datapath.c 58.1 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_GPL(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 nla_policy flow_policy[];

70 71
static const struct genl_multicast_group ovs_dp_flow_multicast_group = {
	.name = OVS_FLOW_MCGROUP,
72 73
};

74 75
static const struct genl_multicast_group ovs_dp_datapath_multicast_group = {
	.name = OVS_DATAPATH_MCGROUP,
76 77
};

78 79
static const struct genl_multicast_group ovs_dp_vport_multicast_group = {
	.name = OVS_VPORT_MCGROUP,
80 81
};

82 83
/* Check if need to build a reply message.
 * OVS userspace sets the NLM_F_ECHO flag if it needs the reply. */
84 85
static bool ovs_must_notify(struct genl_family *family, struct genl_info *info,
			    unsigned int group)
86 87
{
	return info->nlhdr->nlmsg_flags & NLM_F_ECHO ||
88
	       genl_has_listeners(family, genl_info_net(info), group);
89 90
}

91
static void ovs_notify(struct genl_family *family,
92
		       struct sk_buff *skb, struct genl_info *info)
93
{
J
Jiri Benc 已提交
94
	genl_notify(family, skb, info, 0, 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_GPL(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 sw_flow_key *,
140
			     const struct dp_upcall_info *);
141
static int queue_userspace_packet(struct datapath *dp, struct sk_buff *,
142
				  const struct sw_flow_key *,
143 144
				  const struct dp_upcall_info *);

145 146
/* Must be called with rcu_read_lock. */
static struct datapath *get_dp_rcu(struct net *net, int dp_ifindex)
147
{
148
	struct net_device *dev = dev_get_by_index_rcu(net, dp_ifindex);
149 150 151 152

	if (dev) {
		struct vport *vport = ovs_internal_dev_get_vport(dev);
		if (vport)
153
			return vport->dp;
154
	}
155 156 157 158 159 160 161 162 163 164 165 166 167 168

	return NULL;
}

/* The caller must hold either ovs_mutex or rcu_read_lock to keep the
 * returned dp pointer valid.
 */
static inline struct datapath *get_dp(struct net *net, int dp_ifindex)
{
	struct datapath *dp;

	WARN_ON_ONCE(!rcu_read_lock_held() && !lockdep_ovsl_is_held());
	rcu_read_lock();
	dp = get_dp_rcu(net, dp_ifindex);
169 170 171 172 173
	rcu_read_unlock();

	return dp;
}

174
/* Must be called with rcu_read_lock or ovs_mutex. */
175
const char *ovs_dp_name(const struct datapath *dp)
176
{
177
	struct vport *vport = ovs_vport_ovsl_rcu(dp, OVSP_LOCAL);
178
	return ovs_vport_name(vport);
179 180
}

181
static int get_dpifindex(const struct datapath *dp)
182 183 184 185 186 187
{
	struct vport *local;
	int ifindex;

	rcu_read_lock();

188
	local = ovs_vport_rcu(dp, OVSP_LOCAL);
189
	if (local)
190
		ifindex = local->dev->ifindex;
191 192 193 194 195 196 197 198 199 200 201 202
	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);

203
	ovs_flow_tbl_destroy(&dp->table);
204
	free_percpu(dp->stats_percpu);
205
	kfree(dp->ports);
206 207 208
	kfree(dp);
}

209 210 211 212 213 214
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
215
/* Called with ovs_mutex or RCU read lock. */
216 217 218 219 220 221
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);
222
	hlist_for_each_entry_rcu(vport, head, dp_hash_node) {
223 224 225 226 227 228
		if (vport->port_no == port_no)
			return vport;
	}
	return NULL;
}

229
/* Called with ovs_mutex. */
230 231 232 233 234 235 236
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;
237
		struct hlist_head *head = vport_hash_bucket(dp, vport->port_no);
238

239
		hlist_add_head_rcu(&vport->dp_hash_node, head);
240 241 242 243 244 245
	}
	return vport;
}

void ovs_dp_detach_port(struct vport *p)
{
246
	ASSERT_OVSL();
247 248

	/* First drop references to device. */
249
	hlist_del_rcu(&p->dp_hash_node);
250 251 252 253 254 255

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

/* Must be called with rcu_read_lock. */
256
void ovs_dp_process_packet(struct sk_buff *skb, struct sw_flow_key *key)
257
{
258
	const struct vport *p = OVS_CB(skb)->input_vport;
259 260
	struct datapath *dp = p->dp;
	struct sw_flow *flow;
261
	struct sw_flow_actions *sf_acts;
262 263
	struct dp_stats_percpu *stats;
	u64 *stats_counter;
264
	u32 n_mask_hit;
265

266
	stats = this_cpu_ptr(dp->stats_percpu);
267 268

	/* Look up flow. */
269
	flow = ovs_flow_tbl_lookup_stats(&dp->table, key, &n_mask_hit);
270 271
	if (unlikely(!flow)) {
		struct dp_upcall_info upcall;
272
		int error;
273

274
		memset(&upcall, 0, sizeof(upcall));
275
		upcall.cmd = OVS_PACKET_CMD_MISS;
276
		upcall.portid = ovs_vport_find_upcall_portid(p, skb);
J
Joe Stringer 已提交
277
		upcall.mru = OVS_CB(skb)->mru;
278
		error = ovs_dp_upcall(dp, skb, key, &upcall);
279 280 281 282
		if (unlikely(error))
			kfree_skb(skb);
		else
			consume_skb(skb);
283 284 285 286
		stats_counter = &stats->n_missed;
		goto out;
	}

287 288 289
	ovs_flow_stats_update(flow, key->tp.flags, skb);
	sf_acts = rcu_dereference(flow->sf_acts);
	ovs_execute_actions(dp, skb, sf_acts, key);
290

291
	stats_counter = &stats->n_hit;
292 293 294

out:
	/* Update datapath statistics. */
295
	u64_stats_update_begin(&stats->syncp);
296
	(*stats_counter)++;
297
	stats->n_mask_hit += n_mask_hit;
298
	u64_stats_update_end(&stats->syncp);
299 300 301
}

int ovs_dp_upcall(struct datapath *dp, struct sk_buff *skb,
302
		  const struct sw_flow_key *key,
303
		  const struct dp_upcall_info *upcall_info)
304 305 306 307
{
	struct dp_stats_percpu *stats;
	int err;

308
	if (upcall_info->portid == 0) {
309 310 311 312 313
		err = -ENOTCONN;
		goto err;
	}

	if (!skb_is_gso(skb))
314
		err = queue_userspace_packet(dp, skb, key, upcall_info);
315
	else
316
		err = queue_gso_packets(dp, skb, key, upcall_info);
317 318 319 320 321 322
	if (err)
		goto err;

	return 0;

err:
323
	stats = this_cpu_ptr(dp->stats_percpu);
324

325
	u64_stats_update_begin(&stats->syncp);
326
	stats->n_lost++;
327
	u64_stats_update_end(&stats->syncp);
328 329 330 331

	return err;
}

332
static int queue_gso_packets(struct datapath *dp, struct sk_buff *skb,
333
			     const struct sw_flow_key *key,
334 335
			     const struct dp_upcall_info *upcall_info)
{
336
	unsigned short gso_type = skb_shinfo(skb)->gso_type;
337 338 339 340
	struct sw_flow_key later_key;
	struct sk_buff *segs, *nskb;
	int err;

341
	BUILD_BUG_ON(sizeof(*OVS_CB(skb)) > SKB_SGO_CB_OFFSET);
342
	segs = __skb_gso_segment(skb, NETIF_F_SG, false);
343 344
	if (IS_ERR(segs))
		return PTR_ERR(segs);
345 346
	if (segs == NULL)
		return -EINVAL;
347

348 349 350 351 352 353 354 355 356
	if (gso_type & SKB_GSO_UDP) {
		/* The initial flow key extracted by ovs_flow_key_extract()
		 * in this case is for a first fragment, so we need to
		 * properly mark later fragments.
		 */
		later_key = *key;
		later_key.ip.frag = OVS_FRAG_TYPE_LATER;
	}

357 358 359
	/* Queue all of the segments. */
	skb = segs;
	do {
360 361 362 363
		if (gso_type & SKB_GSO_UDP && skb != segs)
			key = &later_key;

		err = queue_userspace_packet(dp, skb, key, upcall_info);
364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380
		if (err)
			break;

	} 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;
}

381
static size_t upcall_msg_size(const struct dp_upcall_info *upcall_info,
382
			      unsigned int hdrlen)
383 384
{
	size_t size = NLMSG_ALIGN(sizeof(struct ovs_header))
385
		+ nla_total_size(hdrlen) /* OVS_PACKET_ATTR_PACKET */
386
		+ nla_total_size(ovs_key_attr_size()); /* OVS_PACKET_ATTR_KEY */
387 388

	/* OVS_PACKET_ATTR_USERDATA */
389 390 391 392 393 394
	if (upcall_info->userdata)
		size += NLA_ALIGN(upcall_info->userdata->nla_len);

	/* OVS_PACKET_ATTR_EGRESS_TUN_KEY */
	if (upcall_info->egress_tun_info)
		size += nla_total_size(ovs_tun_key_attr_size());
395

396 397 398 399
	/* OVS_PACKET_ATTR_ACTIONS */
	if (upcall_info->actions_len)
		size += nla_total_size(upcall_info->actions_len);

J
Joe Stringer 已提交
400 401 402 403
	/* OVS_PACKET_ATTR_MRU */
	if (upcall_info->mru)
		size += nla_total_size(sizeof(upcall_info->mru));

