datapath.c 50.9 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
 *
 * 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>
50 51
#include <linux/genetlink.h>
#include <net/genetlink.h>
52
#include <net/genetlink.h>
53 54
#include <net/net_namespace.h>
#include <net/netns/generic.h>
55 56 57

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

63 64
int ovs_net_id __read_mostly;

65 66 67 68
static struct genl_family dp_packet_genl_family;
static struct genl_family dp_flow_genl_family;
static struct genl_family dp_datapath_genl_family;

69 70
static const struct genl_multicast_group ovs_dp_flow_multicast_group = {
	.name = OVS_FLOW_MCGROUP,
71 72
};

73 74
static const struct genl_multicast_group ovs_dp_datapath_multicast_group = {
	.name = OVS_DATAPATH_MCGROUP,
75 76
};

77 78
static const struct genl_multicast_group ovs_dp_vport_multicast_group = {
	.name = OVS_VPORT_MCGROUP,
79 80
};

81 82 83 84 85 86 87 88 89
/* Check if need to build a reply message.
 * OVS userspace sets the NLM_F_ECHO flag if it needs the reply. */
static bool ovs_must_notify(struct genl_info *info,
			    const struct genl_multicast_group *grp)
{
	return info->nlhdr->nlmsg_flags & NLM_F_ECHO ||
		netlink_has_listeners(genl_info_net(info)->genl_sock, 0);
}

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
 */

114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135
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

136
static struct vport *new_vport(const struct vport_parms *);
137
static int queue_gso_packets(struct datapath *dp, struct sk_buff *,
138
			     const struct dp_upcall_info *);
139
static int queue_userspace_packet(struct datapath *dp, struct sk_buff *,
140 141
				  const struct dp_upcall_info *);

142
/* Must be called with rcu_read_lock or ovs_mutex. */
143
static struct datapath *get_dp(struct net *net, int dp_ifindex)
144 145 146 147 148
{
	struct datapath *dp = NULL;
	struct net_device *dev;

	rcu_read_lock();
149
	dev = dev_get_by_index_rcu(net, dp_ifindex);
150 151 152 153 154 155 156 157 158 159
	if (dev) {
		struct vport *vport = ovs_internal_dev_get_vport(dev);
		if (vport)
			dp = vport->dp;
	}
	rcu_read_unlock();

	return dp;
}

160
/* Must be called with rcu_read_lock or ovs_mutex. */
S
Stephen Hemminger 已提交
161
static const char *ovs_dp_name(const struct datapath *dp)
162
{
163
	struct vport *vport = ovs_vport_ovsl_rcu(dp, OVSP_LOCAL);
164 165 166 167 168 169 170 171 172 173
	return vport->ops->get_name(vport);
}

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

	rcu_read_lock();

174
	local = ovs_vport_rcu(dp, OVSP_LOCAL);
175
	if (local)
176
		ifindex = netdev_vport_priv(local)->dev->ifindex;
177 178 179 180 181 182 183 184 185 186 187 188 189
	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);

	free_percpu(dp->stats_percpu);
190
	release_net(ovs_dp_get_net(dp));
191
	kfree(dp->ports);
192 193 194
	kfree(dp);
}

195 196 197 198 199 200
static struct hlist_head *vport_hash_bucket(const struct datapath *dp,
					    u16 port_no)
{
	return &dp->ports[port_no & (DP_VPORT_HASH_BUCKETS - 1)];
}

J
Jarno Rajahalme 已提交
201
/* Called with ovs_mutex or RCU read lock. */
202 203 204 205 206 207
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);
208
	hlist_for_each_entry_rcu(vport, head, dp_hash_node) {
209 210 211 212 213 214
		if (vport->port_no == port_no)
			return vport;
	}
	return NULL;
}

215
/* Called with ovs_mutex. */
216 217 218 219 220 221 222
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;
223
		struct hlist_head *head = vport_hash_bucket(dp, vport->port_no);
224

225
		hlist_add_head_rcu(&vport->dp_hash_node, head);
226 227 228 229 230 231
	}
	return vport;
}

void ovs_dp_detach_port(struct vport *p)
{
232
	ASSERT_OVSL();
233 234

	/* First drop references to device. */
235
	hlist_del_rcu(&p->dp_hash_node);
236 237 238 239 240 241 242 243 244 245 246 247 248

	/* 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;
249
	u32 n_mask_hit;
250 251
	int error;

252
	stats = this_cpu_ptr(dp->stats_percpu);
253 254

	/* Extract flow from 'skb' into 'key'. */
255
	error = ovs_flow_extract(skb, p->port_no, &key);
256 257 258 259 260 261
	if (unlikely(error)) {
		kfree_skb(skb);
		return;
	}

	/* Look up flow. */
262
	flow = ovs_flow_tbl_lookup_stats(&dp->table, &key, &n_mask_hit);
263 264 265 266 267 268
	if (unlikely(!flow)) {
		struct dp_upcall_info upcall;

		upcall.cmd = OVS_PACKET_CMD_MISS;
		upcall.key = &key;
		upcall.userdata = NULL;
269
		upcall.portid = p->upcall_portid;
270 271 272 273 274 275 276
		ovs_dp_upcall(dp, skb, &upcall);
		consume_skb(skb);
		stats_counter = &stats->n_missed;
		goto out;
	}

	OVS_CB(skb)->flow = flow;
277
	OVS_CB(skb)->pkt_key = &key;
278

279
	ovs_flow_stats_update(OVS_CB(skb)->flow, key.tp.flags, skb);
280
	ovs_execute_actions(dp, skb);
281
	stats_counter = &stats->n_hit;
282 283 284

out:
	/* Update datapath statistics. */
285
	u64_stats_update_begin(&stats->syncp);
286
	(*stats_counter)++;
287
	stats->n_mask_hit += n_mask_hit;
288
	u64_stats_update_end(&stats->syncp);
289 290 291
}

int ovs_dp_upcall(struct datapath *dp, struct sk_buff *skb,
292
		  const struct dp_upcall_info *upcall_info)
293 294 295 296
{
	struct dp_stats_percpu *stats;
	int err;

297
	if (upcall_info->portid == 0) {
298 299 300 301 302
		err = -ENOTCONN;
		goto err;
	}

	if (!skb_is_gso(skb))
303
		err = queue_userspace_packet(dp, skb, upcall_info);
304
	else
305
		err = queue_gso_packets(dp, skb, upcall_info);
306 307 308 309 310 311
	if (err)
		goto err;

	return 0;

err:
312
	stats = this_cpu_ptr(dp->stats_percpu);
313

314
	u64_stats_update_begin(&stats->syncp);
315
	stats->n_lost++;
316
	u64_stats_update_end(&stats->syncp);
317 318 319 320

	return err;
}

321
static int queue_gso_packets(struct datapath *dp, struct sk_buff *skb,
322 323
			     const struct dp_upcall_info *upcall_info)
{
324
	unsigned short gso_type = skb_shinfo(skb)->gso_type;
325 326 327 328 329
	struct dp_upcall_info later_info;
	struct sw_flow_key later_key;
	struct sk_buff *segs, *nskb;
	int err;

330
	segs = __skb_gso_segment(skb, NETIF_F_SG, false);
331 332
	if (IS_ERR(segs))
		return PTR_ERR(segs);
333 334 335 336

	/* Queue all of the segments. */
	skb = segs;
	do {
337
		err = queue_userspace_packet(dp, skb, upcall_info);
338 339 340
		if (err)
			break;

341
		if (skb == segs && gso_type & SKB_GSO_UDP) {
342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366
			/* 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;
}

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

390 391
static size_t upcall_msg_size(const struct nlattr *userdata,
			      unsigned int hdrlen)
392 393
{
	size_t size = NLMSG_ALIGN(sizeof(struct ovs_header))
394
		+ nla_total_size(hdrlen) /* OVS_PACKET_ATTR_PACKET */
395 396 397 398 399 400 401 402 403
		+ nla_total_size(key_attr_size()); /* OVS_PACKET_ATTR_KEY */

