datapath.c 52.1 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)];
}

J
Jarno Rajahalme 已提交
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;
L
Li RongQing 已提交
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);
L
Li RongQing 已提交
500
	user_skb = NULL;
501
out:
502 503
	if (err)
		skb_tx_error(skb);
L
Li RongQing 已提交
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. */
S
Simon Horman 已提交
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 */
}

J
Jarno Rajahalme 已提交
676
/* Called with ovs_mutex or RCU read lock. */
677
static int ovs_flow_cmd_fill_info(const struct sw_flow *flow, int dp_ifindex,
678
				  struct sk_buff *skb, u32 portid,
679 680 681
				  u32 seq, u32 flags, u8 cmd)
{
	const int skb_orig_len = skb->len;
682
	struct nlattr *start;
683
	struct ovs_flow_stats stats;
684 685
	__be16 tcp_flags;
	unsigned long used;
686 687 688 689
	struct ovs_header *ovs_header;
	struct nlattr *nla;
	int err;

690
	ovs_header = genlmsg_put(skb, portid, seq, &dp_flow_genl_family, flags, cmd);
691 692 693
	if (!ovs_header)
		return -EMSGSIZE;

694
	ovs_header->dp_ifindex = dp_ifindex;
695

696
	/* Fill flow key. */
697 698 699
	nla = nla_nest_start(skb, OVS_FLOW_ATTR_KEY);
	if (!nla)
		goto nla_put_failure;
700

701
	err = ovs_nla_put_flow(&flow->unmasked_key, &flow->unmasked_key, skb);
702 703 704 705 706 707 708 709
	if (err)
		goto error;
	nla_nest_end(skb, nla);

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

710
	err = ovs_nla_put_flow(&flow->key, &flow->mask->key, skb);
711 712
	if (err)
		goto error;
713

714 715
	nla_nest_end(skb, nla);

716
	ovs_flow_stats_get(flow, &stats, &used, &tcp_flags);
717

718 719 720
	if (used &&
	    nla_put_u64(skb, OVS_FLOW_ATTR_USED, ovs_flow_used_time(used)))
		goto nla_put_failure;
721

722
	if (stats.n_packets &&
723
	    nla_put(skb, OVS_FLOW_ATTR_STATS, sizeof(struct ovs_flow_stats), &stats))
724
		goto nla_put_failure;
725

726 727
	if ((u8)ntohs(tcp_flags) &&
	     nla_put_u8(skb, OVS_FLOW_ATTR_TCP_FLAGS, (u8)ntohs(tcp_flags)))
728
		goto nla_put_failure;
729 730 731 732 733 734 735 736 737 738 739

	/* 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.
	 */
740 741
	start = nla_nest_start(skb, OVS_FLOW_ATTR_ACTIONS);
	if (start) {
742 743
		const struct sw_flow_actions *sf_acts;

744
		sf_acts = rcu_dereference_ovsl(flow->sf_acts);
745 746
		err = ovs_nla_put_actions(sf_acts->actions,
					  sf_acts->actions_len, skb);
747

748 749 750 751 752 753 754 755 756 757
		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;
758 759 760 761 762 763 764 765 766 767

	return genlmsg_end(skb, ovs_header);

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

768 769
/* May not be called with RCU read lock. */
static struct sk_buff *ovs_flow_cmd_alloc_info(const struct sw_flow_actions *acts,
770 771
					       struct genl_info *info,
					       bool always)
772
{
773
	struct sk_buff *skb;
774

775
	if (!always && !ovs_must_notify(&dp_flow_genl_family, info, 0))
776 777
		return NULL;

778
	skb = genlmsg_new_unicast(ovs_flow_cmd_msg_size(acts), info, GFP_KERNEL);
779 780 781 782
	if (!skb)
		return ERR_PTR(-ENOMEM);

	return skb;
783 784
}

785 786 787 788 789
/* 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)
790 791 792 793
{
	struct sk_buff *skb;
	int retval;

794 795
	skb = ovs_flow_cmd_alloc_info(ovsl_dereference(flow->sf_acts), info,
				      always);
796
	if (IS_ERR_OR_NULL(skb))
797
		return skb;
798

799 800 801
	retval = ovs_flow_cmd_fill_info(flow, dp_ifindex, skb,
					info->snd_portid, info->snd_seq, 0,
					cmd);
802 803 804 805
	BUG_ON(retval < 0);
	return skb;
}

806
static int ovs_flow_cmd_new(struct sk_buff *skb, struct genl_info *info)
807 808 809
{
	struct nlattr **a = info->attrs;
	struct ovs_header *ovs_header = info->userhdr;
810
	struct sw_flow *flow, *new_flow;
811
	struct sw_flow_mask mask;
812 813
	struct sk_buff *reply;
	struct datapath *dp;
814
	struct sw_flow_actions *acts;
815
	struct sw_flow_match match;
816 817
	int error;

818
	/* Must have key and actions. */
819
	error = -EINVAL;
820 821
	if (!a[OVS_FLOW_ATTR_KEY]) {
		OVS_NLERR("Flow key attribute not present in new flow.\n");
822
		goto error;
823 824 825
	}
	if (!a[OVS_FLOW_ATTR_ACTIONS]) {
		OVS_NLERR("Flow actions attribute not present in new flow.\n");
826
		goto error;
827
	}
828