404 405 406
	return size;
}

J
Joe Stringer 已提交
407 408 409 410 411 412 413 414 415 416
static void pad_packet(struct datapath *dp, struct sk_buff *skb)
{
	if (!(dp->user_features & OVS_DP_F_UNALIGNED)) {
		size_t plen = NLA_ALIGN(skb->len) - skb->len;

		if (plen > 0)
			memset(skb_put(skb, plen), 0, plen);
	}
}

417
static int queue_userspace_packet(struct datapath *dp, struct sk_buff *skb,
418
				  const struct sw_flow_key *key,
419 420 421 422
				  const struct dp_upcall_info *upcall_info)
{
	struct ovs_header *upcall;
	struct sk_buff *nskb = NULL;
L
Li RongQing 已提交
423
	struct sk_buff *user_skb = NULL; /* to be queued to userspace */
424
	struct nlattr *nla;
425
	size_t len;
426
	unsigned int hlen;
427 428 429 430 431
	int err, dp_ifindex;

	dp_ifindex = get_dpifindex(dp);
	if (!dp_ifindex)
		return -ENODEV;
432

433
	if (skb_vlan_tag_present(skb)) {
434 435 436 437
		nskb = skb_clone(skb, GFP_ATOMIC);
		if (!nskb)
			return -ENOMEM;

438
		nskb = __vlan_hwaccel_push_inside(nskb);
439
		if (!nskb)
440 441 442 443 444 445 446 447 448 449
			return -ENOMEM;

		skb = nskb;
	}

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

450 451 452 453 454 455 456 457 458 459 460 461 462 463
	/* 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;

464
	len = upcall_msg_size(upcall_info, hlen);
465
	user_skb = genlmsg_new(len, GFP_ATOMIC);
466 467 468 469 470 471 472 473 474
	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;

475
	err = ovs_nla_put_key(key, key, OVS_PACKET_ATTR_KEY, false, user_skb);
476
	BUG_ON(err);
477 478

	if (upcall_info->userdata)
479 480 481
		__nla_put(user_skb, OVS_PACKET_ATTR_USERDATA,
			  nla_len(upcall_info->userdata),
			  nla_data(upcall_info->userdata));
482

483 484
	if (upcall_info->egress_tun_info) {
		nla = nla_nest_start(user_skb, OVS_PACKET_ATTR_EGRESS_TUN_KEY);
485 486
		err = ovs_nla_put_tunnel_info(user_skb,
					      upcall_info->egress_tun_info);
487 488 489 490
		BUG_ON(err);
		nla_nest_end(user_skb, nla);
	}

491 492 493 494 495 496 497 498 499 500 501
	if (upcall_info->actions_len) {
		nla = nla_nest_start(user_skb, OVS_PACKET_ATTR_ACTIONS);
		err = ovs_nla_put_actions(upcall_info->actions,
					  upcall_info->actions_len,
					  user_skb);
		if (!err)
			nla_nest_end(user_skb, nla);
		else
			nla_nest_cancel(user_skb, nla);
	}

J
Joe Stringer 已提交
502 503 504 505 506 507 508 509 510 511
	/* Add OVS_PACKET_ATTR_MRU */
	if (upcall_info->mru) {
		if (nla_put_u16(user_skb, OVS_PACKET_ATTR_MRU,
				upcall_info->mru)) {
			err = -ENOBUFS;
			goto out;
		}
		pad_packet(dp, user_skb);
	}

512 513 514 515 516 517 518
	/* 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);
519

520 521 522
	err = skb_zerocopy(user_skb, skb, skb->len, hlen);
	if (err)
		goto out;
523

524
	/* Pad OVS_PACKET_ATTR_PACKET if linear copy was performed */
J
Joe Stringer 已提交
525
	pad_packet(dp, user_skb);
526

527
	((struct nlmsghdr *) user_skb->data)->nlmsg_len = user_skb->len;
528

529
	err = genlmsg_unicast(ovs_dp_get_net(dp), user_skb, upcall_info->portid);
L
Li RongQing 已提交
530
	user_skb = NULL;
531
out:
532 533
	if (err)
		skb_tx_error(skb);
L
Li RongQing 已提交
534
	kfree_skb(user_skb);
535 536 537 538 539 540 541
	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;
J
Joe Stringer 已提交
542
	struct net *net = sock_net(skb->sk);
543 544 545 546
	struct nlattr **a = info->attrs;
	struct sw_flow_actions *acts;
	struct sk_buff *packet;
	struct sw_flow *flow;
547
	struct sw_flow_actions *sf_acts;
548 549
	struct datapath *dp;
	struct ethhdr *eth;
550
	struct vport *input_vport;
J
Joe Stringer 已提交
551
	u16 mru = 0;
552 553
	int len;
	int err;
554
	bool log = !a[OVS_PACKET_ATTR_PROBE];
555 556 557

	err = -EINVAL;
	if (!a[OVS_PACKET_ATTR_PACKET] || !a[OVS_PACKET_ATTR_KEY] ||
558
	    !a[OVS_PACKET_ATTR_ACTIONS])
559 560 561 562 563 564 565 566 567
		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);

568
	nla_memcpy(__skb_put(packet, len), a[OVS_PACKET_ATTR_PACKET], len);
569 570 571 572 573 574 575

	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. */
576
	if (eth_proto_is_802_3(eth->h_proto))
577 578 579 580
		packet->protocol = eth->h_proto;
	else
		packet->protocol = htons(ETH_P_802_2);

J
Joe Stringer 已提交
581 582 583 584 585 586 587
	/* Set packet's mru */
	if (a[OVS_PACKET_ATTR_MRU]) {
		mru = nla_get_u16(a[OVS_PACKET_ATTR_MRU]);
		packet->ignore_df = 1;
	}
	OVS_CB(packet)->mru = mru;

588
	/* Build an sw_flow for sending this packet. */
589
	flow = ovs_flow_alloc();
590 591 592 593
	err = PTR_ERR(flow);
	if (IS_ERR(flow))
		goto err_kfree_skb;

594 595
	err = ovs_flow_key_extract_userspace(net, a[OVS_PACKET_ATTR_KEY],
					     packet, &flow->key, log);
596 597 598
	if (err)
		goto err_flow_free;

J
Joe Stringer 已提交
599
	err = ovs_nla_copy_actions(net, a[OVS_PACKET_ATTR_ACTIONS],
600
				   &flow->key, &acts, log);
601 602
	if (err)
		goto err_flow_free;
603

604
	rcu_assign_pointer(flow->sf_acts, acts);
605
	packet->priority = flow->key.phy.priority;
606
	packet->mark = flow->key.phy.skb_mark;
607 608

	rcu_read_lock();
J
Joe Stringer 已提交
609
	dp = get_dp_rcu(net, ovs_header->dp_ifindex);
610 611 612 613
	err = -ENODEV;
	if (!dp)
		goto err_unlock;

614 615 616 617 618 619 620
	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;

J
Joe Stringer 已提交
621
	packet->dev = input_vport->dev;
622
	OVS_CB(packet)->input_vport = input_vport;
623
	sf_acts = rcu_dereference(flow->sf_acts);
624

625
	local_bh_disable();
626
	err = ovs_execute_actions(dp, packet, sf_acts, &flow->key);
627 628 629
	local_bh_enable();
	rcu_read_unlock();

630
	ovs_flow_free(flow, false);
631 632 633 634 635
	return err;

err_unlock:
	rcu_read_unlock();
err_flow_free:
636
	ovs_flow_free(flow, false);
637 638 639 640 641 642 643
err_kfree_skb:
	kfree_skb(packet);
err:
	return err;
}

static const struct nla_policy packet_policy[OVS_PACKET_ATTR_MAX + 1] = {
644
	[OVS_PACKET_ATTR_PACKET] = { .len = ETH_HLEN },
645 646
	[OVS_PACKET_ATTR_KEY] = { .type = NLA_NESTED },
	[OVS_PACKET_ATTR_ACTIONS] = { .type = NLA_NESTED },
647
	[OVS_PACKET_ATTR_PROBE] = { .type = NLA_FLAG },
J
Joe Stringer 已提交
648
	[OVS_PACKET_ATTR_MRU] = { .type = NLA_U16 },
649 650
};

651
static const struct genl_ops dp_packet_genl_ops[] = {
652
	{ .cmd = OVS_PACKET_CMD_EXECUTE,
653
	  .flags = GENL_UNS_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
654 655 656 657 658
	  .policy = packet_policy,
	  .doit = ovs_packet_cmd_execute
	}
};

659 660 661 662 663 664 665 666 667 668 669 670
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),
};

671
static void get_dp_stats(const struct datapath *dp, struct ovs_dp_stats *stats,
672
			 struct ovs_dp_megaflow_stats *mega_stats)
673 674 675
{
	int i;

676 677
	memset(mega_stats, 0, sizeof(*mega_stats));

678
	stats->n_flows = ovs_flow_tbl_count(&dp->table);
679
	mega_stats->n_masks = ovs_flow_tbl_num_masks(&dp->table);
680 681

	stats->n_hit = stats->n_missed = stats->n_lost = 0;
682

683 684 685 686 687 688 689 690
	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 {
691
			start = u64_stats_fetch_begin_irq(&percpu_stats->syncp);
692
			local_stats = *percpu_stats;
693
		} while (u64_stats_fetch_retry_irq(&percpu_stats->syncp, start));
694 695 696 697

		stats->n_hit += local_stats.n_hit;
		stats->n_missed += local_stats.n_missed;
		stats->n_lost += local_stats.n_lost;
698
		mega_stats->n_mask_hit += local_stats.n_mask_hit;
699 700 701
	}
}

