datapath.c 51.7 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 62
int ovs_net_id __read_mostly;

63 64 65 66
static struct genl_family dp_packet_genl_family;
static struct genl_family dp_flow_genl_family;
static struct genl_family dp_datapath_genl_family;

67 68
static const struct genl_multicast_group ovs_dp_flow_multicast_group = {
	.name = OVS_FLOW_MCGROUP,
69 70
};

71 72
static const struct genl_multicast_group ovs_dp_datapath_multicast_group = {
	.name = OVS_DATAPATH_MCGROUP,
73 74
};

75 76
static const struct genl_multicast_group ovs_dp_vport_multicast_group = {
	.name = OVS_VPORT_MCGROUP,
77 78
};

79 80
/* Check if need to build a reply message.
 * OVS userspace sets the NLM_F_ECHO flag if it needs the reply. */
81 82
static bool ovs_must_notify(struct genl_family *family, struct genl_info *info,
			    unsigned int group)
83 84
{
	return info->nlhdr->nlmsg_flags & NLM_F_ECHO ||
85 86
	       genl_has_listeners(family, genl_info_net(info)->genl_sock,
				  group);
87 88
}

89
static void ovs_notify(struct genl_family *family,
90
		       struct sk_buff *skb, struct genl_info *info)
91
{
92
	genl_notify(family, skb, genl_info_net(info), info->snd_portid,
93
		    0, info->nlhdr, GFP_KERNEL);
94 95
}

96 97 98
/**
 * DOC: Locking:
 *
99 100 101 102
 * 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.
103 104 105 106 107 108
 *
 * 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.
109 110
 *
 * The RTNL lock nests inside ovs_mutex.
111 112
 */

113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134
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

135
static struct vport *new_vport(const struct vport_parms *);
136
static int queue_gso_packets(struct datapath *dp, struct sk_buff *,
137
			     const struct dp_upcall_info *);
138
static int queue_userspace_packet(struct datapath *dp, struct sk_buff *,
139 140
				  const struct dp_upcall_info *);

141
/* Must be called with rcu_read_lock or ovs_mutex. */
142
static struct datapath *get_dp(struct net *net, int dp_ifindex)
143 144 145 146 147
{
	struct datapath *dp = NULL;
	struct net_device *dev;

	rcu_read_lock();
148
	dev = dev_get_by_index_rcu(net, dp_ifindex);
149 150 151 152 153 154 155 156 157 158
	if (dev) {
		struct vport *vport = ovs_internal_dev_get_vport(dev);
		if (vport)
			dp = vport->dp;
	}
	rcu_read_unlock();

	return dp;
}

159
/* Must be called with rcu_read_lock or ovs_mutex. */
160
const char *ovs_dp_name(const struct datapath *dp)
161
{
162
	struct vport *vport = ovs_vport_ovsl_rcu(dp, OVSP_LOCAL);
163 164 165 166 167 168 169 170 171 172
	return vport->ops->get_name(vport);
}

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

	rcu_read_lock();

173
	local = ovs_vport_rcu(dp, OVSP_LOCAL);
174
	if (local)
175
		ifindex = netdev_vport_priv(local)->dev->ifindex;
176 177 178 179 180 181 182 183 184 185 186 187 188
	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);
189
	release_net(ovs_dp_get_net(dp));
190
	kfree(dp->ports);
191 192 193
	kfree(dp);
}

194 195 196 197 198 199
static struct hlist_head *vport_hash_bucket(const struct datapath *dp,
					    u16 port_no)
{
	return &dp->ports[port_no & (DP_VPORT_HASH_BUCKETS - 1)];
}

200
/* Called with ovs_mutex or RCU read lock. */
201 202 203 204 205 206
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);
207
	hlist_for_each_entry_rcu(vport, head, dp_hash_node) {
208 209 210 211 212 213
		if (vport->port_no == port_no)
			return vport;
	}
	return NULL;
}

214
/* Called with ovs_mutex. */
215 216 217 218 219 220 221
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;
222
		struct hlist_head *head = vport_hash_bucket(dp, vport->port_no);
223

