datapath.c 52.8 KB
Newer Older
1
/*
2
 * Copyright (c) 2007-2014 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 47 48 49 50
 *
 * 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>
#include <linux/openvswitch.h>
#include <linux/rculist.h>
#include <linux/dmi.h>
#include <net/genetlink.h>
51 52
#include <net/net_namespace.h>
#include <net/netns/generic.h>
53 54 55

#include "datapath.h"
#include "flow.h"
56
#include "flow_table.h"
57
#include "flow_netlink.h"
58
#include "vport-internal_dev.h"
59
#include "vport-netdev.h"
60

61
int ovs_net_id __read_mostly;
62
EXPORT_SYMBOL(ovs_net_id);
63

64 65 66 67
static struct genl_family dp_packet_genl_family;
static struct genl_family dp_flow_genl_family;
static struct genl_family dp_datapath_genl_family;

68 69
static const struct genl_multicast_group ovs_dp_flow_multicast_group = {
	.name = OVS_FLOW_MCGROUP,
70 71
};

72 73
static const struct genl_multicast_group ovs_dp_datapath_multicast_group = {
	.name = OVS_DATAPATH_MCGROUP,
74 75
};

76 77
static const struct genl_multicast_group ovs_dp_vport_multicast_group = {
	.name = OVS_VPORT_MCGROUP,
78 79
};

80 81
/* Check if need to build a reply message.
 * OVS userspace sets the NLM_F_ECHO flag if it needs the reply. */
82 83
static bool ovs_must_notify(struct genl_family *family, struct genl_info *info,
			    unsigned int group)
84 85
{
	return info->nlhdr->nlmsg_flags & NLM_F_ECHO ||
86 87
	       genl_has_listeners(family, genl_info_net(info)->genl_sock,
				  group);
88 89
}

90
static void ovs_notify(struct genl_family *family,
91
		       struct sk_buff *skb, struct genl_info *info)
92
{
93
	genl_notify(family, skb, genl_info_net(info), info->snd_portid,
94
		    0, info->nlhdr, GFP_KERNEL);
95 96
}

97 98 99
/**
 * DOC: Locking:
 *
100 101 102 103
 * 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.
104 105 106 107 108 109
 *
 * 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.
110 111
 *
 * The RTNL lock nests inside ovs_mutex.
112 113
 */

114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133
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;
}
134
EXPORT_SYMBOL(lockdep_ovsl_is_held);
135 136
#endif

137
static struct vport *new_vport(const struct vport_parms *);
138
static int queue_gso_packets(struct datapath *dp, struct sk_buff *,
139
			     const struct dp_upcall_info *);
140
static int queue_userspace_packet(struct datapath *dp, struct sk_buff *,
141 142
				  const struct dp_upcall_info *);

143
/* Must be called with rcu_read_lock or ovs_mutex. */
144
static struct datapath *get_dp(struct net *net, int dp_ifindex)
145 146 147 148 149
{
	struct datapath *dp = NULL;
	struct net_device *dev;

	rcu_read_lock();
150
	dev = dev_get_by_index_rcu(net, dp_ifindex);
151 152 153 154 155 156 157 158 159 160
	if (dev) {
		struct vport *vport = ovs_internal_dev_get_vport(dev);
		if (vport)
			dp = vport->dp;
	}
	rcu_read_unlock();

	return dp;
}

161
/* Must be called with rcu_read_lock or ovs_mutex. */
162
const char *ovs_dp_name(const struct datapath *dp)
163
{
164
	struct vport *vport = ovs_vport_ovsl_rcu(dp, OVSP_LOCAL);
165 166 167 168 169 170 171 172 173 174
	return vport->ops->get_name(vport);
}

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

	rcu_read_lock();

175
	local = ovs_vport_rcu(dp, OVSP_LOCAL);
176
	if (local)
177
		ifindex = netdev_vport_priv(local)->dev->ifindex;
178 179 180 181 182 183 184 185 186 187 188 189
	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);

190
	ovs_flow_tbl_destroy(&dp->table);
191
	free_percpu(dp->stats_percpu);
192
	release_net(ovs_dp_get_net(dp));
193
	kfree(dp->ports);
194 195 196
	kfree(dp);
}

197 198 199 200 201 202
static struct hlist_head *vport_hash_bucket(const struct datapath *dp,
					    u16 port_no)
{
	return &dp->ports[port_no & (DP_VPORT_HASH_BUCKETS - 1)];
}