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

	return size;
}

404
static int queue_userspace_packet(struct datapath *dp, struct sk_buff *skb,
405 406 407 408 409 410
				  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;
411
	struct genl_info info = {
412
		.dst_sk = ovs_dp_get_net(dp)->genl_sock,
413 414 415
		.snd_portid = upcall_info->portid,
	};
	size_t len;
416
	unsigned int hlen;
417 418 419 420 421
	int err, dp_ifindex;

	dp_ifindex = get_dpifindex(dp);
	if (!dp_ifindex)
		return -ENODEV;
422 423 424 425 426 427

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

428
		nskb = __vlan_put_tag(nskb, nskb->vlan_proto, vlan_tx_tag_get(nskb));
429
		if (!nskb)
430 431 432 433 434 435 436 437 438 439 440
			return -ENOMEM;

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

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

441 442 443 444 445 446 447 448 449 450 451 452 453 454 455
	/* 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);
456
	user_skb = genlmsg_new_unicast(len, &info, GFP_ATOMIC);
457 458 459 460 461 462 463 464 465 466
	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);
467
	ovs_nla_put_flow(upcall_info->key, upcall_info->key, user_skb);
468 469 470
	nla_nest_end(user_skb, nla);

	if (upcall_info->userdata)
471 472 473
		__nla_put(user_skb, OVS_PACKET_ATTR_USERDATA,
			  nla_len(upcall_info->userdata),
			  nla_data(upcall_info->userdata));
474

475 476 477 478 479 480 481
	/* 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);
482

483 484 485
	err = skb_zerocopy(user_skb, skb, skb->len, hlen);
	if (err)
		goto out;
486

487 488 489 490 491 492 493 494
	/* 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);
	}

495
	((struct nlmsghdr *) user_skb->data)->nlmsg_len = user_skb->len;
496

497
	err = genlmsg_unicast(ovs_dp_get_net(dp), user_skb, upcall_info->portid);
498
out:
499 500
	if (err)
		skb_tx_error(skb);
501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518
	kfree_skb(nskb);
	return err;
}

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

	err = -EINVAL;
	if (!a[OVS_PACKET_ATTR_PACKET] || !a[OVS_PACKET_ATTR_KEY] ||
519
	    !a[OVS_PACKET_ATTR_ACTIONS])
520 521 522 523 524 525 526 527 528
		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);

529
	nla_memcpy(__skb_put(packet, len), a[OVS_PACKET_ATTR_PACKET], len);
530 531 532 533 534 535 536

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

	/* Normally, setting the skb 'protocol' field would be handled by a
	 * call to eth_type_trans(), but it assumes there's a sending
	 * device, which we may not have. */
S
Simon Horman 已提交
537
	if (ntohs(eth->h_proto) >= ETH_P_802_3_MIN)
538 539 540 541 542
		packet->protocol = eth->h_proto;
	else
		packet->protocol = htons(ETH_P_802_2);

	/* Build an sw_flow for sending this packet. */
543
	flow = ovs_flow_alloc();
544 545 546 547
	err = PTR_ERR(flow);
	if (IS_ERR(flow))
		goto err_kfree_skb;

548
	err = ovs_flow_extract(packet, -1, &flow->key);
549 550 551
	if (err)
		goto err_flow_free;

552
	err = ovs_nla_get_flow_metadata(flow, a[OVS_PACKET_ATTR_KEY]);
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
	rcu_assign_pointer(flow->sf_acts, acts);
563 564
	if (err)
		goto err_flow_free;
565 566

	OVS_CB(packet)->flow = flow;
567
	OVS_CB(packet)->pkt_key = &flow->key;
568
	packet->priority = flow->key.phy.priority;
569
	packet->mark = flow->key.phy.skb_mark;
570 571

	rcu_read_lock();
572
	dp = get_dp(sock_net(skb->sk), ovs_header->dp_ifindex);
573 574 575 576 577 578 579 580 581
	err = -ENODEV;
	if (!dp)
		goto err_unlock;

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

582
	ovs_flow_free(flow, false);
583 584 585 586 587
	return err;

err_unlock:
	rcu_read_unlock();
err_flow_free:
588
	ovs_flow_free(flow, false);
589 590 591 592 593 594 595
err_kfree_skb:
	kfree_skb(packet);
err:
	return err;
}

static const struct nla_policy packet_policy[OVS_PACKET_ATTR_MAX + 1] = {
596
	[OVS_PACKET_ATTR_PACKET] = { .len = ETH_HLEN },
597 598 599 600
	[OVS_PACKET_ATTR_KEY] = { .type = NLA_NESTED },
	[OVS_PACKET_ATTR_ACTIONS] = { .type = NLA_NESTED },
};

601
static const struct genl_ops dp_packet_genl_ops[] = {
602 603 604 605 606 607 608
	{ .cmd = OVS_PACKET_CMD_EXECUTE,
	  .flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
	  .policy = packet_policy,
	  .doit = ovs_packet_cmd_execute
	}
};

609 610 611 612 613 614 615 616 617 618 619 620
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),
};

621 622
static void get_dp_stats(struct datapath *dp, struct ovs_dp_stats *stats,
			 struct ovs_dp_megaflow_stats *mega_stats)
623 624 625
{
	int i;

626 627
	memset(mega_stats, 0, sizeof(*mega_stats));

628
	stats->n_flows = ovs_flow_tbl_count(&dp->table);
629
	mega_stats->n_masks = ovs_flow_tbl_num_masks(&dp->table);
630 631

	stats->n_hit = stats->n_missed = stats->n_lost = 0;
632

633 634 635 636 637 638 639 640
	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 {
641
			start = u64_stats_fetch_begin_irq(&percpu_stats->syncp);
642
			local_stats = *percpu_stats;
643
		} while (u64_stats_fetch_retry_irq(&percpu_stats->syncp, start));
644 645 646 647

		stats->n_hit += local_stats.n_hit;
		stats->n_missed += local_stats.n_missed;
		stats->n_lost += local_stats.n_lost;
648
		mega_stats->n_mask_hit += local_stats.n_mask_hit;
649 650 651
	}
}

652 653 654 655
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 */
656
		+ nla_total_size(key_attr_size()) /* OVS_FLOW_ATTR_MASK */
657 658 659 660 661 662
		+ 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 */
}

J
Jarno Rajahalme 已提交
663
/* Called with ovs_mutex or RCU read lock. */
664
static int ovs_flow_cmd_fill_info(const struct sw_flow *flow, int dp_ifindex,
665
				  struct sk_buff *skb, u32 portid,
666 667 668
				  u32 seq, u32 flags, u8 cmd)
{
	const int skb_orig_len = skb->len;
669
	struct nlattr *start;
670
	struct ovs_flow_stats stats;
671 672
	__be16 tcp_flags;
	unsigned long used;
673 674 675 676
	struct ovs_header *ovs_header;
	struct nlattr *nla;
	int err;

677
	ovs_header = genlmsg_put(skb, portid, seq, &dp_flow_genl_family, flags, cmd);
678 679 680
	if (!ovs_header)
		return -EMSGSIZE;

681
	ovs_header->dp_ifindex = dp_ifindex;
682

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

688
	err = ovs_nla_put_flow(&flow->unmasked_key, &flow->unmasked_key, skb);
689 690 691 692 693 694 695 696
	if (err)
		goto error;
	nla_nest_end(skb, nla);

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

697
	err = ovs_nla_put_flow(&flow->key, &flow->mask->key, skb);
698 699
	if (err)
		goto error;
700

701 702
	nla_nest_end(skb, nla);

703
	ovs_flow_stats_get(flow, &stats, &used, &tcp_flags);
704

705 706 707
	if (used &&
	    nla_put_u64(skb, OVS_FLOW_ATTR_USED, ovs_flow_used_time(used)))
		goto nla_put_failure;
708