224
		hlist_add_head_rcu(&vport->dp_hash_node, head);
225 226 227 228 229 230
	}
	return vport;
}

void ovs_dp_detach_port(struct vport *p)
{
231
	ASSERT_OVSL();
232 233

	/* First drop references to device. */
234
	hlist_del_rcu(&p->dp_hash_node);
235 236 237 238 239 240

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

/* Must be called with rcu_read_lock. */
241
void ovs_dp_process_packet(struct sk_buff *skb, struct sw_flow_key *key)
242
{
243
	const struct vport *p = OVS_CB(skb)->input_vport;
244 245 246 247
	struct datapath *dp = p->dp;
	struct sw_flow *flow;
	struct dp_stats_percpu *stats;
	u64 *stats_counter;
248
	u32 n_mask_hit;
249

250
	stats = this_cpu_ptr(dp->stats_percpu);
251 252

	/* Look up flow. */
253
	flow = ovs_flow_tbl_lookup_stats(&dp->table, key, &n_mask_hit);
254 255
	if (unlikely(!flow)) {
		struct dp_upcall_info upcall;
256
		int error;
257 258

		upcall.cmd = OVS_PACKET_CMD_MISS;
259
		upcall.key = key;
260
		upcall.userdata = NULL;
261
		upcall.portid = ovs_vport_find_upcall_portid(p, skb);
262 263 264 265 266
		error = ovs_dp_upcall(dp, skb, &upcall);
		if (unlikely(error))
			kfree_skb(skb);
		else
			consume_skb(skb);
267 268 269 270 271 272
		stats_counter = &stats->n_missed;
		goto out;
	}

	OVS_CB(skb)->flow = flow;

273 274
	ovs_flow_stats_update(OVS_CB(skb)->flow, key->tp.flags, skb);
	ovs_execute_actions(dp, skb, key);
275
	stats_counter = &stats->n_hit;
276 277 278

out:
	/* Update datapath statistics. */
279
	u64_stats_update_begin(&stats->syncp);
280
	(*stats_counter)++;
281
	stats->n_mask_hit += n_mask_hit;
282
	u64_stats_update_end(&stats->syncp);
283 284 285
}

int ovs_dp_upcall(struct datapath *dp, struct sk_buff *skb,
286
		  const struct dp_upcall_info *upcall_info)
287 288 289 290
{
	struct dp_stats_percpu *stats;
	int err;

291
	if (upcall_info->portid == 0) {
292 293 294 295 296
		err = -ENOTCONN;
		goto err;
	}

	if (!skb_is_gso(skb))
297
		err = queue_userspace_packet(dp, skb, upcall_info);
298
	else
299
		err = queue_gso_packets(dp, skb, upcall_info);
300 301 302 303 304 305
	if (err)
		goto err;

	return 0;

err:
306
	stats = this_cpu_ptr(dp->stats_percpu);
307

308
	u64_stats_update_begin(&stats->syncp);
309
	stats->n_lost++;
310
	u64_stats_update_end(&stats->syncp);
311 312 313 314

	return err;
}

315
static int queue_gso_packets(struct datapath *dp, 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, false);
325 326
	if (IS_ERR(segs))
		return PTR_ERR(segs);
327 328
	if (segs == NULL)
		return -EINVAL;
329 330 331 332

	/* Queue all of the segments. */
	skb = segs;
	do {
333
		err = queue_userspace_packet(dp, skb, upcall_info);
334 335 336
		if (err)
			break;

337
		if (skb == segs && gso_type & SKB_GSO_UDP) {
338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362
			/* 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;
}

363 364 365
static size_t key_attr_size(void)
{
	return    nla_total_size(4)   /* OVS_KEY_ATTR_PRIORITY */
366 367 368 369 370 371 372 373
		+ 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 */
374
		  + nla_total_size(0)   /* OVS_TUNNEL_KEY_ATTR_OAM */
375
		  + nla_total_size(256)   /* OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS */
376 377 378 379 380 381 382 383 384 385 386 387
		+ 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 */
}

388 389
static size_t upcall_msg_size(const struct nlattr *userdata,
			      unsigned int hdrlen)
390 391
{
	size_t size = NLMSG_ALIGN(sizeof(struct ovs_header))
392
		+ nla_total_size(hdrlen) /* OVS_PACKET_ATTR_PACKET */
393 394 395 396 397 398 399 400 401
		+ nla_total_size(key_attr_size()); /* OVS_PACKET_ATTR_KEY */