203
/* Called with ovs_mutex or RCU read lock. */
204 205 206 207 208 209
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);
210
	hlist_for_each_entry_rcu(vport, head, dp_hash_node) {
211 212 213 214 215 216
		if (vport->port_no == port_no)
			return vport;
	}
	return NULL;
}

217
/* Called with ovs_mutex. */
218 219 220 221 222 223 224
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;
225
		struct hlist_head *head = vport_hash_bucket(dp, vport->port_no);
226

227
		hlist_add_head_rcu(&vport->dp_hash_node, head);
228 229 230 231 232 233
	}
	return vport;
}

void ovs_dp_detach_port(struct vport *p)
{
234
	ASSERT_OVSL();
235 236

	/* First drop references to device. */
237
	hlist_del_rcu(&p->dp_hash_node);
238 239 240 241 242 243

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

/* Must be called with rcu_read_lock. */
244
void ovs_dp_process_packet(struct sk_buff *skb, struct sw_flow_key *key)
245
{
246
	const struct vport *p = OVS_CB(skb)->input_vport;
247 248 249 250
	struct datapath *dp = p->dp;
	struct sw_flow *flow;
	struct dp_stats_percpu *stats;
	u64 *stats_counter;
251
	u32 n_mask_hit;
252

253
	stats = this_cpu_ptr(dp->stats_percpu);
254 255

	/* Look up flow. */
256
	flow = ovs_flow_tbl_lookup_stats(&dp->table, key, &n_mask_hit);
257 258
	if (unlikely(!flow)) {
		struct dp_upcall_info upcall;
259
		int error;
260 261

		upcall.cmd = OVS_PACKET_CMD_MISS;
262
		upcall.key = key;
263
		upcall.userdata = NULL;
264
		upcall.portid = ovs_vport_find_upcall_portid(p, skb);
265 266 267 268 269
		error = ovs_dp_upcall(dp, skb, &upcall);
		if (unlikely(error))
			kfree_skb(skb);
		else
			consume_skb(skb);
270 271 272 273 274 275
		stats_counter = &stats->n_missed;
		goto out;
	}

	OVS_CB(skb)->flow = flow;

276 277
	ovs_flow_stats_update(OVS_CB(skb)->flow, key->tp.flags, skb);
	ovs_execute_actions(dp, skb, key);
278
	stats_counter = &stats->n_hit;
279 280 281

out:
	/* Update datapath statistics. */
282
	u64_stats_update_begin(&stats->syncp);
283
	(*stats_counter)++;
284
	stats->n_mask_hit += n_mask_hit;
285
	u64_stats_update_end(&stats->syncp);
286 287 288
}

int ovs_dp_upcall(struct datapath *dp, struct sk_buff *skb,
289
		  const struct dp_upcall_info *upcall_info)
290 291 292 293
{
	struct dp_stats_percpu *stats;
	int err;

294
	if (upcall_info->portid == 0) {
295 296 297 298 299
		err = -ENOTCONN;
		goto err;
	}

	if (!skb_is_gso(skb))
300
		err = queue_userspace_packet(dp, skb, upcall_info);
301
	else
302
		err = queue_gso_packets(dp, skb, upcall_info);
303 304 305 306 307 308
	if (err)
		goto err;

	return 0;

err:
309
	stats = this_cpu_ptr(dp->stats_percpu);
310

311
	u64_stats_update_begin(&stats->syncp);
312
	stats->n_lost++;
313
	u64_stats_update_end(&stats->syncp);
314 315 316 317

	return err;
}

318
static int queue_gso_packets(struct datapath *dp, struct sk_buff *skb,
319 320
			     const struct dp_upcall_info *upcall_info)
{
321
	unsigned short gso_type = skb_shinfo(skb)->gso_type;
322 323 324 325 326
	struct dp_upcall_info later_info;
	struct sw_flow_key later_key;
	struct sk_buff *segs, *nskb;
	int err;

327
	segs = __skb_gso_segment(skb, NETIF_F_SG, false);
328 329
	if (IS_ERR(segs))
		return PTR_ERR(segs);
330 331
	if (segs == NULL)
		return -EINVAL;
332 333 334 335

	/* Queue all of the segments. */
	skb = segs;
	do {
336
		err = queue_userspace_packet(dp, skb, upcall_info);
337 338 339
		if (err)
			break;