709
	if (stats.n_packets &&
710
	    nla_put(skb, OVS_FLOW_ATTR_STATS, sizeof(struct ovs_flow_stats), &stats))
711
		goto nla_put_failure;
712

713 714
	if ((u8)ntohs(tcp_flags) &&
	     nla_put_u8(skb, OVS_FLOW_ATTR_TCP_FLAGS, (u8)ntohs(tcp_flags)))
715
		goto nla_put_failure;
716 717 718 719 720 721 722 723 724 725 726

	/* 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.
	 */
727 728
	start = nla_nest_start(skb, OVS_FLOW_ATTR_ACTIONS);
	if (start) {
729 730
		const struct sw_flow_actions *sf_acts;

731
		sf_acts = rcu_dereference_ovsl(flow->sf_acts);
732 733
		err = ovs_nla_put_actions(sf_acts->actions,
					  sf_acts->actions_len, skb);
734

735 736 737 738 739 740 741 742 743 744
		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;
745 746 747 748 749 750 751 752 753 754

	return genlmsg_end(skb, ovs_header);

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

755 756
/* May not be called with RCU read lock. */
static struct sk_buff *ovs_flow_cmd_alloc_info(const struct sw_flow_actions *acts,
757 758
					       struct genl_info *info,
					       bool always)
759
{
760
	struct sk_buff *skb;
761

762 763 764
	if (!always && !ovs_must_notify(info, &ovs_dp_flow_multicast_group))
		return NULL;

765
	skb = genlmsg_new_unicast(ovs_flow_cmd_msg_size(acts), info, GFP_KERNEL);
766 767 768 769
	if (!skb)
		return ERR_PTR(-ENOMEM);

	return skb;
770 771
}

772 773 774 775 776
/* 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)
777 778 779 780
{
	struct sk_buff *skb;
	int retval;

781 782
	skb = ovs_flow_cmd_alloc_info(ovsl_dereference(flow->sf_acts), info,
				      always);
783 784
	if (!skb || IS_ERR(skb))
		return skb;
785

786 787 788
	retval = ovs_flow_cmd_fill_info(flow, dp_ifindex, skb,
					info->snd_portid, info->snd_seq, 0,
					cmd);
789 790 791 792
	BUG_ON(retval < 0);
	return skb;
}

793
static int ovs_flow_cmd_new(struct sk_buff *skb, struct genl_info *info)
794 795 796
{
	struct nlattr **a = info->attrs;
	struct ovs_header *ovs_header = info->userhdr;
797
	struct sw_flow *flow, *new_flow;
798
	struct sw_flow_mask mask;
799 800
	struct sk_buff *reply;
	struct datapath *dp;
801
	struct sw_flow_actions *acts;
802
	struct sw_flow_match match;
803 804
	int error;

805
	/* Must have key and actions. */
806 807 808
	error = -EINVAL;
	if (!a[OVS_FLOW_ATTR_KEY])
		goto error;
809 810
	if (!a[OVS_FLOW_ATTR_ACTIONS])
		goto error;
811

812 813 814 815 816 817 818 819 820 821 822
	/* 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);
823
	error = ovs_nla_get_match(&match,
824
				  a[OVS_FLOW_ATTR_KEY], a[OVS_FLOW_ATTR_MASK]);
825
	if (error)
826
		goto err_kfree_flow;
827

828
	ovs_flow_mask_key(&new_flow->key, &new_flow->unmasked_key, &mask);
829

830
	/* Validate actions. */
831 832 833
	acts = ovs_nla_alloc_flow_actions(nla_len(a[OVS_FLOW_ATTR_ACTIONS]));
	error = PTR_ERR(acts);
	if (IS_ERR(acts))
834
		goto err_kfree_flow;
835

836 837
	error = ovs_nla_copy_actions(a[OVS_FLOW_ATTR_ACTIONS], &new_flow->key,
				     0, &acts);
838 839
	if (error) {
		OVS_NLERR("Flow actions may not be safe on all matching packets.\n");
840 841 842 843 844 845 846
		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;
847 848
	}

849
	ovs_lock();
850
	dp = get_dp(sock_net(skb->sk), ovs_header->dp_ifindex);
851 852
	if (unlikely(!dp)) {
		error = -ENODEV;
853
		goto err_unlock_ovs;
854
	}
855
	/* Check if this is a duplicate flow */
856 857 858
	flow = ovs_flow_tbl_lookup(&dp->table, &new_flow->unmasked_key);
	if (likely(!flow)) {
		rcu_assign_pointer(new_flow->sf_acts, acts);
859 860

		/* Put flow in bucket. */
861 862
		error = ovs_flow_tbl_insert(&dp->table, new_flow, &mask);
		if (unlikely(error)) {
863
			acts = NULL;
864 865 866 867 868 869 870 871 872 873
			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);
874
		}
875
		ovs_unlock();
876
	} else {
877 878
		struct sw_flow_actions *old_acts;

879 880 881 882 883 884
		/* 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.
		 */
885 886 887
		if (unlikely(info->nlhdr->nlmsg_flags & (NLM_F_CREATE
							 | NLM_F_EXCL))) {
			error = -EEXIST;
888
			goto err_unlock_ovs;
889
		}
890
		/* The unmasked key has to be the same for flow updates. */
891
		if (unlikely(!ovs_flow_cmp_unmasked_key(flow, &match))) {
892 893 894 895 896
			flow = ovs_flow_tbl_lookup_exact(&dp->table, &match);
			if (!flow) {
				error = -ENOENT;
				goto err_unlock_ovs;
			}
897
		}
898 899 900 901
		/* Update actions. */
		old_acts = ovsl_dereference(flow->sf_acts);
		rcu_assign_pointer(flow->sf_acts, acts);

902 903 904 905 906 907 908 909 910
		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();
911

912 913
		ovs_nla_free_flow_actions(old_acts);
		ovs_flow_free(new_flow, false);
914
	}
915 916 917

	if (reply)
		ovs_notify(&dp_flow_genl_family, reply, info);
918 919 920 921
	return 0;

err_unlock_ovs:
	ovs_unlock();
922 923
	kfree_skb(reply);
err_kfree_acts:
924
	kfree(acts);
925 926
err_kfree_flow:
	ovs_flow_free(new_flow, false);
927 928 929
error:
	return error;
}
930

931 932 933 934 935 936 937 938 939
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;
	struct sw_flow_key key, masked_key;
	struct sw_flow *flow;
	struct sw_flow_mask mask;
	struct sk_buff *reply = NULL;
	struct datapath *dp;
940
	struct sw_flow_actions *old_acts = NULL, *acts = NULL;
941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966
	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]) {
		acts = ovs_nla_alloc_flow_actions(nla_len(a[OVS_FLOW_ATTR_ACTIONS]));
		error = PTR_ERR(acts);
		if (IS_ERR(acts))
			goto error;

		ovs_flow_mask_key(&masked_key, &key, &mask);
		error = ovs_nla_copy_actions(a[OVS_FLOW_ATTR_ACTIONS],
					     &masked_key, 0, &acts);
		if (error) {
			OVS_NLERR("Flow actions may not be safe on all matching packets.\n");
967 968 969 970 971 972 973 974 975 976
			goto err_kfree_acts;
		}
	}

	/* 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;
977
		}
978
	}
979

980 981
	ovs_lock();
	dp = get_dp(sock_net(skb->sk), ovs_header->dp_ifindex);
982 983
	if (unlikely(!dp)) {
		error = -ENODEV;
984
		goto err_unlock_ovs;
985
	}
986
	/* Check that the flow exists. */
987
	flow = ovs_flow_tbl_lookup_exact(&dp->table, &match);
988 989
	if (unlikely(!flow)) {
		error = -ENOENT;
990
		goto err_unlock_ovs;
991
	}
992