829 830 831 832 833 834 835 836 837 838 839
	/* 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);
840
	error = ovs_nla_get_match(&match,
841
				  a[OVS_FLOW_ATTR_KEY], a[OVS_FLOW_ATTR_MASK]);
842
	if (error)
843
		goto err_kfree_flow;
844

845
	ovs_flow_mask_key(&new_flow->key, &new_flow->unmasked_key, &mask);
846

847
	/* Validate actions. */
848 849 850
	acts = ovs_nla_alloc_flow_actions(nla_len(a[OVS_FLOW_ATTR_ACTIONS]));
	error = PTR_ERR(acts);
	if (IS_ERR(acts))
851
		goto err_kfree_flow;
852

853
	error = ovs_nla_copy_actions(a[OVS_FLOW_ATTR_ACTIONS], &new_flow->key,
854
				     &acts);
855 856
	if (error) {
		OVS_NLERR("Flow actions may not be safe on all matching packets.\n");
857 858 859 860 861 862 863
		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;
864 865
	}

866
	ovs_lock();
867
	dp = get_dp(sock_net(skb->sk), ovs_header->dp_ifindex);
868 869
	if (unlikely(!dp)) {
		error = -ENODEV;
870
		goto err_unlock_ovs;
871
	}
872
	/* Check if this is a duplicate flow */
873 874 875
	flow = ovs_flow_tbl_lookup(&dp->table, &new_flow->unmasked_key);
	if (likely(!flow)) {
		rcu_assign_pointer(new_flow->sf_acts, acts);
876 877

		/* Put flow in bucket. */
878 879
		error = ovs_flow_tbl_insert(&dp->table, new_flow, &mask);
		if (unlikely(error)) {
880
			acts = NULL;
881 882 883 884 885 886 887 888 889 890
			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);
891
		}
892
		ovs_unlock();
893
	} else {
894 895
		struct sw_flow_actions *old_acts;

896 897 898 899 900 901
		/* 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.
		 */
902 903 904
		if (unlikely(info->nlhdr->nlmsg_flags & (NLM_F_CREATE
							 | NLM_F_EXCL))) {
			error = -EEXIST;
905
			goto err_unlock_ovs;
906
		}
907
		/* The unmasked key has to be the same for flow updates. */
908
		if (unlikely(!ovs_flow_cmp_unmasked_key(flow, &match))) {
909 910 911 912 913
			flow = ovs_flow_tbl_lookup_exact(&dp->table, &match);
			if (!flow) {
				error = -ENOENT;
				goto err_unlock_ovs;
			}
914
		}
915 916 917 918
		/* Update actions. */
		old_acts = ovsl_dereference(flow->sf_acts);
		rcu_assign_pointer(flow->sf_acts, acts);

919 920 921 922 923 924 925 926 927
		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();
928

929 930
		ovs_nla_free_flow_actions(old_acts);
		ovs_flow_free(new_flow, false);
931
	}
932 933 934

	if (reply)
		ovs_notify(&dp_flow_genl_family, reply, info);
935 936 937 938
	return 0;

err_unlock_ovs:
	ovs_unlock();
939 940
	kfree_skb(reply);
err_kfree_acts:
941
	kfree(acts);
942 943
err_kfree_flow:
	ovs_flow_free(new_flow, false);
944 945 946
error:
	return error;
}
947

948 949 950 951 952 953 954 955 956 957 958 959 960
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);
961
	error = ovs_nla_copy_actions(a, &masked_key, &acts);
962 963 964 965 966 967 968 969 970
	if (error) {
		OVS_NLERR("Flow actions may not be safe on all matching packets.\n");
		kfree(acts);
		return ERR_PTR(error);
	}

	return acts;
}

971 972 973 974
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;
975
	struct sw_flow_key key;
976 977 978 979
	struct sw_flow *flow;
	struct sw_flow_mask mask;
	struct sk_buff *reply = NULL;
	struct datapath *dp;
980
	struct sw_flow_actions *old_acts = NULL, *acts = NULL;
981 982 983 984 985
	struct sw_flow_match match;
	int error;

	/* Extract key. */
	error = -EINVAL;
986 987
	if (!a[OVS_FLOW_ATTR_KEY]) {
		OVS_NLERR("Flow key attribute not present in set flow.\n");
988
		goto error;
989
	}
990 991 992 993 994 995 996 997 998

	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]) {
999 1000 1001
		acts = get_flow_actions(a[OVS_FLOW_ATTR_ACTIONS], &key, &mask);
		if (IS_ERR(acts)) {
			error = PTR_ERR(acts);
1002
			goto error;
1003 1004 1005 1006 1007 1008 1009 1010 1011
		}
	}

	/* 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;
1012
		}
1013
	}
1014

1015 1016
	ovs_lock();
	dp = get_dp(sock_net(skb->sk), ovs_header->dp_ifindex);
1017 1018
	if (unlikely(!dp)) {
		error = -ENODEV;
1019
		goto err_unlock_ovs;
1020
	}
1021
	/* Check that the flow exists. */
1022
	flow = ovs_flow_tbl_lookup_exact(&dp->table, &match);
1023 1024
	if (unlikely(!flow)) {
		error = -ENOENT;
1025
		goto err_unlock_ovs;
1026
	}
1027

