]> git.proxmox.com Git - mirror_ovs.git/blame - datapath/datapath.c
projects / mirror_ovs.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
datapath: Add conntrack limit netlink definition
[mirror_ovs.git] / datapath / datapath.c
CommitLineData
064af421 1/*
e23775f2 2 * Copyright (c) 2007-2015 Nicira, Inc.
a14bc59f 3 *
a9a29d22
JG		 4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of version 2 of the GNU General Public
6 * License as published by the Free Software Foundation.
7 *
8 * This program is distributed in the hope that it will be useful, but
9 * WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
12 *
13 * You should have received a copy of the GNU General Public License
14 * along with this program; if not, write to the Free Software
15 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
16 * 02110-1301, USA
064af421
BP		 17 */
18
dfffaef1
JP		 19#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
20
064af421
BP		 21#include <linux/init.h>
22#include <linux/module.h>
064af421 23#include <linux/if_arp.h>
064af421
BP		 24#include <linux/if_vlan.h>
25#include <linux/in.h>
26#include <linux/ip.h>
982b8810 27#include <linux/jhash.h>
064af421
BP		 28#include <linux/delay.h>
29#include <linux/time.h>
30#include <linux/etherdevice.h>
ed099e92 31#include <linux/genetlink.h>
064af421
BP		 32#include <linux/kernel.h>
33#include <linux/kthread.h>
064af421
BP		 34#include <linux/mutex.h>
35#include <linux/percpu.h>
36#include <linux/rcupdate.h>
37#include <linux/tcp.h>
38#include <linux/udp.h>
39#include <linux/version.h>
40#include <linux/ethtool.h>
064af421 41#include <linux/wait.h>
064af421 42#include <asm/div64.h>
656a0e37 43#include <linux/highmem.h>
064af421
BP		 44#include <linux/netfilter_bridge.h>
45#include <linux/netfilter_ipv4.h>
46#include <linux/inetdevice.h>
47#include <linux/list.h>
077257b8 48#include <linux/openvswitch.h>
064af421 49#include <linux/rculist.h>
064af421 50#include <linux/dmi.h>
36956a7d 51#include <net/genetlink.h>
2a4999f3
PS		 52#include <net/net_namespace.h>
53#include <net/netns/generic.h>
907c26a8 54#include <net/nsh.h>
064af421 55
064af421 56#include "datapath.h"
038e34ab 57#include "conntrack.h"
064af421 58#include "flow.h"
d103f479 59#include "flow_table.h"
a097c0b2 60#include "flow_netlink.h"
1cb57039 61#include "meter.h"
e23775f2 62#include "gso.h"
f2459fe7 63#include "vport-internal_dev.h"
d5de5b0d 64#include "vport-netdev.h"
064af421 65
f56f0b73 66unsigned int ovs_net_id __read_mostly;
2a4999f3 67
cb25142c
PS		 68static struct genl_family dp_packet_genl_family;
69static struct genl_family dp_flow_genl_family;
70static struct genl_family dp_datapath_genl_family;
71
bc619e29
JS		 72static const struct nla_policy flow_policy[];
73
18fd3a52
PS		 74static struct genl_multicast_group ovs_dp_flow_multicast_group = {
75 .name = OVS_FLOW_MCGROUP
cb25142c
PS		 76};
77
18fd3a52
PS		 78static struct genl_multicast_group ovs_dp_datapath_multicast_group = {
79 .name = OVS_DATAPATH_MCGROUP
cb25142c
PS		 80};
81
18fd3a52
PS		 82struct genl_multicast_group ovs_dp_vport_multicast_group = {
83 .name = OVS_VPORT_MCGROUP
cb25142c
PS		 84};
85
afad3556 86/* Check if need to build a reply message.
af465b67
PS		 87 * OVS userspace sets the NLM_F_ECHO flag if it needs the reply.
88 */
114fce23
SG		 89static bool ovs_must_notify(struct genl_family *family, struct genl_info *info,
90 unsigned int group)
afad3556
JR		 91{
92 return info->nlhdr->nlmsg_flags & NLM_F_ECHO ||
6233a1bd 93 genl_has_listeners(family, genl_info_net(info), group);
afad3556
JR		 94}
95
18fd3a52 96static void ovs_notify(struct genl_family *family, struct genl_multicast_group *grp,
cb25142c 97 struct sk_buff *skb, struct genl_info *info)
e297c6b7 98{
0643a78b 99 genl_notify(family, skb, info, GROUP_ID(grp), GFP_KERNEL);
e297c6b7
TG		 100}
101
ed099e92
BP		 102/**
103 * DOC: Locking:
064af421 104 *
cd2a59e9
PS		 105 * All writes e.g. Writes to device state (add/remove datapath, port, set
106 * operations on vports, etc.), Writes to other state (flow table
107 * modifications, set miscellaneous datapath parameters, etc.) are protected
108 * by ovs_lock.
ed099e92
BP		 109 *
110 * Reads are protected by RCU.
111 *
112 * There are a few special cases (mostly stats) that have their own
113 * synchronization but they nest under all of above and don't interact with
114 * each other.
cd2a59e9
PS		 115 *
116 * The RTNL lock nests inside ovs_mutex.