993
	/* Update actions, if present. */
994
	if (likely(acts)) {
995 996
		old_acts = ovsl_dereference(flow->sf_acts);
		rcu_assign_pointer(flow->sf_acts, acts);
997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013

		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);
		}
	} else {
		/* Could not alloc without acts before locking. */
		reply = ovs_flow_cmd_build_info(flow, ovs_header->dp_ifindex,
						info, OVS_FLOW_CMD_NEW, false);
		if (unlikely(IS_ERR(reply))) {
			error = PTR_ERR(reply);
			goto err_unlock_ovs;
		}
1014
	}
1015 1016 1017 1018

	/* Clear stats. */
	if (a[OVS_FLOW_ATTR_CLEAR])
		ovs_flow_stats_clear(flow);
1019
	ovs_unlock();
1020

1021 1022 1023 1024
	if (reply)
		ovs_notify(&dp_flow_genl_family, reply, info);
	if (old_acts)
		ovs_nla_free_flow_actions(old_acts);
1025

1026 1027
	return 0;

1028 1029
err_unlock_ovs:
	ovs_unlock();
1030 1031
	kfree_skb(reply);
err_kfree_acts:
1032
	kfree(acts);
1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044
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;
1045
	struct sw_flow_match match;
1046 1047
	int err;

1048 1049
	if (!a[OVS_FLOW_ATTR_KEY]) {
		OVS_NLERR("Flow get message rejected, Key attribute missing.\n");
1050
		return -EINVAL;
1051 1052 1053
	}

	ovs_match_init(&match, &key, NULL);
1054
	err = ovs_nla_get_match(&match, a[OVS_FLOW_ATTR_KEY], NULL);
1055 1056 1057
	if (err)
		return err;

1058
	ovs_lock();
1059
	dp = get_dp(sock_net(skb->sk), ovs_header->dp_ifindex);
1060 1061 1062 1063
	if (!dp) {
		err = -ENODEV;
		goto unlock;
	}
1064

1065 1066
	flow = ovs_flow_tbl_lookup_exact(&dp->table, &match);
	if (!flow) {
1067 1068 1069
		err = -ENOENT;
		goto unlock;
	}
1070

1071 1072
	reply = ovs_flow_cmd_build_info(flow, ovs_header->dp_ifindex, info,
					OVS_FLOW_CMD_NEW, true);
1073 1074 1075 1076
	if (IS_ERR(reply)) {
		err = PTR_ERR(reply);
		goto unlock;
	}
1077

1078
	ovs_unlock();
1079
	return genlmsg_reply(reply, info);
1080 1081 1082
unlock:
	ovs_unlock();
	return err;
1083 1084 1085 1086 1087 1088 1089 1090 1091 1092
}

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;
1093
	struct sw_flow_match match;
1094 1095
	int err;

1096 1097 1098 1099 1100 1101 1102
	if (likely(a[OVS_FLOW_ATTR_KEY])) {
		ovs_match_init(&match, &key, NULL);
		err = ovs_nla_get_match(&match, a[OVS_FLOW_ATTR_KEY], NULL);
		if (unlikely(err))
			return err;
	}

1103
	ovs_lock();
1104
	dp = get_dp(sock_net(skb->sk), ovs_header->dp_ifindex);
1105
	if (unlikely(!dp)) {
1106 1107 1108
		err = -ENODEV;
		goto unlock;
	}
1109

1110
	if (unlikely(!a[OVS_FLOW_ATTR_KEY])) {
1111
		err = ovs_flow_tbl_flush(&dp->table);
1112 1113
		goto unlock;
	}
1114

1115 1116
	flow = ovs_flow_tbl_lookup_exact(&dp->table, &match);
	if (unlikely(!flow)) {
1117 1118 1119
		err = -ENOENT;
		goto unlock;
	}
1120

1121
	ovs_flow_tbl_remove(&dp->table, flow);
1122
	ovs_unlock();
1123

1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139
	reply = ovs_flow_cmd_alloc_info((const struct sw_flow_actions __force *) flow->sf_acts,
					info, false);
	if (likely(reply)) {
		if (likely(!IS_ERR(reply))) {
			rcu_read_lock();	/*To keep RCU checker happy. */
			err = ovs_flow_cmd_fill_info(flow, ovs_header->dp_ifindex,
						     reply, info->snd_portid,
						     info->snd_seq, 0,
						     OVS_FLOW_CMD_DEL);
			rcu_read_unlock();
			BUG_ON(err < 0);

			ovs_notify(&dp_flow_genl_family, reply, info);
		} else {
			netlink_set_err(sock_net(skb->sk)->genl_sock, 0, 0, PTR_ERR(reply));
		}
1140
	}
1141

1142
	ovs_flow_free(flow, true);
1143
	return 0;
1144 1145 1146
unlock:
	ovs_unlock();
	return err;
1147 1148 1149 1150 1151
}

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));
1152
	struct table_instance *ti;
1153 1154
	struct datapath *dp;

1155
	rcu_read_lock();
1156
	dp = get_dp(sock_net(skb->sk), ovs_header->dp_ifindex);
1157
	if (!dp) {
1158
		rcu_read_unlock();
1159
		return -ENODEV;
1160
	}
1161

1162
	ti = rcu_dereference(dp->table.ti);
1163 1164 1165 1166 1167 1168
	for (;;) {
		struct sw_flow *flow;
		u32 bucket, obj;

		bucket = cb->args[0];
		obj = cb->args[1];
1169
		flow = ovs_flow_tbl_dump_next(ti, &bucket, &obj);
1170 1171 1172
		if (!flow)
			break;

1173
		if (ovs_flow_cmd_fill_info(flow, ovs_header->dp_ifindex, skb,
1174
					   NETLINK_CB(cb->skb).portid,
1175 1176 1177 1178 1179 1180 1181
					   cb->nlh->nlmsg_seq, NLM_F_MULTI,
					   OVS_FLOW_CMD_NEW) < 0)
			break;

		cb->args[0] = bucket;
		cb->args[1] = obj;
	}
1182
	rcu_read_unlock();
1183 1184 1185
	return skb->len;
}

1186 1187 1188 1189 1190 1191
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 },
};

1192
static const struct genl_ops dp_flow_genl_ops[] = {
1193 1194 1195
	{ .cmd = OVS_FLOW_CMD_NEW,
	  .flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
	  .policy = flow_policy,
1196
	  .doit = ovs_flow_cmd_new
1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211
	},
	{ .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,
1212
	  .doit = ovs_flow_cmd_set,
1213 1214 1215
	},
};

1216
static struct genl_family dp_flow_genl_family = {
1217 1218
	.id = GENL_ID_GENERATE,
	.hdrsize = sizeof(struct ovs_header),
1219 1220 1221
	.name = OVS_FLOW_FAMILY,
	.version = OVS_FLOW_VERSION,
	.maxattr = OVS_FLOW_ATTR_MAX,
1222 1223
	.netnsok = true,
	.parallel_ops = true,
1224 1225 1226 1227
	.ops = dp_flow_genl_ops,
	.n_ops = ARRAY_SIZE(dp_flow_genl_ops),
	.mcgrps = &ovs_dp_flow_multicast_group,
	.n_mcgrps = 1,
1228 1229
};

1230 1231 1232 1233 1234 1235
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));
1236
	msgsize += nla_total_size(sizeof(struct ovs_dp_megaflow_stats));
1237
	msgsize += nla_total_size(sizeof(u32)); /* OVS_DP_ATTR_USER_FEATURES */
1238 1239 1240 1241

	return msgsize;
}

J
Jarno Rajahalme 已提交
1242
/* Called with ovs_mutex or RCU read lock. */
1243
static int ovs_dp_cmd_fill_info(struct datapath *dp, struct sk_buff *skb,
1244
				u32 portid, u32 seq, u32 flags, u8 cmd)
1245 1246 1247
{
	struct ovs_header *ovs_header;
	struct ovs_dp_stats dp_stats;
1248
	struct ovs_dp_megaflow_stats dp_megaflow_stats;
1249 1250
	int err;