702 703 704 705 706 707 708 709 710 711 712 713
static bool should_fill_key(const struct sw_flow_id *sfid, uint32_t ufid_flags)
{
	return ovs_identifier_is_ufid(sfid) &&
	       !(ufid_flags & OVS_UFID_F_OMIT_KEY);
}

static bool should_fill_mask(uint32_t ufid_flags)
{
	return !(ufid_flags & OVS_UFID_F_OMIT_MASK);
}

static bool should_fill_actions(uint32_t ufid_flags)
714
{
715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737
	return !(ufid_flags & OVS_UFID_F_OMIT_ACTIONS);
}

static size_t ovs_flow_cmd_msg_size(const struct sw_flow_actions *acts,
				    const struct sw_flow_id *sfid,
				    uint32_t ufid_flags)
{
	size_t len = NLMSG_ALIGN(sizeof(struct ovs_header));

	/* OVS_FLOW_ATTR_UFID */
	if (sfid && ovs_identifier_is_ufid(sfid))
		len += nla_total_size(sfid->ufid_len);

	/* OVS_FLOW_ATTR_KEY */
	if (!sfid || should_fill_key(sfid, ufid_flags))
		len += nla_total_size(ovs_key_attr_size());

	/* OVS_FLOW_ATTR_MASK */
	if (should_fill_mask(ufid_flags))
		len += nla_total_size(ovs_key_attr_size());

	/* OVS_FLOW_ATTR_ACTIONS */
	if (should_fill_actions(ufid_flags))
738
		len += nla_total_size(acts->orig_len);
739 740

	return len
741 742
		+ nla_total_size(sizeof(struct ovs_flow_stats)) /* OVS_FLOW_ATTR_STATS */
		+ nla_total_size(1) /* OVS_FLOW_ATTR_TCP_FLAGS */
743
		+ nla_total_size(8); /* OVS_FLOW_ATTR_USED */
744 745
}

746 747 748 749 750 751 752
/* Called with ovs_mutex or RCU read lock. */
static int ovs_flow_cmd_fill_stats(const struct sw_flow *flow,
				   struct sk_buff *skb)
{
	struct ovs_flow_stats stats;
	__be16 tcp_flags;
	unsigned long used;
753

754
	ovs_flow_stats_get(flow, &stats, &used, &tcp_flags);
755

756
	if (used &&
N
Nicolas Dichtel 已提交
757 758
	    nla_put_u64_64bit(skb, OVS_FLOW_ATTR_USED, ovs_flow_used_time(used),
			      OVS_FLOW_ATTR_PAD))
759
		return -EMSGSIZE;
760

761
	if (stats.n_packets &&
762
	    nla_put(skb, OVS_FLOW_ATTR_STATS, sizeof(struct ovs_flow_stats), &stats))
763
		return -EMSGSIZE;
764

765 766
	if ((u8)ntohs(tcp_flags) &&
	     nla_put_u8(skb, OVS_FLOW_ATTR_TCP_FLAGS, (u8)ntohs(tcp_flags)))
767 768 769 770 771 772 773 774 775 776 777
		return -EMSGSIZE;

	return 0;
}

/* Called with ovs_mutex or RCU read lock. */
static int ovs_flow_cmd_fill_actions(const struct sw_flow *flow,
				     struct sk_buff *skb, int skb_orig_len)
{
	struct nlattr *start;
	int err;
778 779 780 781 782 783 784 785 786 787 788

	/* 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.
	 */
789 790
	start = nla_nest_start(skb, OVS_FLOW_ATTR_ACTIONS);
	if (start) {
791 792
		const struct sw_flow_actions *sf_acts;

793
		sf_acts = rcu_dereference_ovsl(flow->sf_acts);
794 795
		err = ovs_nla_put_actions(sf_acts->actions,
					  sf_acts->actions_len, skb);
796

797 798 799 800
		if (!err)
			nla_nest_end(skb, start);
		else {
			if (skb_orig_len)
801
				return err;
802 803 804

			nla_nest_cancel(skb, start);
		}
805 806 807 808 809 810 811 812 813 814
	} else if (skb_orig_len) {
		return -EMSGSIZE;
	}

	return 0;
}

/* Called with ovs_mutex or RCU read lock. */
static int ovs_flow_cmd_fill_info(const struct sw_flow *flow, int dp_ifindex,
				  struct sk_buff *skb, u32 portid,
815
				  u32 seq, u32 flags, u8 cmd, u32 ufid_flags)
816 817 818 819 820 821 822 823 824 825 826 827
{
	const int skb_orig_len = skb->len;
	struct ovs_header *ovs_header;
	int err;

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

	ovs_header->dp_ifindex = dp_ifindex;

828
	err = ovs_nla_put_identifier(flow, skb);
829 830 831
	if (err)
		goto error;

832 833 834 835 836 837 838 839 840 841 842
	if (should_fill_key(&flow->id, ufid_flags)) {
		err = ovs_nla_put_masked_key(flow, skb);
		if (err)
			goto error;
	}

	if (should_fill_mask(ufid_flags)) {
		err = ovs_nla_put_mask(flow, skb);
		if (err)
			goto error;
	}
843 844 845 846 847

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

848 849 850 851 852
	if (should_fill_actions(ufid_flags)) {
		err = ovs_flow_cmd_fill_actions(flow, skb, skb_orig_len);
		if (err)
			goto error;
	}
853

854 855
	genlmsg_end(skb, ovs_header);
	return 0;
856 857 858 859 860 861

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

862 863
/* May not be called with RCU read lock. */
static struct sk_buff *ovs_flow_cmd_alloc_info(const struct sw_flow_actions *acts,
864
					       const struct sw_flow_id *sfid,
865
					       struct genl_info *info,
866 867
					       bool always,
					       uint32_t ufid_flags)
868
{
869
	struct sk_buff *skb;
870
	size_t len;
871

872
	if (!always && !ovs_must_notify(&dp_flow_genl_family, info, 0))
873 874
		return NULL;

875
	len = ovs_flow_cmd_msg_size(acts, sfid, ufid_flags);
876
	skb = genlmsg_new(len, GFP_KERNEL);
877 878 879 880
	if (!skb)
		return ERR_PTR(-ENOMEM);

	return skb;
881 882
}

883 884 885 886
/* 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,
887
					       bool always, u32 ufid_flags)
888 889 890 891
{
	struct sk_buff *skb;
	int retval;

892 893
	skb = ovs_flow_cmd_alloc_info(ovsl_dereference(flow->sf_acts),
				      &flow->id, info, always, ufid_flags);
894
	if (IS_ERR_OR_NULL(skb))
895
		return skb;
896

897 898
	retval = ovs_flow_cmd_fill_info(flow, dp_ifindex, skb,
					info->snd_portid, info->snd_seq, 0,
899
					cmd, ufid_flags);
900 901 902 903
	BUG_ON(retval < 0);
	return skb;
}

904
static int ovs_flow_cmd_new(struct sk_buff *skb, struct genl_info *info)
905
{
J
Joe Stringer 已提交
906
	struct net *net = sock_net(skb->sk);
907 908
	struct nlattr **a = info->attrs;
	struct ovs_header *ovs_header = info->userhdr;
909
	struct sw_flow *flow = NULL, *new_flow;
910
	struct sw_flow_mask mask;
911 912
	struct sk_buff *reply;
	struct datapath *dp;
913
	struct sw_flow_key key;
914
	struct sw_flow_actions *acts;
915
	struct sw_flow_match match;
916
	u32 ufid_flags = ovs_nla_get_ufid_flags(a[OVS_FLOW_ATTR_UFID_FLAGS]);
917
	int error;
918
	bool log = !a[OVS_FLOW_ATTR_PROBE];
919

920
	/* Must have key and actions. */
921
	error = -EINVAL;
922
	if (!a[OVS_FLOW_ATTR_KEY]) {
923
		OVS_NLERR(log, "Flow key attr not present in new flow.");
924
		goto error;
925 926
	}
	if (!a[OVS_FLOW_ATTR_ACTIONS]) {
927
		OVS_NLERR(log, "Flow actions attr not present in new flow.");
928
		goto error;
929
	}
930

931 932 933 934 935 936 937 938 939 940
	/* 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. */
941
	ovs_match_init(&match, &key, &mask);
942
	error = ovs_nla_get_match(net, &match, a[OVS_FLOW_ATTR_KEY],
943
				  a[OVS_FLOW_ATTR_MASK], log);
944
	if (error)
945
		goto err_kfree_flow;
946

947
	ovs_flow_mask_key(&new_flow->key, &key, true, &mask);
948 949 950 951 952 953

	/* Extract flow identifier. */
	error = ovs_nla_get_identifier(&new_flow->id, a[OVS_FLOW_ATTR_UFID],
				       &key, log);
	if (error)
		goto err_kfree_flow;
954

955
	/* Validate actions. */
J
Joe Stringer 已提交
956 957
	error = ovs_nla_copy_actions(net, a[OVS_FLOW_ATTR_ACTIONS],
				     &new_flow->key, &acts, log);
958
	if (error) {
959
		OVS_NLERR(log, "Flow actions may not be safe on all matching packets.");
960
		goto err_kfree_flow;
961 962
	}

963 964
	reply = ovs_flow_cmd_alloc_info(acts, &new_flow->id, info, false,
					ufid_flags);
965 966 967
	if (IS_ERR(reply)) {
		error = PTR_ERR(reply);
		goto err_kfree_acts;
968 969
	}

970
	ovs_lock();
J
Joe Stringer 已提交
971
	dp = get_dp(net, ovs_header->dp_ifindex);
972 973
	if (unlikely(!dp)) {
		error = -ENODEV;
974
		goto err_unlock_ovs;
975
	}
976

977
	/* Check if this is a duplicate flow */
978 979 980 981
	if (ovs_identifier_is_ufid(&new_flow->id))
		flow = ovs_flow_tbl_lookup_ufid(&dp->table, &new_flow->id);
	if (!flow)
		flow = ovs_flow_tbl_lookup(&dp->table, &key);
982 983
	if (likely(!flow)) {
		rcu_assign_pointer(new_flow->sf_acts, acts);
984 985