064af421 117 */
ed099e92 118
cd2a59e9
PS		 119static DEFINE_MUTEX(ovs_mutex);
120
121void ovs_lock(void)
122{
123 mutex_lock(&ovs_mutex);
124}
125
126void ovs_unlock(void)
127{
128 mutex_unlock(&ovs_mutex);
129}
130
131#ifdef CONFIG_LOCKDEP
132int lockdep_ovsl_is_held(void)
133{
134 if (debug_locks)
135 return lockdep_is_held(&ovs_mutex);
136 else
137 return 1;
138}
139#endif
140
5ae440c3 141static int queue_gso_packets(struct datapath *dp, struct sk_buff *,
f1f60b85 142 const struct sw_flow_key *,
4c7804f1
WT		 143 const struct dp_upcall_info *,
144 uint32_t cutlen);
5ae440c3 145static int queue_userspace_packet(struct datapath *dp, struct sk_buff *,
7d16c847 146 const struct sw_flow_key *,
4c7804f1
WT		 147 const struct dp_upcall_info *,
148 uint32_t cutlen);
064af421 149
cd2a59e9 150/* Must be called with rcu_read_lock or ovs_mutex. */
850b6b3b 151const char *ovs_dp_name(const struct datapath *dp)
f2459fe7 152{
cd2a59e9 153 struct vport *vport = ovs_vport_ovsl_rcu(dp, OVSP_LOCAL);
e23775f2 154 return ovs_vport_name(vport);
f2459fe7
JG		 155}
156
f1f60b85 157static int get_dpifindex(const struct datapath *dp)
99769a40
JG		 158{
159 struct vport *local;
160 int ifindex;
161
162 rcu_read_lock();
163
95b1d73a 164 local = ovs_vport_rcu(dp, OVSP_LOCAL);
99769a40 165 if (local)
e23775f2 166 ifindex = local->dev->ifindex;
99769a40
JG		 167 else
168 ifindex = 0;
169
170 rcu_read_unlock();
171
172 return ifindex;
173}
174
46c6a11d
JG		 175static void destroy_dp_rcu(struct rcu_head *rcu)
176{
177 struct datapath *dp = container_of(rcu, struct datapath, rcu);
46c6a11d 178
e379e4d1 179 ovs_flow_tbl_destroy(&dp->table);
46c6a11d 180 free_percpu(dp->stats_percpu);
95b1d73a 181 kfree(dp->ports);
1cb57039 182 ovs_meters_exit(dp);
5ca1ba48 183 kfree(dp);
46c6a11d
JG		 184}
185
95b1d73a
PS		 186static struct hlist_head *vport_hash_bucket(const struct datapath *dp,
187 u16 port_no)
188{
189 return &dp->ports[port_no & (DP_VPORT_HASH_BUCKETS - 1)];
190}
191
aa917006 192/* Called with ovs_mutex or RCU read lock. */
95b1d73a
PS		 193struct vport *ovs_lookup_vport(const struct datapath *dp, u16 port_no)
194{
195 struct vport *vport;
95b1d73a
PS		 196 struct hlist_head *head;
197
198 head = vport_hash_bucket(dp, port_no);
f8dfbcb7 199 hlist_for_each_entry_rcu(vport, head, dp_hash_node) {
95b1d73a
PS		 200 if (vport->port_no == port_no)
201 return vport;
202 }
203 return NULL;
204}
205
cd2a59e9 206/* Called with ovs_mutex. */
c19e6535 207static struct vport *new_vport(const struct vport_parms *parms)
064af421 208{
f2459fe7 209 struct vport *vport;
f2459fe7 210
850b6b3b 211 vport = ovs_vport_add(parms);
c19e6535
BP		 212 if (!IS_ERR(vport)) {
213 struct datapath *dp = parms->dp;
95b1d73a 214 struct hlist_head *head = vport_hash_bucket(dp, vport->port_no);
064af421 215
95b1d73a 216 hlist_add_head_rcu(&vport->dp_hash_node, head);
c19e6535 217 }
c19e6535 218 return vport;
064af421
BP		 219}
220
850b6b3b 221void ovs_dp_detach_port(struct vport *p)
064af421 222{
cd2a59e9 223 ASSERT_OVSL();
064af421 224
064af421 225 /* First drop references to device. */
95b1d73a 226 hlist_del_rcu(&p->dp_hash_node);
f2459fe7 227
7237e4f4 228 /* Then destroy it. */
850b6b3b 229 ovs_vport_del(p);
064af421
BP		 230}
231
fb66fbd1 232/* Must be called with rcu_read_lock. */
e74d4817 233void ovs_dp_process_packet(struct sk_buff *skb, struct sw_flow_key *key)
064af421 234{
a6059080 235 const struct vport *p = OVS_CB(skb)->input_vport;