1251
	ovs_header = genlmsg_put(skb, portid, seq, &dp_datapath_genl_family,
1252 1253 1254 1255 1256 1257 1258 1259 1260 1261
				   flags, cmd);
	if (!ovs_header)
		goto error;

	ovs_header->dp_ifindex = get_dpifindex(dp);

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

1262 1263 1264 1265 1266 1267 1268 1269
	get_dp_stats(dp, &dp_stats, &dp_megaflow_stats);
	if (nla_put(skb, OVS_DP_ATTR_STATS, sizeof(struct ovs_dp_stats),
			&dp_stats))
		goto nla_put_failure;

	if (nla_put(skb, OVS_DP_ATTR_MEGAFLOW_STATS,
			sizeof(struct ovs_dp_megaflow_stats),
			&dp_megaflow_stats))
1270
		goto nla_put_failure;
1271

1272 1273 1274
	if (nla_put_u32(skb, OVS_DP_ATTR_USER_FEATURES, dp->user_features))
		goto nla_put_failure;

1275 1276 1277 1278 1279 1280 1281 1282
	return genlmsg_end(skb, ovs_header);

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

1283
static struct sk_buff *ovs_dp_cmd_alloc_info(struct genl_info *info)
1284
{
1285
	return genlmsg_new_unicast(ovs_dp_cmd_msg_size(), info, GFP_KERNEL);
1286 1287
}

J
Jarno Rajahalme 已提交
1288
/* Called with rcu_read_lock or ovs_mutex. */
1289 1290
static struct datapath *lookup_datapath(struct net *net,
					struct ovs_header *ovs_header,
1291 1292 1293 1294 1295
					struct nlattr *a[OVS_DP_ATTR_MAX + 1])
{
	struct datapath *dp;

	if (!a[OVS_DP_ATTR_NAME])
1296
		dp = get_dp(net, ovs_header->dp_ifindex);
1297 1298 1299
	else {
		struct vport *vport;

1300
		vport = ovs_vport_locate(net, nla_data(a[OVS_DP_ATTR_NAME]));
1301 1302 1303 1304 1305
		dp = vport && vport->port_no == OVSP_LOCAL ? vport->dp : NULL;
	}
	return dp ? dp : ERR_PTR(-ENODEV);
}

1306 1307 1308 1309 1310
static void ovs_dp_reset_user_features(struct sk_buff *skb, struct genl_info *info)
{
	struct datapath *dp;

	dp = lookup_datapath(sock_net(skb->sk), info->userhdr, info->attrs);
1311
	if (IS_ERR(dp))
1312 1313 1314 1315 1316 1317
		return;

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

1318 1319 1320 1321 1322 1323
static void ovs_dp_change(struct datapath *dp, struct nlattr **a)
{
	if (a[OVS_DP_ATTR_USER_FEATURES])
		dp->user_features = nla_get_u32(a[OVS_DP_ATTR_USER_FEATURES]);
}

1324 1325 1326 1327 1328 1329 1330
static int ovs_dp_cmd_new(struct sk_buff *skb, struct genl_info *info)
{
	struct nlattr **a = info->attrs;
	struct vport_parms parms;
	struct sk_buff *reply;
	struct datapath *dp;
	struct vport *vport;
1331
	struct ovs_net *ovs_net;
1332
	int err, i;
1333 1334 1335 1336 1337

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

1338 1339 1340
	reply = ovs_dp_cmd_alloc_info(info);
	if (!reply)
		return -ENOMEM;
1341 1342 1343 1344

	err = -ENOMEM;
	dp = kzalloc(sizeof(*dp), GFP_KERNEL);
	if (dp == NULL)
1345
		goto err_free_reply;
1346 1347

	ovs_dp_set_net(dp, hold_net(sock_net(skb->sk)));
1348 1349

	/* Allocate table. */
1350 1351
	err = ovs_flow_tbl_init(&dp->table);
	if (err)
1352 1353
		goto err_free_dp;

1354
	dp->stats_percpu = netdev_alloc_pcpu_stats(struct dp_stats_percpu);
1355 1356 1357 1358 1359
	if (!dp->stats_percpu) {
		err = -ENOMEM;
		goto err_destroy_table;
	}

1360
	dp->ports = kmalloc(DP_VPORT_HASH_BUCKETS * sizeof(struct hlist_head),
1361
			    GFP_KERNEL);
1362 1363 1364 1365 1366 1367 1368 1369
	if (!dp->ports) {
		err = -ENOMEM;
		goto err_destroy_percpu;
	}

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

1370 1371 1372 1373 1374 1375
	/* Set up our datapath device. */
	parms.name = nla_data(a[OVS_DP_ATTR_NAME]);
	parms.type = OVS_VPORT_TYPE_INTERNAL;
	parms.options = NULL;
	parms.dp = dp;
	parms.port_no = OVSP_LOCAL;
1376
	parms.upcall_portid = nla_get_u32(a[OVS_DP_ATTR_UPCALL_PID]);
1377

1378 1379
	ovs_dp_change(dp, a);

1380 1381 1382
	/* So far only local changes have been made, now need the lock. */
	ovs_lock();

1383 1384 1385 1386 1387 1388
	vport = new_vport(&parms);
	if (IS_ERR(vport)) {
		err = PTR_ERR(vport);
		if (err == -EBUSY)
			err = -EEXIST;

1389 1390 1391 1392 1393 1394 1395 1396 1397
		if (err == -EEXIST) {
			/* An outdated user space instance that does not understand
			 * the concept of user_features has attempted to create a new
			 * datapath and is likely to reuse it. Drop all user features.
			 */
			if (info->genlhdr->version < OVS_DP_VER_FEATURES)
				ovs_dp_reset_user_features(skb, info);
		}

1398
		goto err_destroy_ports_array;
1399 1400
	}

1401 1402 1403
	err = ovs_dp_cmd_fill_info(dp, reply, info->snd_portid,
				   info->snd_seq, 0, OVS_DP_CMD_NEW);
	BUG_ON(err < 0);
1404

1405
	ovs_net = net_generic(ovs_dp_get_net(dp), ovs_net_id);
1406
	list_add_tail_rcu(&dp->list_node, &ovs_net->dps);
1407 1408

	ovs_unlock();
1409

1410
	ovs_notify(&dp_datapath_genl_family, reply, info);
1411 1412
	return 0;

1413
err_destroy_ports_array:
1414
	ovs_unlock();
1415
	kfree(dp->ports);
1416 1417 1418
err_destroy_percpu:
	free_percpu(dp->stats_percpu);
err_destroy_table:
1419
	ovs_flow_tbl_destroy(&dp->table, false);
1420
err_free_dp:
1421
	release_net(ovs_dp_get_net(dp));
1422
	kfree(dp);
1423 1424
err_free_reply:
	kfree_skb(reply);
1425 1426 1427 1428
err:
	return err;
}

1429
/* Called with ovs_mutex. */
1430
static void __dp_destroy(struct datapath *dp)
1431
{
1432
	int i;
1433

1434 1435
	for (i = 0; i < DP_VPORT_HASH_BUCKETS; i++) {
		struct vport *vport;
1436
		struct hlist_node *n;
1437

1438
		hlist_for_each_entry_safe(vport, n, &dp->ports[i], dp_hash_node)
1439 1440 1441
			if (vport->port_no != OVSP_LOCAL)
				ovs_dp_detach_port(vport);
	}
1442

1443
	list_del_rcu(&dp->list_node);
1444

1445
	/* OVSP_LOCAL is datapath internal port. We need to make sure that
1446
	 * all ports in datapath are destroyed first before freeing datapath.
1447
	 */
1448
	ovs_dp_detach_port(ovs_vport_ovsl(dp, OVSP_LOCAL));
1449