1028
	/* Update actions, if present. */
1029
	if (likely(acts)) {
1030 1031
		old_acts = ovsl_dereference(flow->sf_acts);
		rcu_assign_pointer(flow->sf_acts, acts);
1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048

		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;
		}
1049
	}
1050 1051 1052 1053

	/* Clear stats. */
	if (a[OVS_FLOW_ATTR_CLEAR])
		ovs_flow_stats_clear(flow);
1054
	ovs_unlock();
1055

1056 1057 1058 1059
	if (reply)
		ovs_notify(&dp_flow_genl_family, reply, info);
	if (old_acts)
		ovs_nla_free_flow_actions(old_acts);
1060

1061 1062
	return 0;

1063 1064
err_unlock_ovs:
	ovs_unlock();
1065 1066
	kfree_skb(reply);
err_kfree_acts:
1067
	kfree(acts);
1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079
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;
1080
	struct sw_flow_match match;
1081 1082
	int err;

1083 1084
	if (!a[OVS_FLOW_ATTR_KEY]) {
		OVS_NLERR("Flow get message rejected, Key attribute missing.\n");
1085
		return -EINVAL;
1086 1087 1088
	}

	ovs_match_init(&match, &key, NULL);
1089
	err = ovs_nla_get_match(&match, a[OVS_FLOW_ATTR_KEY], NULL);
1090 1091 1092
	if (err)
		return err;

1093
	ovs_lock();
1094
	dp = get_dp(sock_net(skb->sk), ovs_header->dp_ifindex);
1095 1096 1097 1098
	if (!dp) {
		err = -ENODEV;
		goto unlock;
	}
1099

1100 1101
	flow = ovs_flow_tbl_lookup_exact(&dp->table, &match);
	if (!flow) {
1102 1103 1104
		err = -ENOENT;
		goto unlock;
	}
1105

1106 1107
	reply = ovs_flow_cmd_build_info(flow, ovs_header->dp_ifindex, info,
					OVS_FLOW_CMD_NEW, true);
1108 1109 1110 1111
	if (IS_ERR(reply)) {
		err = PTR_ERR(reply);
		goto unlock;
	}
1112

1113
	ovs_unlock();
1114
	return genlmsg_reply(reply, info);
1115 1116 1117
unlock:
	ovs_unlock();
	return err;
1118 1119 1120 1121 1122 1123 1124 1125 1126 1127
}

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;
1128
	struct sw_flow_match match;
1129 1130
	int err;

1131 1132 1133 1134 1135 1136 1137
	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;
	}

1138
	ovs_lock();
1139
	dp = get_dp(sock_net(skb->sk), ovs_header->dp_ifindex);
1140
	if (unlikely(!dp)) {
1141 1142 1143
		err = -ENODEV;
		goto unlock;
	}
1144

1145
	if (unlikely(!a[OVS_FLOW_ATTR_KEY])) {
1146
		err = ovs_flow_tbl_flush(&dp->table);
1147 1148
		goto unlock;
	}
1149

1150 1151
	flow = ovs_flow_tbl_lookup_exact(&dp->table, &match);
	if (unlikely(!flow)) {
1152 1153 1154
		err = -ENOENT;
		goto unlock;
	}
1155

1156
	ovs_flow_tbl_remove(&dp->table, flow);
1157
	ovs_unlock();
1158

1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174
	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));
		}
1175
	}
1176

1177
	ovs_flow_free(flow, true);
1178
	return 0;
1179 1180 1181
unlock:
	ovs_unlock();
	return err;
1182 1183 1184 1185 1186
}

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));
1187
	struct table_instance *ti;
1188 1189
	struct datapath *dp;

1190
	rcu_read_lock();
1191
	dp = get_dp(sock_net(skb->sk), ovs_header->dp_ifindex);
1192
	if (!dp) {
1193
		rcu_read_unlock();
1194
		return -ENODEV;
1195
	}
1196

1197
	ti = rcu_dereference(dp->table.ti);
1198 1199 1200 1201 1202 1203
	for (;;) {
		struct sw_flow *flow;
		u32 bucket, obj;

		bucket = cb->args[0];
		obj = cb->args[1];
1204
		flow = ovs_flow_tbl_dump_next(ti, &bucket, &obj);
1205 1206 1207
		if (!flow)
			break;

1208
		if (ovs_flow_cmd_fill_info(flow, ovs_header->dp_ifindex, skb,
1209
					   NETLINK_CB(cb->skb).portid,
1210 1211 1212 1213 1214 1215 1216
					   cb->nlh->nlmsg_seq, NLM_F_MULTI,
					   OVS_FLOW_CMD_NEW) < 0)
			break;

		cb->args[0] = bucket;
		cb->args[1] = obj;
	}
1217
	rcu_read_unlock();
1218 1219 1220
	return skb->len;
}

1221 1222 1223 1224 1225 1226
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 },
};