064af421 236 struct datapath *dp = p->dp;
3544358a 237 struct sw_flow *flow;
ad50cb60 238 struct sw_flow_actions *sf_acts;
064af421 239 struct dp_stats_percpu *stats;
e9141eec 240 u64 *stats_counter;
4fa72a95 241 u32 n_mask_hit;
064af421 242
70dbc259 243 stats = this_cpu_ptr(dp->stats_percpu);
a063b0df 244
52a23d92 245 /* Look up flow. */
e74d4817 246 flow = ovs_flow_tbl_lookup_stats(&dp->table, key, skb_get_hash(skb),
5604935e 247 &n_mask_hit);
52a23d92
JG		 248 if (unlikely(!flow)) {
249 struct dp_upcall_info upcall;
a7d607c5 250 int error;
52a23d92 251
0e469d3b 252 memset(&upcall, 0, sizeof(upcall));
52a23d92 253 upcall.cmd = OVS_PACKET_CMD_MISS;
beb1c69a 254 upcall.portid = ovs_vport_find_upcall_portid(p, skb);
a94ebc39 255 upcall.mru = OVS_CB(skb)->mru;
4c7804f1 256 error = ovs_dp_upcall(dp, skb, key, &upcall, 0);
a7d607c5
LR		 257 if (unlikely(error))
258 kfree_skb(skb);
259 else
260 consume_skb(skb);
52a23d92
JG		 261 stats_counter = &stats->n_missed;
262 goto out;
263 }
264
e74d4817 265 ovs_flow_stats_update(flow, key->tp.flags, skb);
ad50cb60 266 sf_acts = rcu_dereference(flow->sf_acts);
7d16c847
PS		 267 ovs_execute_actions(dp, skb, sf_acts, key);
268
b0b906cc 269 stats_counter = &stats->n_hit;
55574bb0 270
8819fac7 271out:
55574bb0 272 /* Update datapath statistics. */
b81deb15 273 u64_stats_update_begin(&stats->syncp);
e9141eec 274 (*stats_counter)++;
4fa72a95 275 stats->n_mask_hit += n_mask_hit;
b81deb15 276 u64_stats_update_end(&stats->syncp);
064af421
BP		 277}
278
850b6b3b 279int ovs_dp_upcall(struct datapath *dp, struct sk_buff *skb,
f1f60b85 280 const struct sw_flow_key *key,
4c7804f1
WT		 281 const struct dp_upcall_info *upcall_info,
282 uint32_t cutlen)
aa5a8fdc
JG		 283{
284 struct dp_stats_percpu *stats;
285 int err;
286
28aea917 287 if (upcall_info->portid == 0) {
b063d9f0 288 err = -ENOTCONN;
b063d9f0
JG		 289 goto err;
290 }
291
7257b535 292 if (!skb_is_gso(skb))
4c7804f1 293 err = queue_userspace_packet(dp, skb, key, upcall_info, cutlen);
7257b535 294 else
4c7804f1 295 err = queue_gso_packets(dp, skb, key, upcall_info, cutlen);
d76195db
JG		 296 if (err)
297 goto err;
298
299 return 0;
aa5a8fdc 300
aa5a8fdc 301err:
70dbc259 302 stats = this_cpu_ptr(dp->stats_percpu);
aa5a8fdc 303
b81deb15 304 u64_stats_update_begin(&stats->syncp);
aa5a8fdc 305 stats->n_lost++;
b81deb15 306 u64_stats_update_end(&stats->syncp);
aa5a8fdc 307
aa5a8fdc 308 return err;
982b8810
BP		 309}
310
5ae440c3 311static int queue_gso_packets(struct datapath *dp, struct sk_buff *skb,
f1f60b85 312 const struct sw_flow_key *key,
4c7804f1
WT		 313 const struct dp_upcall_info *upcall_info,
314 uint32_t cutlen)
cb5087ca 315{
969b8e6b 316#ifdef HAVE_SKB_GSO_UDP
f54a7a5d 317 unsigned int gso_type = skb_shinfo(skb)->gso_type;
7257b535 318 struct sw_flow_key later_key;
969b8e6b 319#endif
7257b535 320 struct sk_buff *segs, *nskb;
b2a23c4e 321 struct ovs_skb_cb ovs_cb;
7257b535 322 int err;
cb5087ca 323
b2a23c4e 324 ovs_cb = *OVS_CB(skb);
1d04cd4e 325 segs = __skb_gso_segment(skb, NETIF_F_SG, false);
b2a23c4e 326 *OVS_CB(skb) = ovs_cb;
79089764
PS		 327 if (IS_ERR(segs))
328 return PTR_ERR(segs);
d1da7669
PS		 329 if (segs == NULL)
330 return -EINVAL;
969b8e6b 331#ifdef HAVE_SKB_GSO_UDP
9b277b39 332 if (gso_type & SKB_GSO_UDP) {
c135bba1 333 /* The initial flow key extracted by ovs_flow_key_extract()
9b277b39
PS		 334 * in this case is for a first fragment, so we need to
335 * properly mark later fragments.