1450 1451 1452
	/* RCU destroy the flow table */
	ovs_flow_tbl_destroy(&dp->table, true);

1453
	call_rcu(&dp->rcu, destroy_dp_rcu);
1454 1455 1456 1457 1458 1459 1460 1461
}

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

1462 1463 1464 1465
	reply = ovs_dp_cmd_alloc_info(info);
	if (!reply)
		return -ENOMEM;

1466
	ovs_lock();
1467 1468 1469
	dp = lookup_datapath(sock_net(skb->sk), info->userhdr, info->attrs);
	err = PTR_ERR(dp);
	if (IS_ERR(dp))
1470
		goto err_unlock_free;
1471

1472 1473 1474
	err = ovs_dp_cmd_fill_info(dp, reply, info->snd_portid,
				   info->snd_seq, 0, OVS_DP_CMD_DEL);
	BUG_ON(err < 0);
1475 1476

	__dp_destroy(dp);
1477
	ovs_unlock();
1478

1479
	ovs_notify(&dp_datapath_genl_family, reply, info);
1480 1481

	return 0;
1482 1483

err_unlock_free:
1484
	ovs_unlock();
1485
	kfree_skb(reply);
1486
	return err;
1487 1488 1489 1490 1491 1492 1493 1494
}

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

1495 1496 1497 1498
	reply = ovs_dp_cmd_alloc_info(info);
	if (!reply)
		return -ENOMEM;

1499
	ovs_lock();
1500
	dp = lookup_datapath(sock_net(skb->sk), info->userhdr, info->attrs);
1501
	err = PTR_ERR(dp);
1502
	if (IS_ERR(dp))
1503
		goto err_unlock_free;
1504

1505 1506
	ovs_dp_change(dp, info->attrs);

1507 1508 1509
	err = ovs_dp_cmd_fill_info(dp, reply, info->snd_portid,
				   info->snd_seq, 0, OVS_DP_CMD_NEW);
	BUG_ON(err < 0);
1510

1511
	ovs_unlock();
1512
	ovs_notify(&dp_datapath_genl_family, reply, info);
1513 1514

	return 0;
1515 1516

err_unlock_free:
1517
	ovs_unlock();
1518
	kfree_skb(reply);
1519
	return err;
1520 1521 1522 1523 1524 1525
}

static int ovs_dp_cmd_get(struct sk_buff *skb, struct genl_info *info)
{
	struct sk_buff *reply;
	struct datapath *dp;
1526
	int err;
1527

1528 1529 1530 1531 1532
	reply = ovs_dp_cmd_alloc_info(info);
	if (!reply)
		return -ENOMEM;

	rcu_read_lock();
1533
	dp = lookup_datapath(sock_net(skb->sk), info->userhdr, info->attrs);
1534 1535
	if (IS_ERR(dp)) {
		err = PTR_ERR(dp);
1536
		goto err_unlock_free;
1537
	}
1538 1539 1540 1541
	err = ovs_dp_cmd_fill_info(dp, reply, info->snd_portid,
				   info->snd_seq, 0, OVS_DP_CMD_NEW);
	BUG_ON(err < 0);
	rcu_read_unlock();
1542 1543

	return genlmsg_reply(reply, info);
1544

1545 1546 1547
err_unlock_free:
	rcu_read_unlock();
	kfree_skb(reply);
1548
	return err;
1549 1550 1551 1552
}

static int ovs_dp_cmd_dump(struct sk_buff *skb, struct netlink_callback *cb)
{
1553
	struct ovs_net *ovs_net = net_generic(sock_net(skb->sk), ovs_net_id);
1554 1555 1556 1557
	struct datapath *dp;
	int skip = cb->args[0];
	int i = 0;

1558 1559
	rcu_read_lock();
	list_for_each_entry_rcu(dp, &ovs_net->dps, list_node) {
1560
		if (i >= skip &&
1561
		    ovs_dp_cmd_fill_info(dp, skb, NETLINK_CB(cb->skb).portid,
1562 1563 1564 1565 1566
					 cb->nlh->nlmsg_seq, NLM_F_MULTI,
					 OVS_DP_CMD_NEW) < 0)
			break;
		i++;
	}
1567
	rcu_read_unlock();
1568 1569 1570 1571 1572 1573

	cb->args[0] = i;

	return skb->len;
}

1574 1575 1576 1577 1578 1579
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 },
	[OVS_DP_ATTR_USER_FEATURES] = { .type = NLA_U32 },
};

1580
static const struct genl_ops dp_datapath_genl_ops[] = {
1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603
	{ .cmd = OVS_DP_CMD_NEW,
	  .flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
	  .policy = datapath_policy,
	  .doit = ovs_dp_cmd_new
	},
	{ .cmd = OVS_DP_CMD_DEL,
	  .flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
	  .policy = datapath_policy,
	  .doit = ovs_dp_cmd_del
	},
	{ .cmd = OVS_DP_CMD_GET,
	  .flags = 0,		    /* OK for unprivileged users. */
	  .policy = datapath_policy,
	  .doit = ovs_dp_cmd_get,
	  .dumpit = ovs_dp_cmd_dump
	},
	{ .cmd = OVS_DP_CMD_SET,
	  .flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
	  .policy = datapath_policy,
	  .doit = ovs_dp_cmd_set,
	},
};

1604
static struct genl_family dp_datapath_genl_family = {
1605 1606
	.id = GENL_ID_GENERATE,
	.hdrsize = sizeof(struct ovs_header),
1607 1608 1609
	.name = OVS_DATAPATH_FAMILY,
	.version = OVS_DATAPATH_VERSION,
	.maxattr = OVS_DP_ATTR_MAX,
1610 1611
	.netnsok = true,
	.parallel_ops = true,
1612 1613 1614 1615
	.ops = dp_datapath_genl_ops,
	.n_ops = ARRAY_SIZE(dp_datapath_genl_ops),
	.mcgrps = &ovs_dp_datapath_multicast_group,
	.n_mcgrps = 1,
1616 1617
};

1618
/* Called with ovs_mutex or RCU read lock. */
1619
static int ovs_vport_cmd_fill_info(struct vport *vport, struct sk_buff *skb,
1620
				   u32 portid, u32 seq, u32 flags, u8 cmd)
1621 1622 1623 1624 1625
{
	struct ovs_header *ovs_header;
	struct ovs_vport_stats vport_stats;
	int err;

1626
	ovs_header = genlmsg_put(skb, portid, seq, &dp_vport_genl_family,
1627 1628 1629 1630 1631 1632
				 flags, cmd);
	if (!ovs_header)
		return -EMSGSIZE;

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

1633 1634 1635
	if (nla_put_u32(skb, OVS_VPORT_ATTR_PORT_NO, vport->port_no) ||
	    nla_put_u32(skb, OVS_VPORT_ATTR_TYPE, vport->ops->type) ||
	    nla_put_string(skb, OVS_VPORT_ATTR_NAME, vport->ops->get_name(vport)) ||
1636
	    nla_put_u32(skb, OVS_VPORT_ATTR_UPCALL_PID, vport->upcall_portid))
1637
		goto nla_put_failure;
1638 1639

	ovs_vport_get_stats(vport, &vport_stats);
1640 1641 1642
	if (nla_put(skb, OVS_VPORT_ATTR_STATS, sizeof(struct ovs_vport_stats),
		    &vport_stats))
		goto nla_put_failure;
1643 1644 1645 1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656

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

	return genlmsg_end(skb, ovs_header);

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

1657 1658 1659 1660 1661 1662
static struct sk_buff *ovs_vport_cmd_alloc_info(void)
{
	return nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
}

/* Called with ovs_mutex, only via ovs_dp_notify_wq(). */
1663
struct sk_buff *ovs_vport_cmd_build_info(struct vport *vport, u32 portid,
1664 1665 1666 1667 1668 1669 1670 1671 1672
					 u32 seq, u8 cmd)
{
	struct sk_buff *skb;
	int retval;