		/* Put flow in bucket. */
986 987
		error = ovs_flow_tbl_insert(&dp->table, new_flow, &mask);
		if (unlikely(error)) {
988
			acts = NULL;
989 990 991 992 993 994 995 996
			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,
997 998
						       OVS_FLOW_CMD_NEW,
						       ufid_flags);
999
			BUG_ON(error < 0);
1000
		}
1001
		ovs_unlock();
1002
	} else {
1003 1004
		struct sw_flow_actions *old_acts;

1005 1006 1007 1008 1009 1010
		/* 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.
		 */
1011 1012 1013
		if (unlikely(info->nlhdr->nlmsg_flags & (NLM_F_CREATE
							 | NLM_F_EXCL))) {
			error = -EEXIST;
1014
			goto err_unlock_ovs;
1015
		}
1016 1017 1018 1019 1020 1021 1022 1023 1024
		/* The flow identifier has to be the same for flow updates.
		 * Look for any overlapping flow.
		 */
		if (unlikely(!ovs_flow_cmp(flow, &match))) {
			if (ovs_identifier_is_key(&flow->id))
				flow = ovs_flow_tbl_lookup_exact(&dp->table,
								 &match);
			else /* UFID matches but key is different */
				flow = NULL;
1025 1026 1027 1028
			if (!flow) {
				error = -ENOENT;
				goto err_unlock_ovs;
			}
1029
		}
1030 1031 1032 1033
		/* Update actions. */
		old_acts = ovsl_dereference(flow->sf_acts);
		rcu_assign_pointer(flow->sf_acts, acts);

1034 1035 1036 1037 1038
		if (unlikely(reply)) {
			error = ovs_flow_cmd_fill_info(flow,
						       ovs_header->dp_ifindex,
						       reply, info->snd_portid,
						       info->snd_seq, 0,
1039 1040
						       OVS_FLOW_CMD_NEW,
						       ufid_flags);
1041 1042 1043
			BUG_ON(error < 0);
		}
		ovs_unlock();
1044

1045
		ovs_nla_free_flow_actions_rcu(old_acts);
1046
		ovs_flow_free(new_flow, false);
1047
	}
1048 1049 1050

	if (reply)
		ovs_notify(&dp_flow_genl_family, reply, info);
1051 1052 1053 1054
	return 0;

err_unlock_ovs:
	ovs_unlock();
1055 1056
	kfree_skb(reply);
err_kfree_acts:
1057
	ovs_nla_free_flow_actions(acts);
1058 1059
err_kfree_flow:
	ovs_flow_free(new_flow, false);
1060 1061 1062
error:
	return error;
}
1063

1064
/* Factor out action copy to avoid "Wframe-larger-than=1024" warning. */
J
Joe Stringer 已提交
1065 1066
static struct sw_flow_actions *get_flow_actions(struct net *net,
						const struct nlattr *a,
1067
						const struct sw_flow_key *key,
1068 1069
						const struct sw_flow_mask *mask,
						bool log)
1070 1071 1072 1073 1074
{
	struct sw_flow_actions *acts;
	struct sw_flow_key masked_key;
	int error;

1075
	ovs_flow_mask_key(&masked_key, key, true, mask);
J
Joe Stringer 已提交
1076
	error = ovs_nla_copy_actions(net, a, &masked_key, &acts, log);
1077
	if (error) {
1078 1079
		OVS_NLERR(log,
			  "Actions may not be safe on all matching packets");
1080 1081 1082 1083 1084 1085
		return ERR_PTR(error);
	}

	return acts;
}

1086 1087
static int ovs_flow_cmd_set(struct sk_buff *skb, struct genl_info *info)
{
J
Joe Stringer 已提交
1088
	struct net *net = sock_net(skb->sk);
1089 1090
	struct nlattr **a = info->attrs;
	struct ovs_header *ovs_header = info->userhdr;
1091
	struct sw_flow_key key;
1092 1093 1094 1095
	struct sw_flow *flow;
	struct sw_flow_mask mask;
	struct sk_buff *reply = NULL;
	struct datapath *dp;
1096
	struct sw_flow_actions *old_acts = NULL, *acts = NULL;
1097
	struct sw_flow_match match;
1098 1099
	struct sw_flow_id sfid;
	u32 ufid_flags = ovs_nla_get_ufid_flags(a[OVS_FLOW_ATTR_UFID_FLAGS]);
1100
	int error = 0;
1101
	bool log = !a[OVS_FLOW_ATTR_PROBE];
1102
	bool ufid_present;
1103

1104
	ufid_present = ovs_nla_get_ufid(&sfid, a[OVS_FLOW_ATTR_UFID], log);
1105 1106 1107 1108 1109 1110 1111 1112 1113
	if (a[OVS_FLOW_ATTR_KEY]) {
		ovs_match_init(&match, &key, &mask);
		error = ovs_nla_get_match(net, &match, a[OVS_FLOW_ATTR_KEY],
					  a[OVS_FLOW_ATTR_MASK], log);
	} else if (!ufid_present) {
		OVS_NLERR(log,
			  "Flow set message rejected, Key attribute missing.");
		error = -EINVAL;
	}
1114 1115 1116 1117 1118
	if (error)
		goto error;

	/* Validate actions. */
	if (a[OVS_FLOW_ATTR_ACTIONS]) {
1119 1120 1121 1122 1123 1124 1125
		if (!a[OVS_FLOW_ATTR_KEY]) {
			OVS_NLERR(log,
				  "Flow key attribute not present in set flow.");
			error = -EINVAL;
			goto error;
		}

J
Joe Stringer 已提交
1126 1127
		acts = get_flow_actions(net, a[OVS_FLOW_ATTR_ACTIONS], &key,
					&mask, log);
1128 1129
		if (IS_ERR(acts)) {
			error = PTR_ERR(acts);
1130
			goto error;
1131 1132
		}

1133
		/* Can allocate before locking if have acts. */
1134 1135
		reply = ovs_flow_cmd_alloc_info(acts, &sfid, info, false,
						ufid_flags);
1136 1137 1138
		if (IS_ERR(reply)) {
			error = PTR_ERR(reply);
			goto err_kfree_acts;
1139
		}
1140
	}
1141

1142
	ovs_lock();
J
Joe Stringer 已提交
1143
	dp = get_dp(net, ovs_header->dp_ifindex);
1144 1145
	if (unlikely(!dp)) {
		error = -ENODEV;
1146
		goto err_unlock_ovs;
1147
	}
1148
	/* Check that the flow exists. */
1149 1150 1151 1152
	if (ufid_present)
		flow = ovs_flow_tbl_lookup_ufid(&dp->table, &sfid);
	else
		flow = ovs_flow_tbl_lookup_exact(&dp->table, &match);
1153 1154
	if (unlikely(!flow)) {
		error = -ENOENT;
1155
		goto err_unlock_ovs;
1156
	}
1157

1158
	/* Update actions, if present. */
1159
	if (likely(acts)) {
1160 1161
		old_acts = ovsl_dereference(flow->sf_acts);
		rcu_assign_pointer(flow->sf_acts, acts);
1162 1163 1164 1165 1166 1167

		if (unlikely(reply)) {
			error = ovs_flow_cmd_fill_info(flow,
						       ovs_header->dp_ifindex,
						       reply, info->snd_portid,
						       info->snd_seq, 0,
1168 1169
						       OVS_FLOW_CMD_NEW,
						       ufid_flags);
1170 1171 1172 1173 1174
			BUG_ON(error < 0);
		}
	} else {
		/* Could not alloc without acts before locking. */
		reply = ovs_flow_cmd_build_info(flow, ovs_header->dp_ifindex,
1175 1176 1177
						info, OVS_FLOW_CMD_NEW, false,
						ufid_flags);

1178
		if (IS_ERR(reply)) {
1179 1180 1181
			error = PTR_ERR(reply);
			goto err_unlock_ovs;
		}
1182
	}
1183 1184 1185 1186

	/* Clear stats. */
	if (a[OVS_FLOW_ATTR_CLEAR])
		ovs_flow_stats_clear(flow);
1187
	ovs_unlock();
1188

1189 1190 1191
	if (reply)
		ovs_notify(&dp_flow_genl_family, reply, info);
	if (old_acts)
1192
		ovs_nla_free_flow_actions_rcu(old_acts);
1193

1194 1195
	return 0;

1196 1197
err_unlock_ovs:
	ovs_unlock();
1198 1199
	kfree_skb(reply);
err_kfree_acts:
1200
	ovs_nla_free_flow_actions(acts);
1201 1202 1203 1204 1205 1206 1207 1208
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;
1209
	struct net *net = sock_net(skb->sk);
1210 1211 1212 1213
	struct sw_flow_key key;
	struct sk_buff *reply;
	struct sw_flow *flow;
	struct datapath *dp;
1214
	struct sw_flow_match match;
1215 1216 1217
	struct sw_flow_id ufid;
	u32 ufid_flags = ovs_nla_get_ufid_flags(a[OVS_FLOW_ATTR_UFID_FLAGS]);
	int err = 0;
1218
	bool log = !a[OVS_FLOW_ATTR_PROBE];
1219
	bool ufid_present;
1220

1221 1222 1223
	ufid_present = ovs_nla_get_ufid(&ufid, a[OVS_FLOW_ATTR_UFID], log);
	if (a[OVS_FLOW_ATTR_KEY]) {
		ovs_match_init(&match, &key, NULL);
1224
		err = ovs_nla_get_match(net, &match, a[OVS_FLOW_ATTR_KEY], NULL,
1225 1226
					log);
	} else if (!ufid_present) {
1227 1228
		OVS_NLERR(log,
			  "Flow get message rejected, Key attribute missing.");
1229
		err = -EINVAL;
1230
	}
1231 1232 1233
	if (err)
		return err;