336 */
e74d4817 337 later_key = *key;
9b277b39
PS		 338 later_key.ip.frag = OVS_FRAG_TYPE_LATER;
339 }
969b8e6b 340#endif
7257b535
BP		 341 /* Queue all of the segments. */
342 skb = segs;
cb5087ca 343 do {
b2a23c4e 344 *OVS_CB(skb) = ovs_cb;
969b8e6b 345#ifdef HAVE_SKB_GSO_UDP
9b277b39 346 if (gso_type & SKB_GSO_UDP && skb != segs)
e74d4817 347 key = &later_key;
969b8e6b 348#endif
4c7804f1 349 err = queue_userspace_packet(dp, skb, key, upcall_info, cutlen);
982b8810 350 if (err)
7257b535 351 break;
856081f6 352
36ce148c 353 } while ((skb = skb->next));
cb5087ca 354
7257b535
BP		 355 /* Free all of the segments. */
356 skb = segs;
357 do {
358 nskb = skb->next;
359 if (err)
360 kfree_skb(skb);
361 else
362 consume_skb(skb);
363 } while ((skb = nskb));
364 return err;
365}
366
8b7ea2d4 367static size_t upcall_msg_size(const struct dp_upcall_info *upcall_info,
6b330a60 368 unsigned int hdrlen, int actions_attrlen)
0afa2373
TG		 369{
370 size_t size = NLMSG_ALIGN(sizeof(struct ovs_header))
533bea51 371 + nla_total_size(hdrlen) /* OVS_PACKET_ATTR_PACKET */
039fb36c
WT		 372 + nla_total_size(ovs_key_attr_size()) /* OVS_PACKET_ATTR_KEY */
373 + nla_total_size(sizeof(unsigned int)); /* OVS_PACKET_ATTR_LEN */
0afa2373
TG		 374
375 /* OVS_PACKET_ATTR_USERDATA */
8b7ea2d4
WZ		 376 if (upcall_info->userdata)
377 size += NLA_ALIGN(upcall_info->userdata->nla_len);
378
379 /* OVS_PACKET_ATTR_EGRESS_TUN_KEY */
380 if (upcall_info->egress_tun_info)
381 size += nla_total_size(ovs_tun_key_attr_size());
0afa2373 382
0e469d3b
NM		 383 /* OVS_PACKET_ATTR_ACTIONS */
384 if (upcall_info->actions_len)
6b330a60 385 size += nla_total_size(actions_attrlen);
0e469d3b 386
a94ebc39
JS		 387 /* OVS_PACKET_ATTR_MRU */
388 if (upcall_info->mru)
389 size += nla_total_size(sizeof(upcall_info->mru));
390
0afa2373
TG		 391 return size;
392}
393
a94ebc39
JS		 394static void pad_packet(struct datapath *dp, struct sk_buff *skb)
395{
396 if (!(dp->user_features & OVS_DP_F_UNALIGNED)) {
397 size_t plen = NLA_ALIGN(skb->len) - skb->len;
398
399 if (plen > 0)
0ace0a26 400 skb_put_zero(skb, plen);
a94ebc39
JS		 401 }
402}
403
5ae440c3 404static int queue_userspace_packet(struct datapath *dp, struct sk_buff *skb,
f1f60b85 405 const struct sw_flow_key *key,
4c7804f1
WT		 406 const struct dp_upcall_info *upcall_info,
407 uint32_t cutlen)
7257b535
BP		 408{
409 struct ovs_header *upcall;
6161d3fd 410 struct sk_buff *nskb = NULL;
82706a6f 411 struct sk_buff *user_skb = NULL; /* to be queued to userspace */
7257b535 412 struct nlattr *nla;
978188b2 413 size_t len;
533bea51 414 unsigned int hlen;
5ae440c3
TG		 415 int err, dp_ifindex;
416
417 dp_ifindex = get_dpifindex(dp);
418 if (!dp_ifindex)
419 return -ENODEV;
7257b535 420
efd8a18e 421 if (skb_vlan_tag_present(skb)) {
6161d3fd
JG		 422 nskb = skb_clone(skb, GFP_ATOMIC);
423 if (!nskb)
424 return -ENOMEM;
07ac71ea 425
8063e095 426 nskb = __vlan_hwaccel_push_inside(nskb);
07ac71ea
PS		 427 if (!nskb)
428 return -ENOMEM;
429
6161d3fd
JG		 430 skb = nskb;
431 }
432
433 if (nla_attr_size(skb->len) > USHRT_MAX) {
434 err = -EFBIG;
435 goto out;
436 }
7257b535 437
533bea51
TG		 438 /* Complete checksum if needed */
439 if (skb->ip_summed == CHECKSUM_PARTIAL &&
a0c9fedc 440 (err = skb_csum_hwoffload_help(skb, 0)))
533bea51
TG		 441 goto out;
442
443 /* Older versions of OVS user space enforce alignment of the last
444 * Netlink attribute to NLA_ALIGNTO which would require extensive
445 * padding logic. Only perform zerocopy if padding is not required.