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

1673
	retval = ovs_vport_cmd_fill_info(vport, skb, portid, seq, 0, cmd);
1674 1675
	BUG_ON(retval < 0);

1676 1677 1678
	return skb;
}

1679
/* Called with ovs_mutex or RCU read lock. */
1680 1681
static struct vport *lookup_vport(struct net *net,
				  struct ovs_header *ovs_header,
1682 1683 1684 1685 1686 1687
				  struct nlattr *a[OVS_VPORT_ATTR_MAX + 1])
{
	struct datapath *dp;
	struct vport *vport;

	if (a[OVS_VPORT_ATTR_NAME]) {
1688
		vport = ovs_vport_locate(net, nla_data(a[OVS_VPORT_ATTR_NAME]));
1689 1690
		if (!vport)
			return ERR_PTR(-ENODEV);
1691 1692 1693
		if (ovs_header->dp_ifindex &&
		    ovs_header->dp_ifindex != get_dpifindex(vport->dp))
			return ERR_PTR(-ENODEV);
1694 1695 1696 1697 1698 1699 1700
		return vport;
	} else if (a[OVS_VPORT_ATTR_PORT_NO]) {
		u32 port_no = nla_get_u32(a[OVS_VPORT_ATTR_PORT_NO]);

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

1701
		dp = get_dp(net, ovs_header->dp_ifindex);
1702 1703 1704
		if (!dp)
			return ERR_PTR(-ENODEV);

1705
		vport = ovs_vport_ovsl_rcu(dp, port_no);
1706
		if (!vport)
1707
			return ERR_PTR(-ENODEV);
1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725
		return vport;
	} else
		return ERR_PTR(-EINVAL);
}

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

	if (!a[OVS_VPORT_ATTR_NAME] || !a[OVS_VPORT_ATTR_TYPE] ||
	    !a[OVS_VPORT_ATTR_UPCALL_PID])
1726 1727 1728 1729 1730 1731 1732 1733 1734 1735
		return -EINVAL;

	port_no = a[OVS_VPORT_ATTR_PORT_NO]
		? nla_get_u32(a[OVS_VPORT_ATTR_PORT_NO]) : 0;
	if (port_no >= DP_MAX_PORTS)
		return -EFBIG;

	reply = ovs_vport_cmd_alloc_info();
	if (!reply)
		return -ENOMEM;
1736

1737
	ovs_lock();
1738
	dp = get_dp(sock_net(skb->sk), ovs_header->dp_ifindex);
1739 1740
	err = -ENODEV;
	if (!dp)
1741
		goto exit_unlock_free;
1742

1743
	if (port_no) {
1744
		vport = ovs_vport_ovsl(dp, port_no);
1745 1746
		err = -EBUSY;
		if (vport)
1747
			goto exit_unlock_free;
1748 1749 1750 1751
	} else {
		for (port_no = 1; ; port_no++) {
			if (port_no >= DP_MAX_PORTS) {
				err = -EFBIG;
1752
				goto exit_unlock_free;
1753
			}
1754
			vport = ovs_vport_ovsl(dp, port_no);
1755 1756 1757 1758 1759 1760 1761 1762 1763 1764
			if (!vport)
				break;
		}
	}

	parms.name = nla_data(a[OVS_VPORT_ATTR_NAME]);
	parms.type = nla_get_u32(a[OVS_VPORT_ATTR_TYPE]);
	parms.options = a[OVS_VPORT_ATTR_OPTIONS];
	parms.dp = dp;
	parms.port_no = port_no;
1765
	parms.upcall_portid = nla_get_u32(a[OVS_VPORT_ATTR_UPCALL_PID]);
1766 1767 1768 1769

	vport = new_vport(&parms);
	err = PTR_ERR(vport);
	if (IS_ERR(vport))
1770
		goto exit_unlock_free;
1771

1772 1773 1774 1775
	err = ovs_vport_cmd_fill_info(vport, reply, info->snd_portid,
				      info->snd_seq, 0, OVS_VPORT_CMD_NEW);
	BUG_ON(err < 0);
	ovs_unlock();
1776

1777
	ovs_notify(&dp_vport_genl_family, reply, info);
1778
	return 0;
1779

1780
exit_unlock_free:
1781
	ovs_unlock();
1782
	kfree_skb(reply);
1783 1784 1785 1786 1787 1788 1789 1790 1791 1792
	return err;
}

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

1793 1794 1795 1796
	reply = ovs_vport_cmd_alloc_info();
	if (!reply)
		return -ENOMEM;

1797
	ovs_lock();
1798
	vport = lookup_vport(sock_net(skb->sk), info->userhdr, a);
1799 1800
	err = PTR_ERR(vport);
	if (IS_ERR(vport))
1801
		goto exit_unlock_free;
1802 1803

	if (a[OVS_VPORT_ATTR_TYPE] &&
1804
	    nla_get_u32(a[OVS_VPORT_ATTR_TYPE]) != vport->ops->type) {
1805
		err = -EINVAL;
1806
		goto exit_unlock_free;
1807 1808
	}

1809
	if (a[OVS_VPORT_ATTR_OPTIONS]) {
1810
		err = ovs_vport_set_options(vport, a[OVS_VPORT_ATTR_OPTIONS]);
1811
		if (err)
1812
			goto exit_unlock_free;
1813
	}
1814

1815
	if (a[OVS_VPORT_ATTR_UPCALL_PID])
1816
		vport->upcall_portid = nla_get_u32(a[OVS_VPORT_ATTR_UPCALL_PID]);
1817

1818 1819 1820
	err = ovs_vport_cmd_fill_info(vport, reply, info->snd_portid,
				      info->snd_seq, 0, OVS_VPORT_CMD_NEW);
	BUG_ON(err < 0);
1821

1822
	ovs_unlock();
1823
	ovs_notify(&dp_vport_genl_family, reply, info);
1824
	return 0;
1825

1826
exit_unlock_free:
1827
	ovs_unlock();
1828
	kfree_skb(reply);
1829 1830 1831 1832 1833 1834 1835 1836 1837 1838
	return err;
}

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

1839 1840 1841 1842
	reply = ovs_vport_cmd_alloc_info();
	if (!reply)
		return -ENOMEM;

1843
	ovs_lock();
1844
	vport = lookup_vport(sock_net(skb->sk), info->userhdr, a);
1845 1846
	err = PTR_ERR(vport);
	if (IS_ERR(vport))
1847
		goto exit_unlock_free;
1848 1849 1850

	if (vport->port_no == OVSP_LOCAL) {
		err = -EINVAL;
1851
		goto exit_unlock_free;
1852 1853
	}

1854 1855 1856
	err = ovs_vport_cmd_fill_info(vport, reply, info->snd_portid,
				      info->snd_seq, 0, OVS_VPORT_CMD_DEL);
	BUG_ON(err < 0);
1857
	ovs_dp_detach_port(vport);
1858
	ovs_unlock();
1859

1860
	ovs_notify(&dp_vport_genl_family, reply, info);
1861
	return 0;
1862

1863
exit_unlock_free:
1864
	ovs_unlock();
1865
	kfree_skb(reply);
1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876
	return err;
}

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

1877 1878 1879 1880
	reply = ovs_vport_cmd_alloc_info();
	if (!reply)
		return -ENOMEM;

1881
	rcu_read_lock();
1882
	vport = lookup_vport(sock_net(skb->sk), ovs_header, a);
1883 1884
	err = PTR_ERR(vport);
	if (IS_ERR(vport))
1885 1886 1887 1888
		goto exit_unlock_free;
	err = ovs_vport_cmd_fill_info(vport, reply, info->snd_portid,
				      info->snd_seq, 0, OVS_VPORT_CMD_NEW);
	BUG_ON(err < 0);
1889 1890 1891 1892
	rcu_read_unlock();

	return genlmsg_reply(reply, info);