1234
	ovs_lock();
1235
	dp = get_dp(sock_net(skb->sk), ovs_header->dp_ifindex);
1236 1237 1238 1239
	if (!dp) {
		err = -ENODEV;
		goto unlock;
	}
1240

1241 1242 1243 1244
	if (ufid_present)
		flow = ovs_flow_tbl_lookup_ufid(&dp->table, &ufid);
	else
		flow = ovs_flow_tbl_lookup_exact(&dp->table, &match);
1245
	if (!flow) {
1246 1247 1248
		err = -ENOENT;
		goto unlock;
	}
1249

1250
	reply = ovs_flow_cmd_build_info(flow, ovs_header->dp_ifindex, info,
1251
					OVS_FLOW_CMD_NEW, true, ufid_flags);
1252 1253 1254 1255
	if (IS_ERR(reply)) {
		err = PTR_ERR(reply);
		goto unlock;
	}
1256

1257
	ovs_unlock();
1258
	return genlmsg_reply(reply, info);
1259 1260 1261
unlock:
	ovs_unlock();
	return err;
1262 1263 1264 1265 1266 1267
}

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;
1268
	struct net *net = sock_net(skb->sk);
1269 1270
	struct sw_flow_key key;
	struct sk_buff *reply;
1271
	struct sw_flow *flow = NULL;
1272
	struct datapath *dp;
1273
	struct sw_flow_match match;
1274 1275
	struct sw_flow_id ufid;
	u32 ufid_flags = ovs_nla_get_ufid_flags(a[OVS_FLOW_ATTR_UFID_FLAGS]);
1276
	int err;
1277
	bool log = !a[OVS_FLOW_ATTR_PROBE];
1278
	bool ufid_present;
1279

1280 1281
	ufid_present = ovs_nla_get_ufid(&ufid, a[OVS_FLOW_ATTR_UFID], log);
	if (a[OVS_FLOW_ATTR_KEY]) {
1282
		ovs_match_init(&match, &key, NULL);
1283 1284
		err = ovs_nla_get_match(net, &match, a[OVS_FLOW_ATTR_KEY],
					NULL, log);
1285 1286 1287 1288
		if (unlikely(err))
			return err;
	}

1289
	ovs_lock();
1290
	dp = get_dp(sock_net(skb->sk), ovs_header->dp_ifindex);
1291
	if (unlikely(!dp)) {
1292 1293 1294
		err = -ENODEV;
		goto unlock;
	}
1295

1296
	if (unlikely(!a[OVS_FLOW_ATTR_KEY] && !ufid_present)) {
1297
		err = ovs_flow_tbl_flush(&dp->table);
1298 1299
		goto unlock;
	}
1300

1301 1302 1303 1304
	if (ufid_present)
		flow = ovs_flow_tbl_lookup_ufid(&dp->table, &ufid);
	else
		flow = ovs_flow_tbl_lookup_exact(&dp->table, &match);
1305
	if (unlikely(!flow)) {
1306 1307 1308
		err = -ENOENT;
		goto unlock;
	}
1309

1310
	ovs_flow_tbl_remove(&dp->table, flow);
1311
	ovs_unlock();
1312

1313
	reply = ovs_flow_cmd_alloc_info((const struct sw_flow_actions __force *) flow->sf_acts,
1314
					&flow->id, info, false, ufid_flags);
1315 1316 1317 1318 1319 1320
	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,
1321 1322
						     OVS_FLOW_CMD_DEL,
						     ufid_flags);
1323 1324 1325 1326 1327 1328 1329
			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));
		}
1330
	}
1331

1332
	ovs_flow_free(flow, true);
1333
	return 0;
1334 1335 1336
unlock:
	ovs_unlock();
	return err;
1337 1338 1339 1340
}

static int ovs_flow_cmd_dump(struct sk_buff *skb, struct netlink_callback *cb)
{
1341
	struct nlattr *a[__OVS_FLOW_ATTR_MAX];
1342
	struct ovs_header *ovs_header = genlmsg_data(nlmsg_data(cb->nlh));
1343
	struct table_instance *ti;
1344
	struct datapath *dp;
1345 1346 1347 1348 1349 1350 1351 1352
	u32 ufid_flags;
	int err;

	err = genlmsg_parse(cb->nlh, &dp_flow_genl_family, a,
			    OVS_FLOW_ATTR_MAX, flow_policy);
	if (err)
		return err;
	ufid_flags = ovs_nla_get_ufid_flags(a[OVS_FLOW_ATTR_UFID_FLAGS]);
1353

1354
	rcu_read_lock();
1355
	dp = get_dp_rcu(sock_net(skb->sk), ovs_header->dp_ifindex);
1356
	if (!dp) {
1357
		rcu_read_unlock();
1358
		return -ENODEV;
1359
	}
1360

1361
	ti = rcu_dereference(dp->table.ti);
1362 1363 1364 1365 1366 1367
	for (;;) {
		struct sw_flow *flow;
		u32 bucket, obj;

		bucket = cb->args[0];
		obj = cb->args[1];
1368
		flow = ovs_flow_tbl_dump_next(ti, &bucket, &obj);
1369 1370 1371
		if (!flow)
			break;

1372
		if (ovs_flow_cmd_fill_info(flow, ovs_header->dp_ifindex, skb,
1373
					   NETLINK_CB(cb->skb).portid,
1374
					   cb->nlh->nlmsg_seq, NLM_F_MULTI,
1375
					   OVS_FLOW_CMD_NEW, ufid_flags) < 0)
1376 1377 1378 1379 1380
			break;

		cb->args[0] = bucket;
		cb->args[1] = obj;
	}
1381
	rcu_read_unlock();
1382 1383 1384
	return skb->len;
}

1385 1386
static const struct nla_policy flow_policy[OVS_FLOW_ATTR_MAX + 1] = {
	[OVS_FLOW_ATTR_KEY] = { .type = NLA_NESTED },
1387
	[OVS_FLOW_ATTR_MASK] = { .type = NLA_NESTED },
1388 1389
	[OVS_FLOW_ATTR_ACTIONS] = { .type = NLA_NESTED },
	[OVS_FLOW_ATTR_CLEAR] = { .type = NLA_FLAG },
1390
	[OVS_FLOW_ATTR_PROBE] = { .type = NLA_FLAG },
1391 1392
	[OVS_FLOW_ATTR_UFID] = { .type = NLA_UNSPEC, .len = 1 },
	[OVS_FLOW_ATTR_UFID_FLAGS] = { .type = NLA_U32 },
1393 1394
};

1395
static const struct genl_ops dp_flow_genl_ops[] = {
1396
	{ .cmd = OVS_FLOW_CMD_NEW,
1397
	  .flags = GENL_UNS_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
1398
	  .policy = flow_policy,
1399
	  .doit = ovs_flow_cmd_new
1400 1401
	},
	{ .cmd = OVS_FLOW_CMD_DEL,
1402
	  .flags = GENL_UNS_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
1403 1404 1405 1406 1407 1408 1409 1410 1411 1412
	  .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,
1413
	  .flags = GENL_UNS_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
1414
	  .policy = flow_policy,
1415
	  .doit = ovs_flow_cmd_set,
1416 1417 1418
	},
};

1419
static struct genl_family dp_flow_genl_family = {
1420 1421
	.id = GENL_ID_GENERATE,
	.hdrsize = sizeof(struct ovs_header),
1422 1423 1424
	.name = OVS_FLOW_FAMILY,
	.version = OVS_FLOW_VERSION,
	.maxattr = OVS_FLOW_ATTR_MAX,
1425 1426
	.netnsok = true,
	.parallel_ops = true,
1427 1428 1429 1430
	.ops = dp_flow_genl_ops,
	.n_ops = ARRAY_SIZE(dp_flow_genl_ops),
	.mcgrps = &ovs_dp_flow_multicast_group,
	.n_mcgrps = 1,
1431 1432
};

1433 1434 1435 1436 1437 1438
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));
1439
	msgsize += nla_total_size(sizeof(struct ovs_dp_megaflow_stats));
1440
	msgsize += nla_total_size(sizeof(u32)); /* OVS_DP_ATTR_USER_FEATURES */
1441 1442 1443 1444

	return msgsize;
}

1445
/* Called with ovs_mutex. */
1446
static int ovs_dp_cmd_fill_info(struct datapath *dp, struct sk_buff *skb,
1447
				u32 portid, u32 seq, u32 flags, u8 cmd)
1448 1449 1450
{
	struct ovs_header *ovs_header;
	struct ovs_dp_stats dp_stats;
1451
	struct ovs_dp_megaflow_stats dp_megaflow_stats;
1452 1453
	int err;

1454
	ovs_header = genlmsg_put(skb, portid, seq, &dp_datapath_genl_family,
1455 1456 1457 1458 1459 1460 1461 1462 1463 1464
				   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;

1465 1466 1467 1468 1469 1470 1471 1472
	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))
1473
		goto nla_put_failure;
1474

1475 1476 1477
	if (nla_put_u32(skb, OVS_DP_ATTR_USER_FEATURES, dp->user_features))
		goto nla_put_failure;

1478 1479
	genlmsg_end(skb, ovs_header);
	return 0;
1480 1481 1482 1483 1484 1485 1486

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

1487
static struct sk_buff *ovs_dp_cmd_alloc_info(void)
1488
{
1489
	return genlmsg_new(ovs_dp_cmd_msg_size(), GFP_KERNEL);
1490 1491
}