446 */
447 if (dp->user_features & OVS_DP_F_UNALIGNED)
448 hlen = skb_zerocopy_headlen(skb);
449 else
450 hlen = skb->len;
451
6b330a60
GR		 452 len = upcall_msg_size(upcall_info, hlen - cutlen,
453 OVS_CB(skb)->acts_origlen);
40c08cda 454 user_skb = genlmsg_new(len, GFP_ATOMIC);
6161d3fd
JG		 455 if (!user_skb) {
456 err = -ENOMEM;
457 goto out;
458 }
7257b535
BP		 459
460 upcall = genlmsg_put(user_skb, 0, 0, &dp_packet_genl_family,
461 0, upcall_info->cmd);
462 upcall->dp_ifindex = dp_ifindex;
463
db7f2238 464 err = ovs_nla_put_key(key, key, OVS_PACKET_ATTR_KEY, false, user_skb);
9a621f82 465 BUG_ON(err);
7257b535
BP		 466
467 if (upcall_info->userdata)
e995e3df 468 __nla_put(user_skb, OVS_PACKET_ATTR_USERDATA,
462a988b 469 nla_len(upcall_info->userdata),
e995e3df 470 nla_data(upcall_info->userdata));
7257b535 471
e23775f2 472
8b7ea2d4
WZ		 473 if (upcall_info->egress_tun_info) {
474 nla = nla_nest_start(user_skb, OVS_PACKET_ATTR_EGRESS_TUN_KEY);
aad7cb91
PS		 475 err = ovs_nla_put_tunnel_info(user_skb,
476 upcall_info->egress_tun_info);
8b7ea2d4
WZ		 477 BUG_ON(err);
478 nla_nest_end(user_skb, nla);
479 }
480
0e469d3b
NM		 481 if (upcall_info->actions_len) {
482 nla = nla_nest_start(user_skb, OVS_PACKET_ATTR_ACTIONS);
483 err = ovs_nla_put_actions(upcall_info->actions,
484 upcall_info->actions_len,
485 user_skb);
486 if (!err)
487 nla_nest_end(user_skb, nla);
488 else
489 nla_nest_cancel(user_skb, nla);
490 }
491
a94ebc39
JS		 492 /* Add OVS_PACKET_ATTR_MRU */
493 if (upcall_info->mru) {
494 if (nla_put_u16(user_skb, OVS_PACKET_ATTR_MRU,
495 upcall_info->mru)) {
496 err = -ENOBUFS;
497 goto out;
498 }
499 pad_packet(dp, user_skb);
500 }
501
039fb36c
WT		 502 /* Add OVS_PACKET_ATTR_LEN when packet is truncated */
503 if (cutlen > 0) {
504 if (nla_put_u32(user_skb, OVS_PACKET_ATTR_LEN,
505 skb->len)) {
506 err = -ENOBUFS;
507 goto out;
508 }
509 pad_packet(dp, user_skb);
510 }
511
533bea51 512 /* Only reserve room for attribute header, packet data is added
af465b67
PS		 513 * in skb_zerocopy()
514 */
533bea51
TG		 515 if (!(nla = nla_reserve(user_skb, OVS_PACKET_ATTR_PACKET, 0))) {
516 err = -ENOBUFS;
517 goto out;
518 }
4c7804f1 519 nla->nla_len = nla_attr_size(skb->len - cutlen);
bed53bd1 520
4c7804f1 521 err = skb_zerocopy(user_skb, skb, skb->len - cutlen, hlen);
2c272bd9
ZK		 522 if (err)
523 goto out;
7257b535 524
ef507cec 525 /* Pad OVS_PACKET_ATTR_PACKET if linear copy was performed */
a94ebc39 526 pad_packet(dp, user_skb);
ef507cec 527
533bea51 528 ((struct nlmsghdr *) user_skb->data)->nlmsg_len = user_skb->len;
6161d3fd 529
533bea51 530 err = genlmsg_unicast(ovs_dp_get_net(dp), user_skb, upcall_info->portid);
82706a6f 531 user_skb = NULL;
6161d3fd 532out:
2c272bd9
ZK		 533 if (err)
534 skb_tx_error(skb);
82706a6f 535 kfree_skb(user_skb);
6161d3fd
JG		 536 kfree_skb(nskb);
537 return err;
cb5087ca
BP		 538}
539
df2c07f4 540static int ovs_packet_cmd_execute(struct sk_buff *skb, struct genl_info *info)
064af421 541{
df2c07f4 542 struct ovs_header *ovs_header = info->userhdr;
a94ebc39 543 struct net *net = sock_net(skb->sk);
982b8810 544 struct nlattr **a = info->attrs;
e0e57990 545 struct sw_flow_actions *acts;
982b8810 546 struct sk_buff *packet;
e0e57990 547 struct sw_flow *flow;
ad50cb60 548 struct sw_flow_actions *sf_acts;
f7cd0081 549 struct datapath *dp;
a6059080 550 struct vport *input_vport;
a94ebc39 551 u16 mru = 0;
3f19d399 552 int len;
d6569377 553 int err;
2e460098 554 bool log = !a[OVS_PACKET_ATTR_PROBE];
064af421 555
f7cd0081 556 err = -EINVAL;
df2c07f4 557 if (!a[OVS_PACKET_ATTR_PACKET] || !a[OVS_PACKET_ATTR_KEY] ||
7c3072cc 558 !a[OVS_PACKET_ATTR_ACTIONS])
e5cad958 559 goto err;
064af421 560
df2c07f4 561 len = nla_len(a[OVS_PACKET_ATTR_PACKET]);