1227
static const struct genl_ops dp_flow_genl_ops[] = {
1228 1229 1230
	{ .cmd = OVS_FLOW_CMD_NEW,
	  .flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
	  .policy = flow_policy,
1231
	  .doit = ovs_flow_cmd_new
1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246
	},
	{ .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,
1247
	  .doit = ovs_flow_cmd_set,
1248 1249 1250
	},
};

1251
static struct genl_family dp_flow_genl_family = {
1252 1253
	.id = GENL_ID_GENERATE,
	.hdrsize = sizeof(struct ovs_header),
1254 1255 1256
	.name = OVS_FLOW_FAMILY,
	.version = OVS_FLOW_VERSION,
	.maxattr = OVS_FLOW_ATTR_MAX,
1257 1258
	.netnsok = true,
	.parallel_ops = true,
1259 1260 1261 1262
	.ops = dp_flow_genl_ops,
	.n_ops = ARRAY_SIZE(dp_flow_genl_ops),
	.mcgrps = &ovs_dp_flow_multicast_group,
	.n_mcgrps = 1,
1263 1264
};

1265 1266 1267 1268 1269 1270
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));
1271
	msgsize += nla_total_size(sizeof(struct ovs_dp_megaflow_stats));
1272
	msgsize += nla_total_size(sizeof(u32)); /* OVS_DP_ATTR_USER_FEATURES */
1273 1274 1275 1276

	return msgsize;
}

J
Jarno Rajahalme 已提交
1277
/* Called with ovs_mutex or RCU read lock. */
1278
static int ovs_dp_cmd_fill_info(struct datapath *dp, struct sk_buff *skb,
1279
				u32 portid, u32 seq, u32 flags, u8 cmd)
1280 1281 1282
{
	struct ovs_header *ovs_header;
	struct ovs_dp_stats dp_stats;
1283
	struct ovs_dp_megaflow_stats dp_megaflow_stats;
1284 1285
	int err;

1286
	ovs_header = genlmsg_put(skb, portid, seq, &dp_datapath_genl_family,
1287 1288 1289 1290 1291 1292 1293 1294 1295 1296
				   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;

1297 1298 1299 1300 1301 1302 1303 1304
	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))
1305
		goto nla_put_failure;
1306

1307 1308 1309
	if (nla_put_u32(skb, OVS_DP_ATTR_USER_FEATURES, dp->user_features))
		goto nla_put_failure;

1310 1311 1312 1313 1314 1315 1316 1317
	return genlmsg_end(skb, ovs_header);

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

1318
static struct sk_buff *ovs_dp_cmd_alloc_info(struct genl_info *info)
1319
{
1320
	return genlmsg_new_unicast(ovs_dp_cmd_msg_size(), info, GFP_KERNEL);
1321 1322
}

J
Jarno Rajahalme 已提交
1323
/* Called with rcu_read_lock or ovs_mutex. */
1324 1325
static struct datapath *lookup_datapath(struct net *net,
					struct ovs_header *ovs_header,
1326 1327 1328 1329 1330
					struct nlattr *a[OVS_DP_ATTR_MAX + 1])
{
	struct datapath *dp;

	if (!a[OVS_DP_ATTR_NAME])
1331
		dp = get_dp(net, ovs_header->dp_ifindex);
1332 1333 1334
	else {
		struct vport *vport;

1335
		vport = ovs_vport_locate(net, nla_data(a[OVS_DP_ATTR_NAME]));
1336 1337 1338 1339 1340
		dp = vport && vport->port_no == OVSP_LOCAL ? vport->dp : NULL;
	}
	return dp ? dp : ERR_PTR(-ENODEV);
}

1341 1342 1343 1344 1345
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);
1346
	if (IS_ERR(dp))
1347 1348 1349 1350 1351 1352
		return;

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

1353 1354 1355 1356 1357 1358
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]);
}

1359 1360 1361 1362 1363 1364 1365
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;
1366
	struct ovs_net *ovs_net;
1367
	int err, i;
1368 1369 1370 1371 1372

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

1373 1374 1375
	reply = ovs_dp_cmd_alloc_info(info);
	if (!reply)
		return -ENOMEM;
1376 1377 1378 1379

	err = -ENOMEM;
	dp = kzalloc(sizeof(*dp), GFP_KERNEL);
	if (dp == NULL)
1380
		goto err_free_reply;
1381 1382

	ovs_dp_set_net(dp, hold_net(sock_net(skb->sk)));
1383 1384

	/* Allocate table. */
1385 1386
	err = ovs_flow_tbl_init(&dp->table);
	if (err)
1387 1388
		goto err_free_dp;

1389
	dp->stats_percpu = netdev_alloc_pcpu_stats(struct dp_stats_percpu);
1390 1391 1392 1393 1394
	if (!dp->stats_percpu) {
		err = -ENOMEM;
		goto err_destroy_table;
	}

1395
	dp->ports = kmalloc(DP_VPORT_HASH_BUCKETS * sizeof(struct hlist_head),
1396
			    GFP_KERNEL);
1397 1398 1399 1400 1401 1402 1403 1404
	if (!dp->ports) {
		err = -ENOMEM;
		goto err_destroy_percpu;
	}

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

1405 1406 1407 1408 1409 1410
	/* 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;
1411
	parms.upcall_portids = a[OVS_DP_ATTR_UPCALL_PID];
1412

1413 1414
	ovs_dp_change(dp, a);