J
Jarno Rajahalme 已提交
1492
/* Called with rcu_read_lock or ovs_mutex. */
1493
static struct datapath *lookup_datapath(struct net *net,
1494
					const struct ovs_header *ovs_header,
1495 1496 1497 1498 1499
					struct nlattr *a[OVS_DP_ATTR_MAX + 1])
{
	struct datapath *dp;

	if (!a[OVS_DP_ATTR_NAME])
1500
		dp = get_dp(net, ovs_header->dp_ifindex);
1501 1502 1503
	else {
		struct vport *vport;

1504
		vport = ovs_vport_locate(net, nla_data(a[OVS_DP_ATTR_NAME]));
1505 1506 1507 1508 1509
		dp = vport && vport->port_no == OVSP_LOCAL ? vport->dp : NULL;
	}
	return dp ? dp : ERR_PTR(-ENODEV);
}

1510 1511 1512 1513 1514
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);
1515
	if (IS_ERR(dp))
1516 1517 1518 1519 1520 1521
		return;

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

1522
static void ovs_dp_change(struct datapath *dp, struct nlattr *a[])
1523 1524 1525 1526 1527
{
	if (a[OVS_DP_ATTR_USER_FEATURES])
		dp->user_features = nla_get_u32(a[OVS_DP_ATTR_USER_FEATURES]);
}

1528 1529 1530 1531 1532 1533 1534
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;
1535
	struct ovs_net *ovs_net;
1536
	int err, i;
1537 1538 1539 1540 1541

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

1542
	reply = ovs_dp_cmd_alloc_info();
1543 1544
	if (!reply)
		return -ENOMEM;
1545 1546 1547 1548

	err = -ENOMEM;
	dp = kzalloc(sizeof(*dp), GFP_KERNEL);
	if (dp == NULL)
1549
		goto err_free_reply;
1550

1551
	ovs_dp_set_net(dp, sock_net(skb->sk));
1552 1553

	/* Allocate table. */
1554 1555
	err = ovs_flow_tbl_init(&dp->table);
	if (err)
1556 1557
		goto err_free_dp;

1558
	dp->stats_percpu = netdev_alloc_pcpu_stats(struct dp_stats_percpu);
1559 1560 1561 1562 1563
	if (!dp->stats_percpu) {
		err = -ENOMEM;
		goto err_destroy_table;
	}

1564
	dp->ports = kmalloc(DP_VPORT_HASH_BUCKETS * sizeof(struct hlist_head),
1565
			    GFP_KERNEL);
1566 1567 1568 1569 1570 1571 1572 1573
	if (!dp->ports) {
		err = -ENOMEM;
		goto err_destroy_percpu;
	}

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

1574 1575 1576 1577 1578 1579
	/* 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;
1580
	parms.upcall_portids = a[OVS_DP_ATTR_UPCALL_PID];
1581

1582 1583
	ovs_dp_change(dp, a);

1584 1585 1586
	/* So far only local changes have been made, now need the lock. */
	ovs_lock();

1587 1588 1589 1590 1591 1592
	vport = new_vport(&parms);
	if (IS_ERR(vport)) {
		err = PTR_ERR(vport);
		if (err == -EBUSY)
			err = -EEXIST;

1593 1594 1595 1596 1597 1598 1599 1600 1601
		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);
		}

1602
		goto err_destroy_ports_array;
1603 1604
	}

1605 1606 1607
	err = ovs_dp_cmd_fill_info(dp, reply, info->snd_portid,
				   info->snd_seq, 0, OVS_DP_CMD_NEW);
	BUG_ON(err < 0);
1608

1609
	ovs_net = net_generic(ovs_dp_get_net(dp), ovs_net_id);
1610
	list_add_tail_rcu(&dp->list_node, &ovs_net->dps);
1611 1612

	ovs_unlock();
1613

1614
	ovs_notify(&dp_datapath_genl_family, reply, info);
1615 1616
	return 0;

1617
err_destroy_ports_array:
1618
	ovs_unlock();
1619
	kfree(dp->ports);
1620 1621 1622
err_destroy_percpu:
	free_percpu(dp->stats_percpu);
err_destroy_table:
1623
	ovs_flow_tbl_destroy(&dp->table);
1624 1625
err_free_dp:
	kfree(dp);
1626 1627
err_free_reply:
	kfree_skb(reply);
1628 1629 1630 1631
err:
	return err;
}

1632
/* Called with ovs_mutex. */
1633
static void __dp_destroy(struct datapath *dp)
1634
{
1635
	int i;
1636

1637 1638
	for (i = 0; i < DP_VPORT_HASH_BUCKETS; i++) {
		struct vport *vport;
1639
		struct hlist_node *n;
1640

1641
		hlist_for_each_entry_safe(vport, n, &dp->ports[i], dp_hash_node)
1642 1643 1644
			if (vport->port_no != OVSP_LOCAL)
				ovs_dp_detach_port(vport);
	}
1645

1646
	list_del_rcu(&dp->list_node);
1647

1648
	/* OVSP_LOCAL is datapath internal port. We need to make sure that
1649
	 * all ports in datapath are destroyed first before freeing datapath.
1650
	 */
1651
	ovs_dp_detach_port(ovs_vport_ovsl(dp, OVSP_LOCAL));
1652

1653
	/* RCU destroy the flow table */
1654
	call_rcu(&dp->rcu, destroy_dp_rcu);
1655 1656 1657 1658 1659 1660 1661 1662
}

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

1663
	reply = ovs_dp_cmd_alloc_info();
1664 1665 1666
	if (!reply)
		return -ENOMEM;

1667
	ovs_lock();
1668 1669 1670
	dp = lookup_datapath(sock_net(skb->sk), info->userhdr, info->attrs);
	err = PTR_ERR(dp);
	if (IS_ERR(dp))
1671
		goto err_unlock_free;
1672

1673 1674 1675
	err = ovs_dp_cmd_fill_info(dp, reply, info->snd_portid,
				   info->snd_seq, 0, OVS_DP_CMD_DEL);
	BUG_ON(err < 0);
1676 1677

	__dp_destroy(dp);
1678
	ovs_unlock();
1679

1680
	ovs_notify(&dp_datapath_genl_family, reply, info);
1681 1682

	return 0;
1683 1684

err_unlock_free:
1685
	ovs_unlock();
1686
	kfree_skb(reply);
1687
	return err;
1688 1689 1690 1691 1692 1693 1694 1695
}

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

1696
	reply = ovs_dp_cmd_alloc_info();
1697 1698 1699
	if (!reply)
		return -ENOMEM;

1700
	ovs_lock();
1701
	dp = lookup_datapath(sock_net(skb->sk), info->userhdr, info->attrs);
1702
	err = PTR_ERR(dp);
1703
	if (IS_ERR(dp))
1704
		goto err_unlock_free;
1705

1706 1707
	ovs_dp_change(dp, info->attrs);

1708 1709 1710
	err = ovs_dp_cmd_fill_info(dp, reply, info->snd_portid,
				   info->snd_seq, 0, OVS_DP_CMD_NEW);
	BUG_ON(err < 0);
1711

1712
	ovs_unlock();
1713
	ovs_notify(&dp_datapath_genl_family, reply, info);
1714 1715

	return 0;
1716 1717

err_unlock_free:
1718
	ovs_unlock();
1719
	kfree_skb(reply);
1720
	return err;
1721 1722 1723 1724 1725 1726
}

static int ovs_dp_cmd_get(struct sk_buff *skb, struct genl_info *info)
{
	struct sk_buff *reply;
	struct datapath *dp;
1727
	int err;
1728

1729
	reply = ovs_dp_cmd_alloc_info();
1730 1731 1732
	if (!reply)
		return -ENOMEM;

1733
	ovs_lock();
1734
	dp = lookup_datapath(sock_net(skb->sk), info->userhdr, info->attrs);
1735 1736
	if (IS_ERR(dp)) {
		err = PTR_ERR(dp);
1737
		goto err_unlock_free;
1738
	}
1739 1740 1741
	err = ovs_dp_cmd_fill_info(dp, reply, info->snd_portid,
				   info->snd_seq, 0, OVS_DP_CMD_NEW);
	BUG_ON(err < 0);
1742
	ovs_unlock();
1743 1744

	return genlmsg_reply(reply, info);
1745

1746
err_unlock_free:
1747
	ovs_unlock();
1748
	kfree_skb(reply);
1749
	return err;
1750 1751 1752 1753
}

static int ovs_dp_cmd_dump(struct sk_buff *skb, struct netlink_callback *cb)
{
1754
	struct ovs_net *ovs_net = net_generic(sock_net(skb->sk), ovs_net_id);
1755 1756 1757 1758
	struct datapath *dp;
	int skip = cb->args[0];
	int i = 0;

1759 1760
	ovs_lock();
	list_for_each_entry(dp, &ovs_net->dps, list_node) {
1761
		if (i >= skip &&
1762
		    ovs_dp_cmd_fill_info(dp, skb, NETLINK_CB(cb->skb).portid,
1763 1764 1765 1766 1767
					 cb->nlh->nlmsg_seq, NLM_F_MULTI,
					 OVS_DP_CMD_NEW) < 0)
			break;
		i++;
	}
1768
	ovs_unlock();
1769 1770 1771 1772 1773 1774

	cb->args[0] = i;

	return skb->len;
}

1775 1776 1777 1778 1779 1780
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 },
};

1781
static const struct genl_ops dp_datapath_genl_ops[] = {
1782
	{ .cmd = OVS_DP_CMD_NEW,
1783
	  .flags = GENL_UNS_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
1784 1785 1786 1787
	  .policy = datapath_policy,
	  .doit = ovs_dp_cmd_new
	},
	{ .cmd = OVS_DP_CMD_DEL,
1788
	  .flags = GENL_UNS_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
1789 1790 1791 1792 1793 1794 1795 1796 1797 1798
	  .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,
1799
	  .flags = GENL_UNS_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
1800 1801 1802 1803 1804
	  .policy = datapath_policy,
	  .doit = ovs_dp_cmd_set,
	},
};