3f19d399 562 packet = __dev_alloc_skb(NET_IP_ALIGN + len, GFP_KERNEL);
f7cd0081
BP		 563 err = -ENOMEM;
564 if (!packet)
e5cad958 565 goto err;
3f19d399
BP		 566 skb_reserve(packet, NET_IP_ALIGN);
567
bf3d6fce 568 nla_memcpy(__skb_put(packet, len), a[OVS_PACKET_ATTR_PACKET], len);
8d5ebd83 569
a94ebc39
JS		 570 /* Set packet's mru */
571 if (a[OVS_PACKET_ATTR_MRU]) {
572 mru = nla_get_u16(a[OVS_PACKET_ATTR_MRU]);
573 packet->ignore_df = 1;
574 }
575 OVS_CB(packet)->mru = mru;
576
e0e57990 577 /* Build an sw_flow for sending this packet. */
df65fec1 578 flow = ovs_flow_alloc();
e0e57990
BP		 579 err = PTR_ERR(flow);
580 if (IS_ERR(flow))
e5cad958 581 goto err_kfree_skb;
064af421 582
038e34ab
JS		 583 err = ovs_flow_key_extract_userspace(net, a[OVS_PACKET_ATTR_KEY],
584 packet, &flow->key, log);
e0e57990 585 if (err)
9321954a 586 goto err_flow_free;
e0e57990 587
a94ebc39 588 err = ovs_nla_copy_actions(net, a[OVS_PACKET_ATTR_ACTIONS],
9233cef7 589 &flow->key, &acts, log);
9b405f1a
PS		 590 if (err)
591 goto err_flow_free;
e0e57990 592
ff27161e 593 rcu_assign_pointer(flow->sf_acts, acts);
abff858b 594 packet->priority = flow->key.phy.priority;
3025a772 595 packet->mark = flow->key.phy.skb_mark;
e0e57990 596
d6569377 597 rcu_read_lock();
a94ebc39 598 dp = get_dp_rcu(net, ovs_header->dp_ifindex);
f7cd0081 599 err = -ENODEV;
e5cad958
BP		 600 if (!dp)
601 goto err_unlock;
cc4015df 602
a6059080
AZ		 603 input_vport = ovs_vport_rcu(dp, flow->key.phy.in_port);
604 if (!input_vport)
605 input_vport = ovs_vport_rcu(dp, OVSP_LOCAL);
606
607 if (!input_vport)
608 goto err_unlock;
609
e23775f2 610 packet->dev = input_vport->dev;
a6059080 611 OVS_CB(packet)->input_vport = input_vport;
ad50cb60 612 sf_acts = rcu_dereference(flow->sf_acts);
a6059080 613
e9141eec 614 local_bh_disable();
7d16c847 615 err = ovs_execute_actions(dp, packet, sf_acts, &flow->key);
e9141eec 616 local_bh_enable();
d6569377 617 rcu_read_unlock();
e0e57990 618
a1c564be 619 ovs_flow_free(flow, false);
e5cad958 620 return err;
064af421 621
e5cad958
BP		 622err_unlock:
623 rcu_read_unlock();
9321954a 624err_flow_free:
a1c564be 625 ovs_flow_free(flow, false);
e5cad958
BP		 626err_kfree_skb:
627 kfree_skb(packet);
628err:
d6569377 629 return err;
064af421
BP		 630}
631
df2c07f4 632static const struct nla_policy packet_policy[OVS_PACKET_ATTR_MAX + 1] = {
7c3072cc 633 [OVS_PACKET_ATTR_PACKET] = { .len = ETH_HLEN },
df2c07f4
JP		 634 [OVS_PACKET_ATTR_KEY] = { .type = NLA_NESTED },
635 [OVS_PACKET_ATTR_ACTIONS] = { .type = NLA_NESTED },
2e460098 636 [OVS_PACKET_ATTR_PROBE] = { .type = NLA_FLAG },
a94ebc39 637 [OVS_PACKET_ATTR_MRU] = { .type = NLA_U16 },
982b8810
BP		 638};
639
18fd3a52 640static struct genl_ops dp_packet_genl_ops[] = {
df2c07f4 641 { .cmd = OVS_PACKET_CMD_EXECUTE,
a6a8674d 642 .flags = GENL_UNS_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
982b8810 643 .policy = packet_policy,
df2c07f4 644 .doit = ovs_packet_cmd_execute
982b8810
BP		 645 }
646};
647
ba63fe26 648static struct genl_family dp_packet_genl_family __ro_after_init = {
cb25142c
PS		 649 .hdrsize = sizeof(struct ovs_header),
650 .name = OVS_PACKET_FAMILY,
651 .version = OVS_PACKET_VERSION,
652 .maxattr = OVS_PACKET_ATTR_MAX,
653 .netnsok = true,
654 .parallel_ops = true,
655 .ops = dp_packet_genl_ops,
656 .n_ops = ARRAY_SIZE(dp_packet_genl_ops),
ba63fe26 657 .module = THIS_MODULE,
cb25142c
PS		 658};
659
f1f60b85 660static void get_dp_stats(const struct datapath *dp, struct ovs_dp_stats *stats,
4fa72a95 661 struct ovs_dp_megaflow_stats *mega_stats)
064af421 662{
d6569377 663 int i;
f180c2e2 664
4fa72a95
AZ		 665 memset(mega_stats, 0, sizeof(*mega_stats));
666
994dc286 667 stats->n_flows = ovs_flow_tbl_count(&dp->table);