1415 1416 1417
	/* So far only local changes have been made, now need the lock. */
	ovs_lock();

1418 1419 1420 1421 1422 1423
	vport = new_vport(&parms);
	if (IS_ERR(vport)) {
		err = PTR_ERR(vport);
		if (err == -EBUSY)
			err = -EEXIST;

1424 1425 1426 1427 1428 1429 1430 1431 1432
		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);
		}

1433
		goto err_destroy_ports_array;
1434 1435
	}

1436 1437 1438
	err = ovs_dp_cmd_fill_info(dp, reply, info->snd_portid,
				   info->snd_seq, 0, OVS_DP_CMD_NEW);
	BUG_ON(err < 0);
1439

1440
	ovs_net = net_generic(ovs_dp_get_net(dp), ovs_net_id);
1441
	list_add_tail_rcu(&dp->list_node, &ovs_net->dps);
1442 1443

	ovs_unlock();
1444

1445
	ovs_notify(&dp_datapath_genl_family, reply, info);
1446 1447
	return 0;

1448
err_destroy_ports_array:
1449
	ovs_unlock();
1450
	kfree(dp->ports);
1451 1452 1453
err_destroy_percpu:
	free_percpu(dp->stats_percpu);
err_destroy_table:
1454
	ovs_flow_tbl_destroy(&dp->table);
1455
err_free_dp:
1456
	release_net(ovs_dp_get_net(dp));
1457
	kfree(dp);
1458 1459
err_free_reply:
	kfree_skb(reply);
1460 1461 1462 1463
err:
	return err;
}

1464
/* Called with ovs_mutex. */
1465
static void __dp_destroy(struct datapath *dp)
1466
{
1467
	int i;
1468

1469 1470
	for (i = 0; i < DP_VPORT_HASH_BUCKETS; i++) {
		struct vport *vport;
1471
		struct hlist_node *n;
1472

1473
		hlist_for_each_entry_safe(vport, n, &dp->ports[i], dp_hash_node)
1474 1475 1476
			if (vport->port_no != OVSP_LOCAL)
				ovs_dp_detach_port(vport);
	}
1477

1478
	list_del_rcu(&dp->list_node);
1479

1480
	/* OVSP_LOCAL is datapath internal port. We need to make sure that
1481
	 * all ports in datapath are destroyed first before freeing datapath.
1482
	 */
1483
	ovs_dp_detach_port(ovs_vport_ovsl(dp, OVSP_LOCAL));
1484

1485
	/* RCU destroy the flow table */
1486
	call_rcu(&dp->rcu, destroy_dp_rcu);
1487 1488 1489 1490 1491 1492 1493 1494
}

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

1505 1506 1507
	err = ovs_dp_cmd_fill_info(dp, reply, info->snd_portid,
				   info->snd_seq, 0, OVS_DP_CMD_DEL);
	BUG_ON(err < 0);
1508 1509

	__dp_destroy(dp);
1510
	ovs_unlock();
1511

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 1526 1527
}

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

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

1532
	ovs_lock();
1533
	dp = lookup_datapath(sock_net(skb->sk), info->userhdr, info->attrs);
1534
	err = PTR_ERR(dp);
1535
	if (IS_ERR(dp))
1536
		goto err_unlock_free;
1537

1538 1539
	ovs_dp_change(dp, info->attrs);

1540 1541 1542
	err = ovs_dp_cmd_fill_info(dp, reply, info->snd_portid,
				   info->snd_seq, 0, OVS_DP_CMD_NEW);
	BUG_ON(err < 0);
1543

1544
	ovs_unlock();
1545
	ovs_notify(&dp_datapath_genl_family, reply, info);
1546 1547

	return 0;
1548 1549

err_unlock_free:
1550
	ovs_unlock();
1551
	kfree_skb(reply);
1552
	return err;
1553 1554 1555 1556 1557 1558
}

static int ovs_dp_cmd_get(struct sk_buff *skb, struct genl_info *info)
{
	struct sk_buff *reply;
	struct datapath *dp;
1559
	int err;
1560

1561 1562 1563 1564 1565
	reply = ovs_dp_cmd_alloc_info(info);
	if (!reply)
		return -ENOMEM;

	rcu_read_lock();
1566
	dp = lookup_datapath(sock_net(skb->sk), info->userhdr, info->attrs);
1567 1568
	if (IS_ERR(dp)) {
		err = PTR_ERR(dp);
1569
		goto err_unlock_free;
1570
	}
1571 1572 1573 1574
	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();
1575 1576

	return genlmsg_reply(reply, info);
1577

1578 1579 1580
err_unlock_free:
	rcu_read_unlock();
	kfree_skb(reply);
1581
	return err;
1582 1583 1584 1585
}

static int ovs_dp_cmd_dump(struct sk_buff *skb, struct netlink_callback *cb)
{
1586
	struct ovs_net *ovs_net = net_generic(sock_net(skb->sk), ovs_net_id);
1587 1588 1589 1590
	struct datapath *dp;
	int skip = cb->args[0];
	int i = 0;