1805
static struct genl_family dp_datapath_genl_family = {
1806 1807
	.id = GENL_ID_GENERATE,
	.hdrsize = sizeof(struct ovs_header),
1808 1809 1810
	.name = OVS_DATAPATH_FAMILY,
	.version = OVS_DATAPATH_VERSION,
	.maxattr = OVS_DP_ATTR_MAX,
1811 1812
	.netnsok = true,
	.parallel_ops = true,
1813 1814 1815 1816
	.ops = dp_datapath_genl_ops,
	.n_ops = ARRAY_SIZE(dp_datapath_genl_ops),
	.mcgrps = &ovs_dp_datapath_multicast_group,
	.n_mcgrps = 1,
1817 1818
};

1819
/* Called with ovs_mutex or RCU read lock. */
1820
static int ovs_vport_cmd_fill_info(struct vport *vport, struct sk_buff *skb,
1821
				   u32 portid, u32 seq, u32 flags, u8 cmd)
1822 1823 1824 1825 1826
{
	struct ovs_header *ovs_header;
	struct ovs_vport_stats vport_stats;
	int err;

1827
	ovs_header = genlmsg_put(skb, portid, seq, &dp_vport_genl_family,
1828 1829 1830 1831 1832 1833
				 flags, cmd);
	if (!ovs_header)
		return -EMSGSIZE;

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

1834 1835
	if (nla_put_u32(skb, OVS_VPORT_ATTR_PORT_NO, vport->port_no) ||
	    nla_put_u32(skb, OVS_VPORT_ATTR_TYPE, vport->ops->type) ||
1836
	    nla_put_string(skb, OVS_VPORT_ATTR_NAME,
1837
			   ovs_vport_name(vport)))
1838
		goto nla_put_failure;
1839 1840

	ovs_vport_get_stats(vport, &vport_stats);
1841 1842 1843
	if (nla_put(skb, OVS_VPORT_ATTR_STATS, sizeof(struct ovs_vport_stats),
		    &vport_stats))
		goto nla_put_failure;
1844

1845 1846 1847
	if (ovs_vport_get_upcall_portids(vport, skb))
		goto nla_put_failure;

1848 1849 1850 1851
	err = ovs_vport_get_options(vport, skb);
	if (err == -EMSGSIZE)
		goto error;

1852 1853
	genlmsg_end(skb, ovs_header);
	return 0;
1854 1855 1856 1857 1858 1859 1860 1861

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

1862 1863 1864 1865 1866 1867
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(). */
1868
struct sk_buff *ovs_vport_cmd_build_info(struct vport *vport, u32 portid,
1869 1870 1871 1872 1873 1874 1875 1876 1877
					 u32 seq, u8 cmd)
{
	struct sk_buff *skb;
	int retval;

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

1878
	retval = ovs_vport_cmd_fill_info(vport, skb, portid, seq, 0, cmd);
1879 1880
	BUG_ON(retval < 0);

1881 1882 1883
	return skb;
}

1884
/* Called with ovs_mutex or RCU read lock. */
1885
static struct vport *lookup_vport(struct net *net,
1886
				  const struct ovs_header *ovs_header,
1887 1888 1889 1890 1891 1892
				  struct nlattr *a[OVS_VPORT_ATTR_MAX + 1])
{
	struct datapath *dp;
	struct vport *vport;

	if (a[OVS_VPORT_ATTR_NAME]) {
1893
		vport = ovs_vport_locate(net, nla_data(a[OVS_VPORT_ATTR_NAME]));
1894 1895
		if (!vport)
			return ERR_PTR(-ENODEV);
1896 1897 1898
		if (ovs_header->dp_ifindex &&
		    ovs_header->dp_ifindex != get_dpifindex(vport->dp))
			return ERR_PTR(-ENODEV);
1899 1900 1901 1902 1903 1904 1905
		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);

1906
		dp = get_dp(net, ovs_header->dp_ifindex);
1907 1908 1909
		if (!dp)
			return ERR_PTR(-ENODEV);

1910
		vport = ovs_vport_ovsl_rcu(dp, port_no);
1911
		if (!vport)
1912
			return ERR_PTR(-ENODEV);
1913 1914 1915 1916 1917
		return vport;
	} else
		return ERR_PTR(-EINVAL);
}

1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940
/* Called with ovs_mutex */
static void update_headroom(struct datapath *dp)
{
	unsigned dev_headroom, max_headroom = 0;
	struct net_device *dev;
	struct vport *vport;
	int i;

	for (i = 0; i < DP_VPORT_HASH_BUCKETS; i++) {
		hlist_for_each_entry_rcu(vport, &dp->ports[i], dp_hash_node) {
			dev = vport->dev;
			dev_headroom = netdev_get_fwd_headroom(dev);
			if (dev_headroom > max_headroom)
				max_headroom = dev_headroom;
		}
	}

	dp->max_headroom = max_headroom;
	for (i = 0; i < DP_VPORT_HASH_BUCKETS; i++)
		hlist_for_each_entry_rcu(vport, &dp->ports[i], dp_hash_node)
			netdev_set_rx_headroom(vport->dev, max_headroom);
}

1941 1942 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953
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])
1954 1955 1956 1957 1958 1959 1960 1961 1962 1963
		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;
1964

1965
	ovs_lock();
1966
restart:
1967
	dp = get_dp(sock_net(skb->sk), ovs_header->dp_ifindex);
1968 1969
	err = -ENODEV;
	if (!dp)
1970
		goto exit_unlock_free;
1971

1972
	if (port_no) {
1973
		vport = ovs_vport_ovsl(dp, port_no);
1974 1975
		err = -EBUSY;
		if (vport)
1976
			goto exit_unlock_free;
1977 1978 1979 1980
	} else {
		for (port_no = 1; ; port_no++) {
			if (port_no >= DP_MAX_PORTS) {
				err = -EFBIG;
1981
				goto exit_unlock_free;
1982
			}
1983
			vport = ovs_vport_ovsl(dp, port_no);
1984 1985 1986 1987 1988 1989 1990 1991 1992 1993
			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;
1994
	parms.upcall_portids = a[OVS_VPORT_ATTR_UPCALL_PID];
1995 1996 1997

	vport = new_vport(&parms);
	err = PTR_ERR(vport);
1998 1999 2000
	if (IS_ERR(vport)) {
		if (err == -EAGAIN)
			goto restart;
2001
		goto exit_unlock_free;
2002
	}
2003

2004 2005
	err = ovs_vport_cmd_fill_info(vport, reply, info->snd_portid,
				      info->snd_seq, 0, OVS_VPORT_CMD_NEW);
2006 2007 2008 2009 2010 2011

	if (netdev_get_fwd_headroom(vport->dev) > dp->max_headroom)
		update_headroom(dp);
	else
		netdev_set_rx_headroom(vport->dev, dp->max_headroom);

2012 2013
	BUG_ON(err < 0);
	ovs_unlock();
2014

2015
	ovs_notify(&dp_vport_genl_family, reply, info);
2016
	return 0;
2017

2018
exit_unlock_free:
2019
	ovs_unlock();
2020
	kfree_skb(reply);
2021 2022 2023 2024 2025 2026 2027 2028 2029 2030
	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;

2031 2032 2033 2034
	reply = ovs_vport_cmd_alloc_info();
	if (!reply)
		return -ENOMEM;

2035
	ovs_lock();
2036
	vport = lookup_vport(sock_net(skb->sk), info->userhdr, a);
2037 2038
	err = PTR_ERR(vport);
	if (IS_ERR(vport))
2039
		goto exit_unlock_free;
2040 2041

	if (a[OVS_VPORT_ATTR_TYPE] &&
2042
	    nla_get_u32(a[OVS_VPORT_ATTR_TYPE]) != vport->ops->type) {
2043
		err = -EINVAL;
2044
		goto exit_unlock_free;
2045 2046
	}

2047
	if (a[OVS_VPORT_ATTR_OPTIONS]) {
2048
		err = ovs_vport_set_options(vport, a[OVS_VPORT_ATTR_OPTIONS]);
2049
		if (err)
2050
			goto exit_unlock_free;
2051
	}
2052

2053 2054 2055 2056 2057 2058 2059 2060

	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;
	}
2061

2062 2063 2064
	err = ovs_vport_cmd_fill_info(vport, reply, info->snd_portid,
				      info->snd_seq, 0, OVS_VPORT_CMD_NEW);
	BUG_ON(err < 0);
2065

2066
	ovs_unlock();
2067
	ovs_notify(&dp_vport_genl_family, reply, info);
2068
	return 0;
2069

2070
exit_unlock_free:
2071
	ovs_unlock();
2072
	kfree_skb(reply);
2073 2074 2075 2076 2077
	return err;
}

static int ovs_vport_cmd_del(struct sk_buff *skb, struct genl_info *info)
{
2078
	bool must_update_headroom = false;
2079 2080
	struct nlattr **a = info->attrs;
	struct sk_buff *reply;
2081
	struct datapath *dp;
2082 2083 2084
	struct vport *vport;
	int err;

2085 2086 2087 2088
	reply = ovs_vport_cmd_alloc_info();
	if (!reply)
		return -ENOMEM;

2089
	ovs_lock();
2090
	vport = lookup_vport(sock_net(skb->sk), info->userhdr, a);
2091 2092
	err = PTR_ERR(vport);
	if (IS_ERR(vport))
2093
		goto exit_unlock_free;
2094 2095 2096

	if (vport->port_no == OVSP_LOCAL) {
		err = -EINVAL;
2097
		goto exit_unlock_free;
2098 2099
	}