1893
exit_unlock_free:
1894
	rcu_read_unlock();
1895
	kfree_skb(reply);
1896 1897 1898 1899 1900 1901 1902
	return err;
}

static int ovs_vport_cmd_dump(struct sk_buff *skb, struct netlink_callback *cb)
{
	struct ovs_header *ovs_header = genlmsg_data(nlmsg_data(cb->nlh));
	struct datapath *dp;
1903 1904
	int bucket = cb->args[0], skip = cb->args[1];
	int i, j = 0;
1905

J
Jarno Rajahalme 已提交
1906
	rcu_read_lock();
1907
	dp = get_dp(sock_net(skb->sk), ovs_header->dp_ifindex);
J
Jarno Rajahalme 已提交
1908 1909
	if (!dp) {
		rcu_read_unlock();
1910
		return -ENODEV;
J
Jarno Rajahalme 已提交
1911
	}
1912
	for (i = bucket; i < DP_VPORT_HASH_BUCKETS; i++) {
1913
		struct vport *vport;
1914 1915

		j = 0;
1916
		hlist_for_each_entry_rcu(vport, &dp->ports[i], dp_hash_node) {
1917 1918
			if (j >= skip &&
			    ovs_vport_cmd_fill_info(vport, skb,
1919
						    NETLINK_CB(cb->skb).portid,
1920 1921 1922 1923 1924 1925 1926 1927
						    cb->nlh->nlmsg_seq,
						    NLM_F_MULTI,
						    OVS_VPORT_CMD_NEW) < 0)
				goto out;

			j++;
		}
		skip = 0;
1928
	}
1929
out:
1930 1931
	rcu_read_unlock();

1932 1933
	cb->args[0] = i;
	cb->args[1] = j;
1934

1935
	return skb->len;
1936 1937
}

1938 1939 1940 1941 1942 1943 1944 1945 1946
static const struct nla_policy vport_policy[OVS_VPORT_ATTR_MAX + 1] = {
	[OVS_VPORT_ATTR_NAME] = { .type = NLA_NUL_STRING, .len = IFNAMSIZ - 1 },
	[OVS_VPORT_ATTR_STATS] = { .len = sizeof(struct ovs_vport_stats) },
	[OVS_VPORT_ATTR_PORT_NO] = { .type = NLA_U32 },
	[OVS_VPORT_ATTR_TYPE] = { .type = NLA_U32 },
	[OVS_VPORT_ATTR_UPCALL_PID] = { .type = NLA_U32 },
	[OVS_VPORT_ATTR_OPTIONS] = { .type = NLA_NESTED },
};

1947
static const struct genl_ops dp_vport_genl_ops[] = {
1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970
	{ .cmd = OVS_VPORT_CMD_NEW,
	  .flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
	  .policy = vport_policy,
	  .doit = ovs_vport_cmd_new
	},
	{ .cmd = OVS_VPORT_CMD_DEL,
	  .flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
	  .policy = vport_policy,
	  .doit = ovs_vport_cmd_del
	},
	{ .cmd = OVS_VPORT_CMD_GET,
	  .flags = 0,		    /* OK for unprivileged users. */
	  .policy = vport_policy,
	  .doit = ovs_vport_cmd_get,
	  .dumpit = ovs_vport_cmd_dump
	},
	{ .cmd = OVS_VPORT_CMD_SET,
	  .flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
	  .policy = vport_policy,
	  .doit = ovs_vport_cmd_set,
	},
};

1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982
struct genl_family dp_vport_genl_family = {
	.id = GENL_ID_GENERATE,
	.hdrsize = sizeof(struct ovs_header),
	.name = OVS_VPORT_FAMILY,
	.version = OVS_VPORT_VERSION,
	.maxattr = OVS_VPORT_ATTR_MAX,
	.netnsok = true,
	.parallel_ops = true,
	.ops = dp_vport_genl_ops,
	.n_ops = ARRAY_SIZE(dp_vport_genl_ops),
	.mcgrps = &ovs_dp_vport_multicast_group,
	.n_mcgrps = 1,
1983 1984
};

1985 1986 1987 1988 1989
static struct genl_family * const dp_genl_families[] = {
	&dp_datapath_genl_family,
	&dp_vport_genl_family,
	&dp_flow_genl_family,
	&dp_packet_genl_family,
1990 1991 1992 1993 1994 1995 1996
};

static void dp_unregister_genl(int n_families)
{
	int i;

	for (i = 0; i < n_families; i++)
1997
		genl_unregister_family(dp_genl_families[i]);
1998 1999 2000 2001 2002 2003 2004 2005 2006
}

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

	for (i = 0; i < ARRAY_SIZE(dp_genl_families); i++) {

2007
		err = genl_register_family(dp_genl_families[i]);
2008 2009 2010 2011 2012 2013 2014
		if (err)
			goto error;
	}

	return 0;

error:
2015
	dp_unregister_genl(i);
2016 2017 2018
	return err;
}

2019 2020 2021 2022 2023
static int __net_init ovs_init_net(struct net *net)
{
	struct ovs_net *ovs_net = net_generic(net, ovs_net_id);

	INIT_LIST_HEAD(&ovs_net->dps);
2024
	INIT_WORK(&ovs_net->dp_notify_work, ovs_dp_notify_wq);
2025 2026 2027 2028 2029 2030
	return 0;
}

static void __net_exit ovs_exit_net(struct net *net)
{
	struct datapath *dp, *dp_next;
2031
	struct ovs_net *ovs_net = net_generic(net, ovs_net_id);
2032

2033
	ovs_lock();
2034 2035
	list_for_each_entry_safe(dp, dp_next, &ovs_net->dps, list_node)
		__dp_destroy(dp);
2036 2037 2038
	ovs_unlock();

	cancel_work_sync(&ovs_net->dp_notify_work);
2039 2040 2041 2042 2043 2044 2045 2046 2047
}

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

2048 2049 2050 2051
static int __init dp_init(void)
{
	int err;

2052
	BUILD_BUG_ON(sizeof(struct ovs_skb_cb) > FIELD_SIZEOF(struct sk_buff, cb));
2053 2054 2055

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

2056
	err = ovs_internal_dev_rtnl_link_register();
2057 2058 2059
	if (err)
		goto error;

2060 2061 2062 2063
	err = ovs_flow_init();
	if (err)
		goto error_unreg_rtnl_link;

2064 2065 2066 2067
	err = ovs_vport_init();
	if (err)
		goto error_flow_exit;

2068
	err = register_pernet_device(&ovs_net_ops);
2069 2070 2071
	if (err)
		goto error_vport_exit;

2072 2073 2074 2075
	err = register_netdevice_notifier(&ovs_dp_device_notifier);
	if (err)
		goto error_netns_exit;

2076 2077 2078 2079 2080 2081 2082 2083
	err = dp_register_genl();
	if (err < 0)
		goto error_unreg_notifier;

	return 0;

error_unreg_notifier:
	unregister_netdevice_notifier(&ovs_dp_device_notifier);
2084 2085
error_netns_exit:
	unregister_pernet_device(&ovs_net_ops);
2086 2087 2088 2089
error_vport_exit:
	ovs_vport_exit();
error_flow_exit:
	ovs_flow_exit();
2090 2091
error_unreg_rtnl_link:
	ovs_internal_dev_rtnl_link_unregister();
2092 2093 2094 2095 2096 2097 2098 2099
error:
	return err;
}

static void dp_cleanup(void)
{
	dp_unregister_genl(ARRAY_SIZE(dp_genl_families));
	unregister_netdevice_notifier(&ovs_dp_device_notifier);
2100 2101
	unregister_pernet_device(&ovs_net_ops);
	rcu_barrier();
2102 2103
	ovs_vport_exit();
	ovs_flow_exit();
2104
	ovs_internal_dev_rtnl_link_unregister();
2105 2106 2107 2108 2109 2110 2111
}

module_init(dp_init);
module_exit(dp_cleanup);

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