1591 1592
	rcu_read_lock();
	list_for_each_entry_rcu(dp, &ovs_net->dps, list_node) {
1593
		if (i >= skip &&
1594
		    ovs_dp_cmd_fill_info(dp, skb, NETLINK_CB(cb->skb).portid,
1595 1596 1597 1598 1599
					 cb->nlh->nlmsg_seq, NLM_F_MULTI,
					 OVS_DP_CMD_NEW) < 0)
			break;
		i++;
	}
1600
	rcu_read_unlock();
1601 1602 1603 1604 1605 1606

	cb->args[0] = i;

	return skb->len;
}

1607 1608 1609 1610 1611 1612
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 },
};

1613
static const struct genl_ops dp_datapath_genl_ops[] = {
1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636
	{ .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,
	},
};

1637
static struct genl_family dp_datapath_genl_family = {
1638 1639
	.id = GENL_ID_GENERATE,
	.hdrsize = sizeof(struct ovs_header),
1640 1641 1642
	.name = OVS_DATAPATH_FAMILY,
	.version = OVS_DATAPATH_VERSION,
	.maxattr = OVS_DP_ATTR_MAX,
1643 1644
	.netnsok = true,
	.parallel_ops = true,
1645 1646 1647 1648
	.ops = dp_datapath_genl_ops,
	.n_ops = ARRAY_SIZE(dp_datapath_genl_ops),
	.mcgrps = &ovs_dp_datapath_multicast_group,
	.n_mcgrps = 1,
1649 1650
};

1651
/* Called with ovs_mutex or RCU read lock. */
1652
static int ovs_vport_cmd_fill_info(struct vport *vport, struct sk_buff *skb,
1653
				   u32 portid, u32 seq, u32 flags, u8 cmd)
1654 1655 1656 1657 1658
{
	struct ovs_header *ovs_header;
	struct ovs_vport_stats vport_stats;
	int err;

1659
	ovs_header = genlmsg_put(skb, portid, seq, &dp_vport_genl_family,
1660 1661 1662 1663 1664 1665
				 flags, cmd);
	if (!ovs_header)
		return -EMSGSIZE;

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

1666 1667
	if (nla_put_u32(skb, OVS_VPORT_ATTR_PORT_NO, vport->port_no) ||
	    nla_put_u32(skb, OVS_VPORT_ATTR_TYPE, vport->ops->type) ||
1668 1669
	    nla_put_string(skb, OVS_VPORT_ATTR_NAME,
			   vport->ops->get_name(vport)))
1670
		goto nla_put_failure;
1671 1672

	ovs_vport_get_stats(vport, &vport_stats);
1673 1674 1675
	if (nla_put(skb, OVS_VPORT_ATTR_STATS, sizeof(struct ovs_vport_stats),
		    &vport_stats))
		goto nla_put_failure;
1676

1677 1678 1679
	if (ovs_vport_get_upcall_portids(vport, skb))
		goto nla_put_failure;

1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692
	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;
}

1693 1694 1695 1696 1697 1698
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(). */
1699
struct sk_buff *ovs_vport_cmd_build_info(struct vport *vport, u32 portid,
1700 1701 1702 1703 1704 1705 1706 1707 1708
					 u32 seq, u8 cmd)
{
	struct sk_buff *skb;
	int retval;

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

1709
	retval = ovs_vport_cmd_fill_info(vport, skb, portid, seq, 0, cmd);
1710 1711
	BUG_ON(retval < 0);

1712 1713 1714
	return skb;
}

1715
/* Called with ovs_mutex or RCU read lock. */
1716 1717
static struct vport *lookup_vport(struct net *net,
				  struct ovs_header *ovs_header,
1718 1719 1720 1721 1722 1723
				  struct nlattr *a[OVS_VPORT_ATTR_MAX + 1])
{
	struct datapath *dp;
	struct vport *vport;

	if (a[OVS_VPORT_ATTR_NAME]) {
1724
		vport = ovs_vport_locate(net, nla_data(a[OVS_VPORT_ATTR_NAME]));
1725 1726
		if (!vport)
			return ERR_PTR(-ENODEV);
1727 1728 1729
		if (ovs_header->dp_ifindex &&
		    ovs_header->dp_ifindex != get_dpifindex(vport->dp))
			return ERR_PTR(-ENODEV);
1730 1731 1732 1733 1734 1735 1736
		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);

1737
		dp = get_dp(net, ovs_header->dp_ifindex);
1738 1739 1740
		if (!dp)
			return ERR_PTR(-ENODEV);

1741
		vport = ovs_vport_ovsl_rcu(dp, port_no);
1742
		if (!vport)
1743
			return ERR_PTR(-ENODEV);
1744 1745 1746 1747 1748 1749 1750 1751 1752 1753 1754 1755 1756 1757 1758 1759 1760 1761
		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])
1762 1763 1764 1765 1766 1767 1768 1769 1770 1771
		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;
1772

1773
	ovs_lock();
1774
restart:
1775
	dp = get_dp(sock_net(skb->sk), ovs_header->dp_ifindex);
1776 1777
	err = -ENODEV;
	if (!dp)
1778
		goto exit_unlock_free;
1779