340
		if (skb == segs && gso_type & SKB_GSO_UDP) {
341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365
			/* 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;
}

366 367 368
static size_t key_attr_size(void)
{
	return    nla_total_size(4)   /* OVS_KEY_ATTR_PRIORITY */
369 370 371 372 373 374 375 376
		+ 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 */
377
		  + nla_total_size(0)   /* OVS_TUNNEL_KEY_ATTR_OAM */
378
		  + nla_total_size(256)   /* OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS */
379 380 381 382 383 384 385 386 387 388 389 390
		+ 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 */
}

391 392
static size_t upcall_msg_size(const struct nlattr *userdata,
			      unsigned int hdrlen)
393 394
{
	size_t size = NLMSG_ALIGN(sizeof(struct ovs_header))
395
		+ nla_total_size(hdrlen) /* OVS_PACKET_ATTR_PACKET */
396 397 398 399 400 401 402 403 404
		+ nla_total_size(key_attr_size()); /* OVS_PACKET_ATTR_KEY */

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

	return size;
}

405
static int queue_userspace_packet(struct datapath *dp, struct sk_buff *skb,
406 407 408 409
				  const struct dp_upcall_info *upcall_info)
{
	struct ovs_header *upcall;
	struct sk_buff *nskb = NULL;
Li RongQing's avatar
Li RongQing committed
410
	struct sk_buff *user_skb = NULL; /* to be queued to userspace */
411
	struct nlattr *nla;
412
	struct genl_info info = {
413
		.dst_sk = ovs_dp_get_net(dp)->genl_sock,
414 415 416
		.snd_portid = upcall_info->portid,
	};
	size_t len;
417
	unsigned int hlen;
418 419 420 421 422
	int err, dp_ifindex;

	dp_ifindex = get_dpifindex(dp);
	if (!dp_ifindex)
		return -ENODEV;
423 424 425 426 427 428

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

429
		nskb = __vlan_put_tag(nskb, nskb->vlan_proto, vlan_tx_tag_get(nskb));
430
		if (!nskb)
431 432 433 434 435 436 437 438 439 440 441
			return -ENOMEM;

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

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

442 443 444 445 446 447 448 449 450 451 452 453 454 455 456
	/* 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);
457
	user_skb = genlmsg_new_unicast(len, &info, GFP_ATOMIC);
458 459 460 461 462 463 464 465 466 467
	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);
468 469
	err = ovs_nla_put_flow(upcall_info->key, upcall_info->key, user_skb);
	BUG_ON(err);
470 471 472
	nla_nest_end(user_skb, nla);

	if (upcall_info->userdata)
473 474 475
		__nla_put(user_skb, OVS_PACKET_ATTR_USERDATA,
			  nla_len(upcall_info->userdata),
			  nla_data(upcall_info->userdata));
476

477 478 479 480 481 482 483
	/* 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);
484

485 486 487
	err = skb_zerocopy(user_skb, skb, skb->len, hlen);
	if (err)
		goto out;
488

489 490 491 492 493 494 495 496
	/* 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);
	}

497
	((struct nlmsghdr *) user_skb->data)->nlmsg_len = user_skb->len;
498

499
	err = genlmsg_unicast(ovs_dp_get_net(dp), user_skb, upcall_info->portid);
Li RongQing's avatar
Li RongQing committed
500
	user_skb = NULL;
501
out:
502 503
	if (err)
		skb_tx_error(skb);
Li RongQing's avatar
Li RongQing committed
504
	kfree_skb(user_skb);
505 506 507 508 509 510 511 512 513 514 515 516 517
	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;
518
	struct vport *input_vport;
519 520 521 522 523
	int len;
	int err;

	err = -EINVAL;
	if (!a[OVS_PACKET_ATTR_PACKET] || !a[OVS_PACKET_ATTR_KEY] ||
524
	    !a[OVS_PACKET_ATTR_ACTIONS])
525 526 527 528 529 530 531 532 533
		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);

534
	nla_memcpy(__skb_put(packet, len), a[OVS_PACKET_ATTR_PACKET], len);
535 536 537 538 539 540 541

	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
542
	if (ntohs(eth->h_proto) >= ETH_P_802_3_MIN)
543 544 545 546 547
		packet->protocol = eth->h_proto;
	else
		packet->protocol = htons(ETH_P_802_2);