4fa72a95 668 mega_stats->n_masks = ovs_flow_tbl_num_masks(&dp->table);
064af421 669
7257b535 670 stats->n_hit = stats->n_missed = stats->n_lost = 0;
4fa72a95 671
d6569377
BP		 672 for_each_possible_cpu(i) {
673 const struct dp_stats_percpu *percpu_stats;
674 struct dp_stats_percpu local_stats;
821cb9fa 675 unsigned int start;
44e05eca 676
d6569377 677 percpu_stats = per_cpu_ptr(dp->stats_percpu, i);
064af421 678
d6569377 679 do {
b81deb15 680 start = u64_stats_fetch_begin_irq(&percpu_stats->syncp);
d6569377 681 local_stats = *percpu_stats;
b81deb15 682 } while (u64_stats_fetch_retry_irq(&percpu_stats->syncp, start));
064af421 683
d6569377
BP		 684 stats->n_hit += local_stats.n_hit;
685 stats->n_missed += local_stats.n_missed;
686 stats->n_lost += local_stats.n_lost;
4fa72a95 687 mega_stats->n_mask_hit += local_stats.n_mask_hit;
d6569377
BP		 688 }
689}
064af421 690
bc619e29
JS		 691static bool should_fill_key(const struct sw_flow_id *sfid, uint32_t ufid_flags)
692{
693 return ovs_identifier_is_ufid(sfid) &&
694 !(ufid_flags & OVS_UFID_F_OMIT_KEY);
695}
696
697static bool should_fill_mask(uint32_t ufid_flags)
698{
699 return !(ufid_flags & OVS_UFID_F_OMIT_MASK);
700}
701
702static bool should_fill_actions(uint32_t ufid_flags)
0afa2373 703{
bc619e29
JS		 704 return !(ufid_flags & OVS_UFID_F_OMIT_ACTIONS);
705}
706
707static size_t ovs_flow_cmd_msg_size(const struct sw_flow_actions *acts,
708 const struct sw_flow_id *sfid,
709 uint32_t ufid_flags)
710{
711 size_t len = NLMSG_ALIGN(sizeof(struct ovs_header));
712
713 /* OVS_FLOW_ATTR_UFID */
714 if (sfid && ovs_identifier_is_ufid(sfid))
715 len += nla_total_size(sfid->ufid_len);
716
717 /* OVS_FLOW_ATTR_KEY */
718 if (!sfid || should_fill_key(sfid, ufid_flags))
719 len += nla_total_size(ovs_key_attr_size());
720
721 /* OVS_FLOW_ATTR_MASK */
722 if (should_fill_mask(ufid_flags))
723 len += nla_total_size(ovs_key_attr_size());
724
725 /* OVS_FLOW_ATTR_ACTIONS */
726 if (should_fill_actions(ufid_flags))
c3bb15b3 727 len += nla_total_size(acts->orig_len);
bc619e29
JS		 728
729 return len
91b37647 730 + nla_total_size_64bit(sizeof(struct ovs_flow_stats)) /* OVS_FLOW_ATTR_STATS */
0afa2373 731 + nla_total_size(1) /* OVS_FLOW_ATTR_TCP_FLAGS */
91b37647 732 + nla_total_size_64bit(8); /* OVS_FLOW_ATTR_USED */
0afa2373
TG		 733}
734
f1948bb9
JS		 735/* Called with ovs_mutex or RCU read lock. */
736static int ovs_flow_cmd_fill_stats(const struct sw_flow *flow,
737 struct sk_buff *skb)
738{
739 struct ovs_flow_stats stats;
740 __be16 tcp_flags;
741 unsigned long used;
742
b0f3a2fe 743 ovs_flow_stats_get(flow, &stats, &used, &tcp_flags);
f71db6b1 744
b0f3a2fe 745 if (used &&
89be7da8
PS		 746 nla_put_u64_64bit(skb, OVS_FLOW_ATTR_USED, ovs_flow_used_time(used),
747 OVS_FLOW_ATTR_PAD))
f1948bb9 748 return -EMSGSIZE;
d6569377 749
b0f3a2fe 750 if (stats.n_packets &&
91b37647
PS		 751 nla_put_64bit(skb, OVS_FLOW_ATTR_STATS,
752 sizeof(struct ovs_flow_stats), &stats,
753 OVS_FLOW_ATTR_PAD))
f1948bb9 754 return -EMSGSIZE;
b0b906cc 755
b0f3a2fe
PS		 756 if ((u8)ntohs(tcp_flags) &&
757 nla_put_u8(skb, OVS_FLOW_ATTR_TCP_FLAGS, (u8)ntohs(tcp_flags)))
f1948bb9
JS		 758 return -EMSGSIZE;
759
760 return 0;
761}
762
763/* Called with ovs_mutex or RCU read lock. */
764static int ovs_flow_cmd_fill_actions(const struct sw_flow *flow,
765 struct sk_buff *skb, int skb_orig_len)
766{
767 struct nlattr *start;
768 int err;
d6569377 769
df2c07f4 770 /* If OVS_FLOW_ATTR_ACTIONS doesn't fit, skip dumping the actions if
30053024
BP		 771 * this is the first flow to be dumped into 'skb'. This is unusual for
772 * Netlink but individual action lists can be longer than
773 * NLMSG_GOODSIZE and thus entirely undumpable if we didn't do this.
