp_flow_multicast_group.id, PTR_ERR(reply));
886 887
	return 0;

888 889
err_flow_free:
	ovs_flow_free(flow, false);
890 891
err_unlock_ovs:
	ovs_unlock();
892 893
err_kfree:
	kfree(acts);
894 895 896 897 898 899 900 901 902 903 904 905
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;
906
	struct sw_flow_match match;
907 908
	int err;

909 910
	if (!a[OVS_FLOW_ATTR_KEY]) {
		OVS_NLERR("Flow get message rejected, Key attribute missing.\n");
911
		return -EINVAL;
912 913 914
	}

	ovs_match_init(&match, &key, NULL);
915
	err = ovs_nla_get_match(&match, a[OVS_FLOW_ATTR_KEY], NULL);
916 917 918
	if (err)
		return err;

919
	ovs_lock();
920
	dp = get_dp(sock_net(skb->sk), ovs_header->dp_ifindex);
921 922 923 924
	if (!dp) {
		err = -ENODEV;
		goto unlock;
	}
925

926
	flow = __ovs_flow_tbl_lookup(&dp->table, &key);
927
	if (!flow || !ovs_flow_cmp_unmasked_key(flow, &match)) {
928 929 930
		err = -ENOENT;
		goto unlock;
	}
931

932
	reply = ovs_flow_cmd_build_info(flow, dp, info->snd_portid,
933
					info->snd_seq, OVS_FLOW_CMD_NEW);
934 935 936 937
	if (IS_ERR(reply)) {
		err = PTR_ERR(reply);
		goto unlock;
	}
938

939
	ovs_unlock();
940
	return genlmsg_reply(reply, info);
941 942 943
unlock:
	ovs_unlock();
	return err;
944 945 946 947 948 949 950 951 952 953
}

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;
954
	struct sw_flow_match match;
955 956
	int err;

957
	ovs_lock();
958
	dp = get_dp(sock_net(skb->sk), ovs_header->dp_ifindex);
959 960 961 962
	if (!dp) {
		err = -ENODEV;
		goto unlock;
	}
963

964
	if (!a[OVS_FLOW_ATTR_KEY]) {
965
		err = ovs_flow_tbl_flush(&dp->table);
966 967
		goto unlock;
	}
968 969

	ovs_match_init(&match, &key, NULL);
970
	err = ovs_nla_get_match(&match, a[OVS_FLOW_ATTR_KEY], NULL);
971
	if (err)
972
		goto unlock;
973

974
	flow = __ovs_flow_tbl_lookup(&dp->table, &key);
975
	if (!flow || !ovs_flow_cmp_unmasked_key(flow, &match)) {
976 977 978
		err = -ENOENT;
		goto unlock;
	}
979 980

	reply = ovs_flow_cmd_alloc_info(flow);
981 982 983 984
	if (!reply) {
		err = -ENOMEM;
		goto unlock;
	}
985

986
	ovs_flow_tbl_remove(&dp->table, flow);
987

988
	err = ovs_flow_cmd_fill_info(flow, dp, reply, info->snd_portid,
989 990 991
				     info->snd_seq, 0, OVS_FLOW_CMD_DEL);
	BUG_ON(err < 0);

992
	ovs_flow_free(flow, true);
993
	ovs_unlock();
994

995 996
	ovs_notify(&dp_flow_genl_family, reply, info,
		   &ovs_dp_flow_multicast_group);
997
	return 0;
998 999 1000
unlock:
	ovs_unlock();
	return err;
1001 1002 1003 1004 1005
}

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));
1006
	struct table_instance *ti;
1007 1008
	struct datapath *dp;

1009
	rcu_read_lock();
1010
	dp = get_dp(sock_net(skb->sk), ovs_header->dp_ifindex);
1011
	if (!dp) {
1012
		rcu_read_unlock();
1013
		return -ENODEV;
1014
	}
1015

1016
	ti = rcu_dereference(dp->table.ti);
1017 1018 1019 1020 1021 1022
	for (;;) {
		struct sw_flow *flow;
		u32 bucket, obj;

		bucket = cb->args[0];
		obj = cb->args[1];
1023
		flow = ovs_flow_tbl_dump_next(ti, &bucket, &obj);
1024 1025 1026 1027
		if (!flow)
			break;

		if (ovs_flow_cmd_fill_info(flow, dp, skb,
1028
					   NETLINK_CB(cb->skb).portid,
1029 1030 1031 1032 1033 1034 1035
					   cb->nlh->nlmsg_seq, NLM_F_MULTI,
					   OVS_FLOW_CMD_NEW) < 0)
			break;

		cb->args[0] = bucket;
		cb->args[1] = obj;
	}
1036
	rcu_read_unlock();
1037 1038 1039
	return skb->len;
}

1040
static const struct genl_ops dp_flow_genl_ops[] = {
1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073
	{ .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 },
};

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,
1074
	.maxattr = OVS_DP_ATTR_MAX,
1075 1076
	.netnsok = true,
	.parallel_ops = true,
1077 1078 1079 1080 1081 1082
};

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

1083 1084 1085 1086 1087 1088
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));
1089
	msgsize += nla_total_size(sizeof(struct ovs_dp_megaflow_stats));
1090 1091 1092 1093

	return msgsize;
}

1094
static int ovs_dp_cmd_fill_info(struct datapath *dp, struct sk_buff *skb,
1095
				u32 portid, u32 seq, u32 flags, u8 cmd)
1096 1097 1098
{
	struct ovs_header *ovs_header;
	struct ovs_dp_stats dp_stats;
1099
	struct ovs_dp_megaflow_stats dp_megaflow_stats;
1100 1101
	int err;

1102
	ovs_header = genlmsg_put(skb, portid, seq, &dp_datapath_genl_family,
1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114
				   flags, cmd);
	if (!ovs_header)
		goto error;

	ovs_header->dp_ifindex = get_dpifindex(dp);

	rcu_read_lock();