	/* Build an sw_flow for sending this packet. */
548
	flow = ovs_flow_alloc();
549 550 551 552
	err = PTR_ERR(flow);
	if (IS_ERR(flow))
		goto err_kfree_skb;

553 554
	err = ovs_flow_key_extract_userspace(a[OVS_PACKET_ATTR_KEY], packet,
					     &flow->key);
555 556 557
	if (err)
		goto err_flow_free;

558
	acts = ovs_nla_alloc_flow_actions(nla_len(a[OVS_PACKET_ATTR_ACTIONS]));
559 560 561
	err = PTR_ERR(acts);
	if (IS_ERR(acts))
		goto err_flow_free;
562

563
	err = ovs_nla_copy_actions(a[OVS_PACKET_ATTR_ACTIONS],
564
				   &flow->key, &acts);
565 566
	if (err)
		goto err_flow_free;
567

568 569 570
	rcu_assign_pointer(flow->sf_acts, acts);

	OVS_CB(packet)->egress_tun_info = NULL;
571 572
	OVS_CB(packet)->flow = flow;
	packet->priority = flow->key.phy.priority;
573
	packet->mark = flow->key.phy.skb_mark;
574 575

	rcu_read_lock();
576
	dp = get_dp(sock_net(skb->sk), ovs_header->dp_ifindex);
577 578 579 580
	err = -ENODEV;
	if (!dp)
		goto err_unlock;

581 582 583 584 585 586 587 588 589
	input_vport = ovs_vport_rcu(dp, flow->key.phy.in_port);
	if (!input_vport)
		input_vport = ovs_vport_rcu(dp, OVSP_LOCAL);

	if (!input_vport)
		goto err_unlock;

	OVS_CB(packet)->input_vport = input_vport;

590
	local_bh_disable();
591
	err = ovs_execute_actions(dp, packet, &flow->key);
592 593 594
	local_bh_enable();
	rcu_read_unlock();

595
	ovs_flow_free(flow, false);
596 597 598 599 600
	return err;

err_unlock:
	rcu_read_unlock();
err_flow_free:
601
	ovs_flow_free(flow, false);
602 603 604 605 606 607 608
err_kfree_skb:
	kfree_skb(packet);
err:
	return err;
}

static const struct nla_policy packet_policy[OVS_PACKET_ATTR_MAX + 1] = {
609
	[OVS_PACKET_ATTR_PACKET] = { .len = ETH_HLEN },
610 611 612 613
	[OVS_PACKET_ATTR_KEY] = { .type = NLA_NESTED },
	[OVS_PACKET_ATTR_ACTIONS] = { .type = NLA_NESTED },
};

614
static const struct genl_ops dp_packet_genl_ops[] = {
615 616 617 618 619 620 621
	{ .cmd = OVS_PACKET_CMD_EXECUTE,
	  .flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
	  .policy = packet_policy,
	  .doit = ovs_packet_cmd_execute
	}
};

622 623 624 625 626 627 628 629 630 631 632 633
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,
	.maxattr = OVS_PACKET_ATTR_MAX,
	.netnsok = true,
	.parallel_ops = true,
	.ops = dp_packet_genl_ops,
	.n_ops = ARRAY_SIZE(dp_packet_genl_ops),
};

634 635
static void get_dp_stats(struct datapath *dp, struct ovs_dp_stats *stats,
			 struct ovs_dp_megaflow_stats *mega_stats)
636 637 638
{
	int i;

639 640
	memset(mega_stats, 0, sizeof(*mega_stats));

641
	stats->n_flows = ovs_flow_tbl_count(&dp->table);
642
	mega_stats->n_masks = ovs_flow_tbl_num_masks(&dp->table);
643 644

	stats->n_hit = stats->n_missed = stats->n_lost = 0;
645

646 647 648 649 650 651 652 653
	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 {
654
			start = u64_stats_fetch_begin_irq(&percpu_stats->syncp);
655
			local_stats = *percpu_stats;
656
		} while (u64_stats_fetch_retry_irq(&percpu_stats->syncp, start));
657 658 659 660

		stats->n_hit += local_stats.n_hit;
		stats->n_missed += local_stats.n_missed;
		stats->n_lost += local_stats.n_lost;
661
		mega_stats->n_mask_hit += local_stats.n_mask_hit;
662 663 664
	}
}