2100 2101 2102
	err = ovs_vport_cmd_fill_info(vport, reply, info->snd_portid,
				      info->snd_seq, 0, OVS_VPORT_CMD_DEL);
	BUG_ON(err < 0);
2103 2104 2105 2106 2107 2108

	/* the vport deletion may trigger dp headroom update */
	dp = vport->dp;
	if (netdev_get_fwd_headroom(vport->dev) == dp->max_headroom)
		must_update_headroom = true;
	netdev_reset_rx_headroom(vport->dev);
2109
	ovs_dp_detach_port(vport);
2110 2111 2112

	if (must_update_headroom)
		update_headroom(dp);
2113
	ovs_unlock();
2114

2115
	ovs_notify(&dp_vport_genl_family, reply, info);
2116
	return 0;
2117

2118
exit_unlock_free:
2119
	ovs_unlock();
2120
	kfree_skb(reply);
2121 2122 2123 2124 2125 2126 2127 2128 2129 2130 2131
	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;

2132 2133 2134 2135
	reply = ovs_vport_cmd_alloc_info();
	if (!reply)
		return -ENOMEM;

2136
	rcu_read_lock();
2137
	vport = lookup_vport(sock_net(skb->sk), ovs_header, a);
2138 2139
	err = PTR_ERR(vport);
	if (IS_ERR(vport))
2140 2141 2142 2143
		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);
2144 2145 2146 2147
	rcu_read_unlock();

	return genlmsg_reply(reply, info);

2148
exit_unlock_free:
2149
	rcu_read_unlock();
2150
	kfree_skb(reply);
2151 2152 2153 2154 2155 2156 2157
	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;
2158 2159
	int bucket = cb->args[0], skip = cb->args[1];
	int i, j = 0;
2160

J
Jarno Rajahalme 已提交
2161
	rcu_read_lock();
2162
	dp = get_dp_rcu(sock_net(skb->sk), ovs_header->dp_ifindex);
J
Jarno Rajahalme 已提交
2163 2164
	if (!dp) {
		rcu_read_unlock();
2165
		return -ENODEV;
J
Jarno Rajahalme 已提交
2166
	}
2167
	for (i = bucket; i < DP_VPORT_HASH_BUCKETS; i++) {
2168
		struct vport *vport;
2169 2170

		j = 0;
2171
		hlist_for_each_entry_rcu(vport, &dp->ports[i], dp_hash_node) {
2172 2173
			if (j >= skip &&
			    ovs_vport_cmd_fill_info(vport, skb,
2174
						    NETLINK_CB(cb->skb).portid,
2175 2176 2177 2178 2179 2180 2181 2182
						    cb->nlh->nlmsg_seq,
						    NLM_F_MULTI,
						    OVS_VPORT_CMD_NEW) < 0)
				goto out;

			j++;
		}
		skip = 0;
2183
	}
2184
out:
2185 2186
	rcu_read_unlock();

2187 2188
	cb->args[0] = i;
	cb->args[1] = j;
2189

2190
	return skb->len;
2191 2192
}

2193 2194 2195 2196 2197 2198 2199 2200 2201
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 },
};

2202
static const struct genl_ops dp_vport_genl_ops[] = {
2203
	{ .cmd = OVS_VPORT_CMD_NEW,
2204
	  .flags = GENL_UNS_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
2205 2206 2207 2208
	  .policy = vport_policy,
	  .doit = ovs_vport_cmd_new
	},
	{ .cmd = OVS_VPORT_CMD_DEL,
2209
	  .flags = GENL_UNS_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
2210 2211 2212 2213 2214 2215 2216 2217 2218 2219
	  .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,
2220
	  .flags = GENL_UNS_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
2221 2222 2223 2224 2225
	  .policy = vport_policy,
	  .doit = ovs_vport_cmd_set,
	},
};

2226 2227 2228 2229 2230 2231 2232 2233 2234 2235 2236 2237
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,
2238 2239
};

2240 2241 2242 2243 2244
static struct genl_family * const dp_genl_families[] = {
	&dp_datapath_genl_family,
	&dp_vport_genl_family,
	&dp_flow_genl_family,
	&dp_packet_genl_family,
2245 2246 2247 2248 2249 2250 2251
};

static void dp_unregister_genl(int n_families)
{
	int i;

	for (i = 0; i < n_families; i++)
2252
		genl_unregister_family(dp_genl_families[i]);
2253 2254 2255 2256 2257 2258 2259 2260 2261
}

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

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

2262
		err = genl_register_family(dp_genl_families[i]);
2263 2264 2265 2266 2267 2268 2269
		if (err)
			goto error;
	}

	return 0;

error:
2270
	dp_unregister_genl(i);
2271 2272 2273
	return err;
}

2274 2275 2276 2277 2278
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);
2279
	INIT_WORK(&ovs_net->dp_notify_work, ovs_dp_notify_wq);
2280
	ovs_ct_init(net);
2281 2282 2283
	return 0;
}

P
Pravin B Shelar 已提交
2284 2285
static void __net_exit list_vports_from_net(struct net *net, struct net *dnet,
					    struct list_head *head)
2286
{
2287
	struct ovs_net *ovs_net = net_generic(net, ovs_net_id);
P
Pravin B Shelar 已提交
2288 2289 2290 2291 2292 2293 2294 2295 2296 2297 2298 2299
	struct datapath *dp;

	list_for_each_entry(dp, &ovs_net->dps, list_node) {
		int i;

		for (i = 0; i < DP_VPORT_HASH_BUCKETS; i++) {
			struct vport *vport;

			hlist_for_each_entry(vport, &dp->ports[i], dp_hash_node) {
				if (vport->ops->type != OVS_VPORT_TYPE_INTERNAL)
					continue;

2300
				if (dev_net(vport->dev) == dnet)
P
Pravin B Shelar 已提交
2301 2302 2303 2304 2305 2306 2307 2308 2309 2310 2311 2312 2313
					list_add(&vport->detach_list, head);
			}
		}
	}
}

static void __net_exit ovs_exit_net(struct net *dnet)
{
	struct datapath *dp, *dp_next;
	struct ovs_net *ovs_net = net_generic(dnet, ovs_net_id);
	struct vport *vport, *vport_next;
	struct net *net;
	LIST_HEAD(head);
2314

2315
	ovs_ct_exit(dnet);
2316
	ovs_lock();
2317 2318
	list_for_each_entry_safe(dp, dp_next, &ovs_net->dps, list_node)
		__dp_destroy(dp);
P
Pravin B Shelar 已提交
2319 2320 2321 2322 2323 2324 2325 2326 2327 2328 2329 2330

	rtnl_lock();
	for_each_net(net)
		list_vports_from_net(net, dnet, &head);
	rtnl_unlock();

	/* Detach all vports from given namespace. */
	list_for_each_entry_safe(vport, vport_next, &head, detach_list) {
		list_del(&vport->detach_list);
		ovs_dp_detach_port(vport);
	}

2331 2332 2333
	ovs_unlock();

	cancel_work_sync(&ovs_net->dp_notify_work);
2334 2335 2336 2337 2338 2339 2340 2341 2342
}

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

2343 2344 2345 2346
static int __init dp_init(void)
{
	int err;

2347
	BUILD_BUG_ON(sizeof(struct ovs_skb_cb) > FIELD_SIZEOF(struct sk_buff, cb));
2348 2349 2350

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

2351
	err = action_fifos_init();
2352 2353 2354
	if (err)
		goto error;

2355 2356 2357 2358
	err = ovs_internal_dev_rtnl_link_register();
	if (err)
		goto error_action_fifos_exit;

2359 2360 2361 2362
	err = ovs_flow_init();
	if (err)
		goto error_unreg_rtnl_link;

2363 2364 2365 2366
	err = ovs_vport_init();
	if (err)
		goto error_flow_exit;

2367
	err = register_pernet_device(&ovs_net_ops);
2368 2369 2370
	if (err)
		goto error_vport_exit;

2371 2372 2373 2374
	err = register_netdevice_notifier(&ovs_dp_device_notifier);
	if (err)
		goto error_netns_exit;

2375 2376 2377 2378
	err = ovs_netdev_init();
	if (err)
		goto error_unreg_notifier;

2379 2380
	err = dp_register_genl();
	if (err < 0)
2381
		goto error_unreg_netdev;
2382 2383 2384

	return 0;

2385 2386
error_unreg_netdev:
	ovs_netdev_exit();
2387 2388
error_unreg_notifier:
	unregister_netdevice_notifier(&ovs_dp_device_notifier);
2389 2390
error_netns_exit:
	unregister_pernet_device(&ovs_net_ops);
2391 2392 2393 2394
error_vport_exit:
	ovs_vport_exit();
error_flow_exit:
	ovs_flow_exit();
2395 2396
error_unreg_rtnl_link:
	ovs_internal_dev_rtnl_link_unregister();
2397 2398
error_action_fifos_exit:
	action_fifos_exit();
2399 2400 2401 2402 2403 2404 2405
error:
	return err;
}

static void dp_cleanup(void)
{
	dp_unregister_genl(ARRAY_SIZE(dp_genl_families));
2406
	ovs_netdev_exit();
2407
	unregister_netdevice_notifier(&ovs_dp_device_notifier);
2408 2409
	unregister_pernet_device(&ovs_net_ops);
	rcu_barrier();
2410 2411
	ovs_vport_exit();
	ovs_flow_exit();
2412
	ovs_internal_dev_rtnl_link_unregister();
2413
	action_fifos_exit();
2414 2415 2416 2417 2418 2419 2420
}

module_init(dp_init);
module_exit(dp_cleanup);

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