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

	return size;
}

402
static int queue_userspace_packet(struct datapath *dp, struct sk_buff *skb,
403 404 405 406
				  const struct dp_upcall_info *upcall_info)
{
	struct ovs_header *upcall;
	struct sk_buff *nskb = NULL;
Li RongQing's avatar
Li RongQing committed
407
	struct sk_buff *user_skb = NULL; /* to be queued to userspace */
408
	struct nlattr *nla;
409
	struct genl_info info = {
410
		.dst_sk = ovs_dp_get_net(dp)->genl_sock,
411 412 413
		.snd_portid = upcall_info->portid,
	};
	size_t len;
414
	unsigned int hlen;
415 416 417 418 419
	int err, dp_ifindex;

	dp_ifindex = get_dpifindex(dp);
	if (!dp_ifindex)
		return -ENODEV;
420 421 422 423 424 425

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

426
		nskb = __vlan_put_tag(nskb, nskb->vlan_proto, vlan_tx_tag_get(nskb));
427
		if (!nskb)
428 429 430 431 432 433 434 435 436 437 438
			return -ENOMEM;

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

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

439 440 441 442 443 444 445 446 447 448 449 450 451 452 453
	/* 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);
454
	user_skb = genlmsg_new_unicast(len, &info, GFP_ATOMIC);
455 456 457 458 459 460 461 462 463 464
	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);
465 466
	err = ovs_nla_put_flow(upcall_info->key, upcall_info->key, user_skb);
	BUG_ON(err);
467 468 469
	nla_nest_end(user_skb, nla);

	if (upcall_info->userdata)
470 471 472
		__nla_put(user_skb, OVS_PACKET_ATTR_USERDATA,
			  nla_len(upcall_info->userdata),
			  nla_data(upcall_info->userdata));
473

474 475 476 477 478 479 480
	/* 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);
481

482 483 484
	err = skb_zerocopy(user_skb, skb, skb->len, hlen);
	if (err)
		goto out;
485

486 487 488 489 490 491 492 493
	/* 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);
	}

494
	((struct nlmsghdr *) user_skb->data)->nlmsg_len = user_skb->len;
495

496
	err = genlmsg_unicast(ovs_dp_get_net(dp), user_skb, upcall_info->portid);
Li RongQing's avatar
Li RongQing committed
497
	user_skb = NULL;
498
out:
499 500
	if (err)
		skb_tx_error(skb);
Li RongQing's avatar
Li RongQing committed
501
	kfree_skb(user_skb);
502 503 504 505 506 507 508 509 510 511 512 513 514
	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;
515
	struct vport *input_vport;
516 517 518 519 520
	int len;
	int err;

	err = -EINVAL;
	if (!a[OVS_PACKET_ATTR_PACKET] || !a[OVS_PACKET_ATTR_KEY] ||
521
	    !a[OVS_PACKET_ATTR_ACTIONS])
522 523 524 525 526 527 528 529 530
		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);

531
	nla_memcpy(__skb_put(packet, len), a[OVS_PACKET_ATTR_PACKET], len);
532 533 534 535 536 537 538

	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
539
	if (ntohs(eth->h_proto) >= ETH_P_802_3_MIN)
540 541 542 543 544
		packet->protocol = eth->h_proto;
	else
		packet->protocol = htons(ETH_P_802_2);

	/* Build an sw_flow for sending this packet. */
545
	flow = ovs_flow_alloc();
546 547 548 549
	err = PTR_ERR(flow);
	if (IS_ERR(flow))
		goto err_kfree_skb;

550 551
	err = ovs_flow_key_extract_userspace(a[OVS_PACKET_ATTR_KEY], packet,
					     &flow->key);
552 553 554
	if (err)
		goto err_flow_free;

555
	acts = ovs_nla_alloc_flow_actions(nla_len(a[OVS_PACKET_ATTR_ACTIONS]));
556 557 558
	err = PTR_ERR(acts);
	if (IS_ERR(acts))
		goto err_flow_free;
559

560 561
	err = ovs_nla_copy_actions(a[OVS_PACKET_ATTR_ACTIONS],
				   &flow->key, 0, &acts);
562 563
	if (err)
		goto err_flow_free;
564