665 666 667 668
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 */
669
		+ nla_total_size(key_attr_size()) /* OVS_FLOW_ATTR_MASK */
670 671 672 673 674 675
		+ 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 */
}

676
/* Called with ovs_mutex or RCU read lock. */
677 678
static int ovs_flow_cmd_fill_match(const struct sw_flow *flow,
				   struct sk_buff *skb)
679 680 681 682
{
	struct nlattr *nla;
	int err;

683
	/* Fill flow key. */
684 685
	nla = nla_nest_start(skb, OVS_FLOW_ATTR_KEY);
	if (!nla)
686
		return -EMSGSIZE;
687

688
	err = ovs_nla_put_flow(&flow->unmasked_key, &flow->unmasked_key, skb);
689
	if (err)
690 691
		return err;

692 693
	nla_nest_end(skb, nla);

694
	/* Fill flow mask. */
695 696
	nla = nla_nest_start(skb, OVS_FLOW_ATTR_MASK);
	if (!nla)
697
		return -EMSGSIZE;
698

699
	err = ovs_nla_put_flow(&flow->key, &flow->mask->key, skb);
700
	if (err)
701
		return err;
702

703
	nla_nest_end(skb, nla);
704 705 706 707 708 709 710 711 712 713
	return 0;
}

/* Called with ovs_mutex or RCU read lock. */
static int ovs_flow_cmd_fill_stats(const struct sw_flow *flow,
				   struct sk_buff *skb)
{
	struct ovs_flow_stats stats;
	__be16 tcp_flags;
	unsigned long used;
714

715
	ovs_flow_stats_get(flow, &stats, &used, &tcp_flags);
716

717 718
	if (used &&
	    nla_put_u64(skb, OVS_FLOW_ATTR_USED, ovs_flow_used_time(used)))
719
		return -EMSGSIZE;
720

721
	if (stats.n_packets &&
722
	    nla_put(skb, OVS_FLOW_ATTR_STATS, sizeof(struct ovs_flow_stats), &stats))
723
		return -EMSGSIZE;
724

725 726
	if ((u8)ntohs(tcp_flags) &&
	     nla_put_u8(skb, OVS_FLOW_ATTR_TCP_FLAGS, (u8)ntohs(tcp_flags)))
727 728 729 730 731 732 733 734 735 736 737
		return -EMSGSIZE;

	return 0;
}

/* Called with ovs_mutex or RCU read lock. */
static int ovs_flow_cmd_fill_actions(const struct sw_flow *flow,
				     struct sk_buff *skb, int skb_orig_len)
{
	struct nlattr *start;
	int err;
738 739 740 741 742 743 744 745 746 747 748

	/* 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.
	 */
749 750
	start = nla_nest_start(skb, OVS_FLOW_ATTR_ACTIONS);
	if (start) {
751 752
		const struct sw_flow_actions *sf_acts;

753
		sf_acts = rcu_dereference_ovsl(flow->sf_acts);
754 755
		err = ovs_nla_put_actions(sf_acts->actions,
					  sf_acts->actions_len, skb);
756

757 758 759 760
		if (!err)
			nla_nest_end(skb, start);
		else {
			if (skb_orig_len)
761
				return err;
762 763 764

			nla_nest_cancel(skb, start);
		}
765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798
	} else if (skb_orig_len) {
		return -EMSGSIZE;
	}

	return 0;
}

/* Called with ovs_mutex or RCU read lock. */
static int ovs_flow_cmd_fill_info(const struct sw_flow *flow, int dp_ifindex,
				  struct sk_buff *skb, u32 portid,
				  u32 seq, u32 flags, u8 cmd)
{
	const int skb_orig_len = skb->len;
	struct ovs_header *ovs_header;
	int err;

	ovs_header = genlmsg_put(skb, portid, seq, &dp_flow_genl_family,
				 flags, cmd);
	if (!ovs_header)
		return -EMSGSIZE;

	ovs_header->dp_ifindex = dp_ifindex;

	err = ovs_flow_cmd_fill_match(flow, skb);
	if (err)
		goto error;

	err = ovs_flow_cmd_fill_stats(flow, skb);
	if (err)
		goto error;

	err = ovs_flow_cmd_fill_actions(flow, skb, skb_orig_len);
	if (err)
		goto error;
799 800 801 802 803 804 805 806

	return genlmsg_end(skb, ovs_header);

error:
	genlmsg_cancel(skb, ovs_header);
	return err;
}

807 808
/* May not be called with RCU read lock. */
static struct sk_buff *ovs_flow_cmd_alloc_info(const struct sw_flow_actions *acts,
809 810
					       struct genl_info *info,
					       bool always)
811
{
812
	struct sk_buff *skb;
813