1780
	if (port_no) {
1781
		vport = ovs_vport_ovsl(dp, port_no);
1782 1783
		err = -EBUSY;
		if (vport)
1784
			goto exit_unlock_free;
1785 1786 1787 1788
	} else {
		for (port_no = 1; ; port_no++) {
			if (port_no >= DP_MAX_PORTS) {
				err = -EFBIG;
1789
				goto exit_unlock_free;
1790
			}
1791
			vport = ovs_vport_ovsl(dp, port_no);
1792 1793 1794 1795 1796 1797 1798 1799 1800 1801
			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;
1802
	parms.upcall_portids = a[OVS_VPORT_ATTR_UPCALL_PID];
1803 1804 1805

	vport = new_vport(&parms);
	err = PTR_ERR(vport);
1806 1807 1808
	if (IS_ERR(vport)) {
		if (err == -EAGAIN)
			goto restart;
1809
		goto exit_unlock_free;
1810
	}
1811

1812 1813 1814 1815
	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();
1816

1817
	ovs_notify(&dp_vport_genl_family, reply, info);
1818
	return 0;
1819

1820
exit_unlock_free:
1821
	ovs_unlock();
1822
	kfree_skb(reply);
1823 1824 1825 1826 1827 1828 1829 1830 1831 1832
	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;

1833 1834 1835 1836
	reply = ovs_vport_cmd_alloc_info();
	if (!reply)
		return -ENOMEM;

1837
	ovs_lock();
1838
	vport = lookup_vport(sock_net(skb->sk), info->userhdr, a);
1839 1840
	err = PTR_ERR(vport);
	if (IS_ERR(vport))
1841
		goto exit_unlock_free;
1842 1843

	if (a[OVS_VPORT_ATTR_TYPE] &&
1844
	    nla_get_u32(a[OVS_VPORT_ATTR_TYPE]) != vport->ops->type) {
1845
		err = -EINVAL;
1846
		goto exit_unlock_free;
1847 1848
	}

1849
	if (a[OVS_VPORT_ATTR_OPTIONS]) {
1850
		err = ovs_vport_set_options(vport, a[OVS_VPORT_ATTR_OPTIONS]);
1851
		if (err)
1852
			goto exit_unlock_free;
1853
	}
1854

1855 1856 1857 1858 1859 1860 1861 1862

	if (a[OVS_VPORT_ATTR_UPCALL_PID]) {
		struct nlattr *ids = a[OVS_VPORT_ATTR_UPCALL_PID];

		err = ovs_vport_set_upcall_portids(vport, ids);
		if (err)
			goto exit_unlock_free;
	}
1863

1864 1865 1866
	err = ovs_vport_cmd_fill_info(vport, reply, info->snd_portid,
				      info->snd_seq, 0, OVS_VPORT_CMD_NEW);
	BUG_ON(err < 0);
1867

1868
	ovs_unlock();
1869
	ovs_notify(&dp_vport_genl_family, reply, info);
1870
	return 0;
1871

1872
exit_unlock_free:
1873
	ovs_unlock();
1874
	kfree_skb(reply);
1875 1876 1877 1878 1879 1880 1881 1882 1883 1884
	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;

1885 1886 1887 1888
	reply = ovs_vport_cmd_alloc_info();
	if (!reply)
		return -ENOMEM;

1889
	ovs_lock();
1890
	vport = lookup_vport(sock_net(skb->sk), info->userhdr, a);
1891 1892
	err = PTR_ERR(vport);
	if (IS_ERR(vport))
1893
		goto exit_unlock_free;
1894 1895 1896

	if (vport->port_no == OVSP_LOCAL) {
		err = -EINVAL;
1897
		goto exit_unlock_free;
1898 1899
	}

1900 1901 1902
	err = ovs_vport_cmd_fill_info(vport, reply, info->snd_portid,
				      info->snd_seq, 0, OVS_VPORT_CMD_DEL);
	BUG_ON(err < 0);
1903
	ovs_dp_detach_port(vport);
1904
	ovs_unlock();
1905

1906
	ovs_notify(&dp_vport_genl_family, reply, info);
1907
	return 0;
1908

1909
exit_unlock_free:
1910
	ovs_unlock();
1911
	kfree_skb(reply);
1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922
	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;

1923 1924 1925 1926
	reply = ovs_vport_cmd_alloc_info();
	if (!reply)
		return -ENOMEM;

1927
	rcu_read_lock();
1928
	vport = lookup_vport(sock_net(skb->sk), ovs_header, a);
1929 1930
	err = PTR_ERR(vport);
	if (IS_ERR(vport))
1931 1932 1933 1934
		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);
1935 1936 1937 1938
	rcu_read_unlock();

	return genlmsg_reply(reply, info);

1939
exit_unlock_free:
1940
	rcu_read_unlock();
1941
	kfree_skb(reply);
1942 1943 1944 1945 1946 1947 1948
	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;
1949 1950
	int bucket = cb->args[0], skip = cb->args[1];
	int i, j = 0;
1951