565 566 567
	rcu_assign_pointer(flow->sf_acts, acts);

	OVS_CB(packet)->egress_tun_info = NULL;
568 569
	OVS_CB(packet)->flow = flow;
	packet->priority = flow->key.phy.priority;
570
	packet->mark = flow->key.phy.skb_mark;
571 572

	rcu_read_lock();
573
	dp = get_dp(sock_net(skb->sk), ovs_header->dp_ifindex);
574 575 576 577
	err = -ENODEV;
	if (!dp)
		goto err_unlock;

578 579 580 581 582 583 584 585 586
	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;

587
	local_bh_disable();
588
	err = ovs_execute_actions(dp, packet, &flow->key);
589 590 591
	local_bh_enable();
	rcu_read_unlock();

592
	ovs_flow_free(flow, false);
593 594 595 596 597
	return err;

err_unlock:
	rcu_read_unlock();
err_flow_free:
598
	ovs_flow_free(flow, false);
599 600 601 602 603 604 605
err_kfree_skb:
	kfree_skb(packet);
err:
	return err;
}

static const struct nla_policy packet_policy[OVS_PACKET_ATTR_MAX + 1] = {
606
	[OVS_PACKET_ATTR_PACKET] = { .len = ETH_HLEN },
607 608 609 610
	[OVS_PACKET_ATTR_KEY] = { .type = NLA_NESTED },
	[OVS_PACKET_ATTR_ACTIONS] = { .type = NLA_NESTED },
};

611
static const struct genl_ops dp_packet_genl_ops[] = {
612 613 614 615 616 617 618
	{ .cmd = OVS_PACKET_CMD_EXECUTE,
	  .flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
	  .policy = packet_policy,
	  .doit = ovs_packet_cmd_execute
	}
};

619 620 621 622 623 624 625 626 627 628 629 630
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),
};

631 632
static void get_dp_stats(struct datapath *dp, struct ovs_dp_stats *stats,
			 struct ovs_dp_megaflow_stats *mega_stats)
633 634 635
{
	int i;

636 637
	memset(mega_stats, 0, sizeof(*mega_stats));

638
	stats->n_flows = ovs_flow_tbl_count(&dp->table);
639
	mega_stats->n_masks = ovs_flow_tbl_num_masks(&dp->table);
640 641

	stats->n_hit = stats->n_missed = stats->n_lost = 0;
642

643 644 645 646 647 648 649 650
	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 {
651
			start = u64_stats_fetch_begin_irq(&percpu_stats->syncp);
652
			local_stats = *percpu_stats;
653
		} while (u64_stats_fetch_retry_irq(&percpu_stats->syncp, start));
654 655 656 657

		stats->n_hit += local_stats.n_hit;
		stats->n_missed += local_stats.n_missed;
		stats->n_lost += local_stats.n_lost;
658
		mega_stats->n_mask_hit += local_stats.n_mask_hit;
659 660 661
	}
}

662 663 664 665
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 */
666
		+ nla_total_size(key_attr_size()) /* OVS_FLOW_ATTR_MASK */
667 668 669 670 671 672
		+ 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 */
}

673
/* Called with ovs_mutex or RCU read lock. */
674
static int ovs_flow_cmd_fill_info(const struct sw_flow *flow, int dp_ifindex,
675
				  struct sk_buff *skb, u32 portid,
676 677 678
				  u32 seq, u32 flags, u8 cmd)
{
	const int skb_orig_len = skb->len;
679
	struct nlattr *start;
680
	struct ovs_flow_stats stats;
681 682
	__be16 tcp_flags;
	unsigned long used;
683 684 685 686
	struct ovs_header *ovs_header;
	struct nlattr *nla;
	int err;

687
	ovs_header = genlmsg_put(skb, portid, seq, &dp_flow_genl_family, flags, cmd);
688 689 690
	if (!ovs_header)
		return -EMSGSIZE;

691
	ovs_header->dp_ifindex = dp_ifindex;
692

693
	/* Fill flow key. */
694 695 696
	nla = nla_nest_start(skb, OVS_FLOW_ATTR_KEY);
	if (!nla)
		goto nla_put_failure;
697