814
	if (!always && !ovs_must_notify(&dp_flow_genl_family, info, 0))
815 816
		return NULL;

817
	skb = genlmsg_new_unicast(ovs_flow_cmd_msg_size(acts), info, GFP_KERNEL);
818 819 820 821
	if (!skb)
		return ERR_PTR(-ENOMEM);

	return skb;
822 823
}

824 825 826 827 828
/* Called with ovs_mutex. */
static struct sk_buff *ovs_flow_cmd_build_info(const struct sw_flow *flow,
					       int dp_ifindex,
					       struct genl_info *info, u8 cmd,
					       bool always)
829 830 831 832
{
	struct sk_buff *skb;
	int retval;

833 834
	skb = ovs_flow_cmd_alloc_info(ovsl_dereference(flow->sf_acts), info,
				      always);
835
	if (IS_ERR_OR_NULL(skb))
836
		return skb;
837

838 839 840
	retval = ovs_flow_cmd_fill_info(flow, dp_ifindex, skb,
					info->snd_portid, info->snd_seq, 0,
					cmd);
841 842 843 844
	BUG_ON(retval < 0);
	return skb;
}

845
static int ovs_flow_cmd_new(struct sk_buff *skb, struct genl_info *info)
846 847 848
{
	struct nlattr **a = info->attrs;
	struct ovs_header *ovs_header = info->userhdr;
849
	struct sw_flow *flow, *new_flow;
850
	struct sw_flow_mask mask;
851 852
	struct sk_buff *reply;
	struct datapath *dp;
853
	struct sw_flow_actions *acts;
854
	struct sw_flow_match match;
855 856
	int error;

857
	/* Must have key and actions. */
858
	error = -EINVAL;
859 860
	if (!a[OVS_FLOW_ATTR_KEY]) {
		OVS_NLERR("Flow key attribute not present in new flow.\n");
861
		goto error;
862 863 864
	}
	if (!a[OVS_FLOW_ATTR_ACTIONS]) {
		OVS_NLERR("Flow actions attribute not present in new flow.\n");
865
		goto error;
866
	}
867

868 869 870 871 872 873 874 875 876 877 878
	/* Most of the time we need to allocate a new flow, do it before
	 * locking.
	 */
	new_flow = ovs_flow_alloc();
	if (IS_ERR(new_flow)) {
		error = PTR_ERR(new_flow);
		goto error;
	}

	/* Extract key. */
	ovs_match_init(&match, &new_flow->unmasked_key, &mask);
879
	error = ovs_nla_get_match(&match,
880
				  a[OVS_FLOW_ATTR_KEY], a[OVS_FLOW_ATTR_MASK]);
881
	if (error)
882
		goto err_kfree_flow;
883

884
	ovs_flow_mask_key(&new_flow->key, &new_flow->unmasked_key, &mask);
885

886
	/* Validate actions. */
887 888 889
	acts = ovs_nla_alloc_flow_actions(nla_len(a[OVS_FLOW_ATTR_ACTIONS]));
	error = PTR_ERR(acts);
	if (IS_ERR(acts))
890
		goto err_kfree_flow;
891

892
	error = ovs_nla_copy_actions(a[OVS_FLOW_ATTR_ACTIONS], &new_flow->key,
893
				     &acts);
894 895
	if (error) {
		OVS_NLERR("Flow actions may not be safe on all matching packets.\n");
896 897 898 899 900 901 902
		goto err_kfree_acts;
	}

	reply = ovs_flow_cmd_alloc_info(acts, info, false);
	if (IS_ERR(reply)) {
		error = PTR_ERR(reply);
		goto err_kfree_acts;
903 904
	}

905
	ovs_lock();
906
	dp = get_dp(sock_net(skb->sk), ovs_header->dp_ifindex);
907 908
	if (unlikely(!dp)) {
		error = -ENODEV;
909
		goto err_unlock_ovs;
910
	}
911
	/* Check if this is a duplicate flow */
912 913 914
	flow = ovs_flow_tbl_lookup(&dp->table, &new_flow->unmasked_key);
	if (likely(!flow)) {
		rcu_assign_pointer(new_flow->sf_acts, acts);
915 916