J
Jarno Rajahalme 已提交
1952
	rcu_read_lock();
1953
	dp = get_dp(sock_net(skb->sk), ovs_header->dp_ifindex);
J
Jarno Rajahalme 已提交
1954 1955
	if (!dp) {
		rcu_read_unlock();
1956
		return -ENODEV;
J
Jarno Rajahalme 已提交
1957
	}
1958
	for (i = bucket; i < DP_VPORT_HASH_BUCKETS; i++) {
1959
		struct vport *vport;
1960 1961

		j = 0;
1962
		hlist_for_each_entry_rcu(vport, &dp->ports[i], dp_hash_node) {
1963 1964
			if (j >= skip &&
			    ovs_vport_cmd_fill_info(vport, skb,
1965
						    NETLINK_CB(cb->skb).portid,
1966 1967 1968 1969 1970 1971 1972 1973
						    cb->nlh->nlmsg_seq,
						    NLM_F_MULTI,
						    OVS_VPORT_CMD_NEW) < 0)
				goto out;

			j++;
		}
		skip = 0;
1974
	}
1975
out:
1976 1977
	rcu_read_unlock();

1978 1979
	cb->args[0] = i;
	cb->args[1] = j;
1980

1981
	return skb->len;
1982 1983
}

1984 1985 1986 1987 1988 1989 1990 1991 1992
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 },
};

1993
static const struct genl_ops dp_vport_genl_ops[] = {
1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016
	{ .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,
	},
};

2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028
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,
2029 2030
};

2031 2032 2033 2034 2035
static struct genl_family * const dp_genl_families[] = {
	&dp_datapath_genl_family,
	&dp_vport_genl_family,
	&dp_flow_genl_family,
	&dp_packet_genl_family,
2036 2037 2038 2039 2040 2041 2042
};

static void dp_unregister_genl(int n_families)
{
	int i;

	for (i = 0; i < n_families; i++)
2043
		genl_unregister_family(dp_genl_families[i]);
2044 2045 2046 2047 2048 2049 2050 2051 2052
}

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

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

2053
		err = genl_register_family(dp_genl_families[i]);
2054 2055 2056 2057 2058 2059 2060
		if (err)
			goto error;
	}

	return 0;

error:
2061
	dp_unregister_genl(i);
2062 2063 2064
	return err;
}

2065 2066 2067 2068 2069
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);
2070
	INIT_WORK(&ovs_net->dp_notify_work, ovs_dp_notify_wq);
2071 2072 2073 2074 2075 2076
	return 0;
}

static void __net_exit ovs_exit_net(struct net *net)
{
	struct datapath *dp, *dp_next;
2077
	struct ovs_net *ovs_net = net_generic(net, ovs_net_id);
2078

2079
	ovs_lock();
2080 2081
	list_for_each_entry_safe(dp, dp_next, &ovs_net->dps, list_node)
		__dp_destroy(dp);
2082 2083 2084
	ovs_unlock();

	cancel_work_sync(&ovs_net->dp_notify_work);
2085 2086 2087 2088 2089 2090 2091 2092 2093
}

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

2094 2095 2096 2097
static int __init dp_init(void)
{
	int err;

2098
	BUILD_BUG_ON(sizeof(struct ovs_skb_cb) > FIELD_SIZEOF(struct sk_buff, cb));
2099 2100 2101

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

2102
	err = action_fifos_init();
2103 2104 2105
	if (err)
		goto error;

2106 2107 2108 2109
	err = ovs_internal_dev_rtnl_link_register();
	if (err)
		goto error_action_fifos_exit;

2110 2111 2112 2113
	err = ovs_flow_init();
	if (err)
		goto error_unreg_rtnl_link;

2114 2115 2116 2117
	err = ovs_vport_init();
	if (err)
		goto error_flow_exit;

2118
	err = register_pernet_device(&ovs_net_ops);
2119 2120 2121
	if (err)
		goto error_vport_exit;

2122 2123 2124 2125
	err = register_netdevice_notifier(&ovs_dp_device_notifier);
	if (err)
		goto error_netns_exit;

2126 2127 2128 2129
	err = ovs_netdev_init();
	if (err)
		goto error_unreg_notifier;

2130 2131
	err = dp_register_genl();
	if (err < 0)
2132
		goto error_unreg_netdev;
2133 2134 2135

	return 0;

2136 2137
error_unreg_netdev:
	ovs_netdev_exit();
2138 2139
error_unreg_notifier:
	unregister_netdevice_notifier(&ovs_dp_device_notifier);
2140 2141
error_netns_exit:
	unregister_pernet_device(&ovs_net_ops);
2142 2143 2144 2145
error_vport_exit:
	ovs_vport_exit();
error_flow_exit:
	ovs_flow_exit();
2146 2147
error_unreg_rtnl_link:
	ovs_internal_dev_rtnl_link_unregister();
2148 2149
error_action_fifos_exit:
	action_fifos_exit();
2150 2151 2152 2153 2154 2155 2156
error:
	return err;
}

static void dp_cleanup(void)
{
	dp_unregister_genl(ARRAY_SIZE(dp_genl_families));
2157
	ovs_netdev_exit();
2158
	unregister_netdevice_notifier(&ovs_dp_device_notifier);
2159 2160
	unregister_pernet_device(&ovs_net_ops);
	rcu_barrier();
2161 2162
	ovs_vport_exit();
	ovs_flow_exit();
2163
	ovs_internal_dev_rtnl_link_unregister();
2164
	action_fifos_exit();
2165 2166 2167 2168 2169 2170 2171
}

module_init(dp_init);
module_exit(dp_cleanup);

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