698
	err = ovs_nla_put_flow(&flow->unmasked_key, &flow->unmasked_key, skb);
699 700 701 702 703 704 705 706
	if (err)
		goto error;
	nla_nest_end(skb, nla);

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

707
	err = ovs_nla_put_flow(&flow->key, &flow->mask->key, skb);
708 709
	if (err)
		goto error;
710

711 712
	nla_nest_end(skb, nla);

713
	ovs_flow_stats_get(flow, &stats, &used, &tcp_flags);
714

715 716 717
	if (used &&
	    nla_put_u64(skb, OVS_FLOW_ATTR_USED, ovs_flow_used_time(used)))
		goto nla_put_failure;
718

719
	if (stats.n_packets &&
720
	    nla_put(skb, OVS_FLOW_ATTR_STATS, sizeof(struct ovs_flow_stats), &stats))
721
		goto nla_put_failure;
722

723 724
	if ((u8)ntohs(tcp_flags) &&
	     nla_put_u8(skb, OVS_FLOW_ATTR_TCP_FLAGS, (u8)ntohs(tcp_flags)))
725
		goto nla_put_failure;
726 727 728 729 730 731 732 733 734 735 736

	/* 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.
	 */
737 738
	start = nla_nest_start(skb, OVS_FLOW_ATTR_ACTIONS);
	if (start) {
739 740
		const struct sw_flow_actions *sf_acts;

741
		sf_acts = rcu_dereference_ovsl(flow->sf_acts);
742 743
		err = ovs_nla_put_actions(sf_acts->actions,
					  sf_acts->actions_len, skb);
744

745 746 747 748 749 750 751 752 753 754
		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;
755 756 757 758 759 760 761 762 763 764

	return genlmsg_end(skb, ovs_header);

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

765 766
/* May not be called with RCU read lock. */
static struct sk_buff *ovs_flow_cmd_alloc_info(const struct sw_flow_actions *acts,
767 768
					       struct genl_info *info,
					       bool always)
769
{
770
	struct sk_buff *skb;
771

772
	if (!always && !ovs_must_notify(&dp_flow_genl_family, info, 0))
773 774
		return NULL;

775
	skb = genlmsg_new_unicast(ovs_flow_cmd_msg_size(acts), info, GFP_KERNEL);
776 777 778 779
	if (!skb)
		return ERR_PTR(-ENOMEM);

	return skb;
780 781
}

782 783 784 785 786
/* 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)
787 788 789 790
{
	struct sk_buff *skb;
	int retval;

791 792
	skb = ovs_flow_cmd_alloc_info(ovsl_dereference(flow->sf_acts), info,
				      always);
793
	if (IS_ERR_OR_NULL(skb))
794
		return skb;
795

796 797 798
	retval = ovs_flow_cmd_fill_info(flow, dp_ifindex, skb,
					info->snd_portid, info->snd_seq, 0,
					cmd);
799 800 801 802
	BUG_ON(retval < 0);
	return skb;
}

803
static int ovs_flow_cmd_new(struct sk_buff *skb, struct genl_info *info)
804 805 806
{
	struct nlattr **a = info->attrs;
	struct ovs_header *ovs_header = info->userhdr;
807
	struct sw_flow *flow, *new_flow;
808
	struct sw_flow_mask mask;
809 810
	struct sk_buff *reply;
	struct datapath *dp;
811
	struct sw_flow_actions *acts;
812
	struct sw_flow_match match;
813 814
	int error;

815
	/* Must have key and actions. */
816 817 818
	error = -EINVAL;
	if (!a[OVS_FLOW_ATTR_KEY])
		goto error;
819 820
	if (!a[OVS_FLOW_ATTR_ACTIONS])
		goto error;
821

822 823 824 825 826 827 828 829 830 831 832
	/* 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);
833
	error = ovs_nla_get_match(&match,
834
				  a[OVS_FLOW_ATTR_KEY], a[OVS_FLOW_ATTR_MASK]);
835
	if (error)
836
		goto err_kfree_flow;
837

838
	ovs_flow_mask_key(&new_flow->key, &new_flow->unmasked_key, &mask);
839

840
	/* Validate actions. */
841 842 843
	acts = ovs_nla_alloc_flow_actions(nla_len(a[OVS_FLOW_ATTR_ACTIONS]));
	error = PTR_ERR(acts);
	if (IS_ERR(acts))
844
		goto err_kfree_flow;
845