		/* Put flow in bucket. */
917 918
		error = ovs_flow_tbl_insert(&dp->table, new_flow, &mask);
		if (unlikely(error)) {
919
			acts = NULL;
920 921 922 923 924 925 926 927 928 929
			goto err_unlock_ovs;
		}

		if (unlikely(reply)) {
			error = ovs_flow_cmd_fill_info(new_flow,
						       ovs_header->dp_ifindex,
						       reply, info->snd_portid,
						       info->snd_seq, 0,
						       OVS_FLOW_CMD_NEW);
			BUG_ON(error < 0);
930
		}
931
		ovs_unlock();
932
	} else {
933 934
		struct sw_flow_actions *old_acts;

935 936 937 938 939 940
		/* 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.
		 */
941 942 943
		if (unlikely(info->nlhdr->nlmsg_flags & (NLM_F_CREATE
							 | NLM_F_EXCL))) {
			error = -EEXIST;
944
			goto err_unlock_ovs;
945
		}
946
		/* The unmasked key has to be the same for flow updates. */
947
		if (unlikely(!ovs_flow_cmp_unmasked_key(flow, &match))) {
948 949 950 951 952
			flow = ovs_flow_tbl_lookup_exact(&dp->table, &match);
			if (!flow) {
				error = -ENOENT;
				goto err_unlock_ovs;
			}
953
		}
954 955 956 957
		/* Update actions. */
		old_acts = ovsl_dereference(flow->sf_acts);
		rcu_assign_pointer(flow->sf_acts, acts);

958 959 960 961 962 963 964 965 966
		if (unlikely(reply)) {
			error = ovs_flow_cmd_fill_info(flow,
						       ovs_header->dp_ifindex,
						       reply, info->snd_portid,
						       info->snd_seq, 0,
						       OVS_FLOW_CMD_NEW);
			BUG_ON(error < 0);
		}
		ovs_unlock();
967

968 969
		ovs_nla_free_flow_actions(old_acts);
		ovs_flow_free(new_flow, false);
970
	}
971 972 973

	if (reply)
		ovs_notify(&dp_flow_genl_family, reply, info);
974 975 976 977
	return 0;

err_unlock_ovs:
	ovs_unlock();
978 979
	kfree_skb(reply);
err_kfree_acts:
980
	kfree(acts);
981 982
err_kfree_flow:
	ovs_flow_free(new_flow, false);
983 984 985
error:
	return error;
}
986

987 988 989 990 991 992 993 994 995 996 997 998 999
static struct sw_flow_actions *get_flow_actions(const struct nlattr *a,
						const struct sw_flow_key *key,
						const struct sw_flow_mask *mask)
{
	struct sw_flow_actions *acts;
	struct sw_flow_key masked_key;
	int error;

	acts = ovs_nla_alloc_flow_actions(nla_len(a));
	if (IS_ERR(acts))
		return acts;

	ovs_flow_mask_key(&masked_key, key, mask);
1000
	error = ovs_nla_copy_actions(a, &masked_key, &acts);
1001 1002 1003 1004 1005 1006 1007 1008 1009
	if (error) {
		OVS_NLERR("Flow actions may not be safe on all matching packets.\n");
		kfree(acts);
		return ERR_PTR(error);
	}

	return acts;
}

1010 1011 1012 1013
static int ovs_flow_cmd_set(struct sk_buff *skb, struct genl_info *info)
{
	struct nlattr **a = info->attrs;
	struct ovs_header *ovs_header = info->userhdr;
1014
	struct sw_flow_key key;
1015 1016 1017 1018
	struct sw_flow *flow;
	struct sw_flow_mask mask;
	struct sk_buff *reply = NULL;
	struct datapath *dp;
1019
	struct sw_flow_actions *old_acts = NULL, *acts = NULL;
1020 1021 1022 1023 1024
	struct sw_flow_match match;
	int error;

	/* Extract key. */
	error = -EINVAL;
1025 1026
	if (!a[OVS_FLOW_ATTR_KEY]) {
		OVS_NLERR("Flow key attribute not present in set flow.\n");
1027
		goto error;
1028
	}