846 847
	error = ovs_nla_copy_actions(a[OVS_FLOW_ATTR_ACTIONS], &new_flow->key,
				     0, &acts);
848 849
	if (error) {
		OVS_NLERR("Flow actions may not be safe on all matching packets.\n");
850 851 852 853 854 855 856
		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;
857 858
	}

859
	ovs_lock();
860
	dp = get_dp(sock_net(skb->sk), ovs_header->dp_ifindex);
861 862
	if (unlikely(!dp)) {
		error = -ENODEV;
863
		goto err_unlock_ovs;
864
	}
865
	/* Check if this is a duplicate flow */
866 867 868
	flow = ovs_flow_tbl_lookup(&dp->table, &new_flow->unmasked_key);
	if (likely(!flow)) {
		rcu_assign_pointer(new_flow->sf_acts, acts);
869 870

		/* Put flow in bucket. */
871 872
		error = ovs_flow_tbl_insert(&dp->table, new_flow, &mask);
		if (unlikely(error)) {
873
			acts = NULL;
874 875 876 877 878 879 880 881 882 883
			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);
884
		}
885
		ovs_unlock();
886
	} else {
887 888
		struct sw_flow_actions *old_acts;

889 890 891 892 893 894
		/* 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.
		 */
895 896 897
		if (unlikely(info->nlhdr->nlmsg_flags & (NLM_F_CREATE
							 | NLM_F_EXCL))) {
			error = -EEXIST;
898
			goto err_unlock_ovs;
899
		}
900
		/* The unmasked key has to be the same for flow updates. */
901
		if (unlikely(!ovs_flow_cmp_unmasked_key(flow, &match))) {
902 903 904 905 906
			flow = ovs_flow_tbl_lookup_exact(&dp->table, &match);
			if (!flow) {
				error = -ENOENT;
				goto err_unlock_ovs;
			}
907
		}
908 909 910 911
		/* Update actions. */
		old_acts = ovsl_dereference(flow->sf_acts);
		rcu_assign_pointer(flow->sf_acts, acts);

912 913 914 915 916 917 918 919 920
		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();
921

922 923
		ovs_nla_free_flow_actions(old_acts);
		ovs_flow_free(new_flow, false);
924
	}
925 926 927

	if (reply)
		ovs_notify(&dp_flow_genl_family, reply, info);
928 929 930 931
	return 0;

err_unlock_ovs:
	ovs_unlock();
932 933
	kfree_skb(reply);
err_kfree_acts:
934
	kfree(acts);
935 936
err_kfree_flow:
	ovs_flow_free(new_flow, false);
937 938 939
error:
	return error;
}
940

941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963
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);
	error = ovs_nla_copy_actions(a, &masked_key, 0, &acts);
	if (error) {
		OVS_NLERR("Flow actions may not be safe on all matching packets.\n");
		kfree(acts);
		return ERR_PTR(error);
	}

	return acts;
}

964 965 966 967
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;
968
	struct sw_flow_key key;
969 970 971 972
	struct sw_flow *flow;
	struct sw_flow_mask mask;
	struct sk_buff *reply = NULL;
	struct datapath *dp;
973
	struct sw_flow_actions *old_acts = NULL, *acts = NULL;
974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989
	struct sw_flow_match match;
	int error;

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

	ovs_match_init(&match, &key, &mask);
	error = ovs_nla_get_match(&match,
				  a[OVS_FLOW_ATTR_KEY], a[OVS_FLOW_ATTR_MASK]);
	if (error)
		goto error;

	/* Validate actions. */
	if (a[OVS_FLOW_ATTR_ACTIONS]) {
990 991 992
		acts = get_flow_actions(a[OVS_FLOW_ATTR_ACTIONS], &key, &mask);
		if (IS_ERR(acts)) {
			error = PTR_ERR(acts);
993
			goto error;
994 995 996 997 998 999 1000 1001 1002
		}
	}

	/* Can allocate before locking if have acts. */
	if (acts) {
		reply = ovs_flow_cmd_alloc_info(acts, info, false);
		if (IS_ERR(reply)) {
			error = PTR_ERR(reply);
			goto err_kfree_acts;