774 * The userspace caller can always fetch the actions separately if it
775 * really wants them. (Most userspace callers in fact don't care.)
776 *
777 * This can only fail for dump operations because the skb is always
778 * properly sized for single flows.
779 */
9b405f1a 780 start = nla_nest_start(skb, OVS_FLOW_ATTR_ACTIONS);
f6f481ee 781 if (start) {
f44ccce1
PS		 782 const struct sw_flow_actions *sf_acts;
783
780ec6ae 784 sf_acts = rcu_dereference_ovsl(flow->sf_acts);
a097c0b2
PS		 785 err = ovs_nla_put_actions(sf_acts->actions,
786 sf_acts->actions_len, skb);
f71db6b1 787
0a25b039
BP		 788 if (!err)
789 nla_nest_end(skb, start);
790 else {
791 if (skb_orig_len)
f1948bb9 792 return err;
0a25b039
BP		 793
794 nla_nest_cancel(skb, start);
795 }
f1948bb9
JS		 796 } else if (skb_orig_len) {
797 return -EMSGSIZE;
798 }
799
800 return 0;
801}
802
803/* Called with ovs_mutex or RCU read lock. */
2c622e5a 804static int ovs_flow_cmd_fill_info(const struct sw_flow *flow, int dp_ifindex,
f1948bb9 805 struct sk_buff *skb, u32 portid,
bc619e29 806 u32 seq, u32 flags, u8 cmd, u32 ufid_flags)
f1948bb9
JS		 807{
808 const int skb_orig_len = skb->len;
809 struct ovs_header *ovs_header;
810 int err;
811
7d16c847
PS		 812 ovs_header = genlmsg_put(skb, portid, seq, &dp_flow_genl_family,
813 flags, cmd);
f1948bb9
JS		 814 if (!ovs_header)
815 return -EMSGSIZE;
7d16c847 816
f1948bb9
JS		 817 ovs_header->dp_ifindex = dp_ifindex;
818
bc619e29 819 err = ovs_nla_put_identifier(flow, skb);
db7f2238
JS		 820 if (err)
821 goto error;
822
bc619e29
JS		 823 if (should_fill_key(&flow->id, ufid_flags)) {
824 err = ovs_nla_put_masked_key(flow, skb);
825 if (err)
826 goto error;
827 }
828
829 if (should_fill_mask(ufid_flags)) {
830 err = ovs_nla_put_mask(flow, skb);
831 if (err)
832 goto error;
833 }
f1948bb9
JS		 834
835 err = ovs_flow_cmd_fill_stats(flow, skb);
836 if (err)
837 goto error;
838
bc619e29
JS		 839 if (should_fill_actions(ufid_flags)) {
840 err = ovs_flow_cmd_fill_actions(flow, skb, skb_orig_len);
841 if (err)
842 goto error;
843 }
37a1300c 844
23b48dc1
TG		 845 genlmsg_end(skb, ovs_header);
846 return 0;
d6569377 847
37a1300c 848error:
df2c07f4 849 genlmsg_cancel(skb, ovs_header);
d6569377 850 return err;
44e05eca
BP		 851}
852
f71db6b1
JR		 853/* May not be called with RCU read lock. */
854static struct sk_buff *ovs_flow_cmd_alloc_info(const struct sw_flow_actions *acts,
bc619e29 855 const struct sw_flow_id *sfid,
afad3556 856 struct genl_info *info,
bc619e29
JS		 857 bool always,
858 uint32_t ufid_flags)
44e05eca 859{
afad3556 860 struct sk_buff *skb;
bc619e29 861 size_t len;
d6569377 862
114fce23
SG		 863 if (!always && !ovs_must_notify(&dp_flow_genl_family, info,
864 GROUP_ID(&ovs_dp_flow_multicast_group)))
afad3556
JR		 865 return NULL;
866
bc619e29 867 len = ovs_flow_cmd_msg_size(acts, sfid, ufid_flags);
40c08cda 868 skb = genlmsg_new(len, GFP_KERNEL);
afad3556
JR		 869 if (!skb)
870 return ERR_PTR(-ENOMEM);
871
872 return skb;
37a1300c 873}
8d5ebd83 874
f71db6b1 875/* Called with ovs_mutex. */
7d16c847 876static struct sk_buff *ovs_flow_cmd_build_info(const struct sw_flow *flow,
f71db6b1
JR		 877 int dp_ifindex,
878 struct genl_info *info, u8 cmd,
bc619e29 879 bool always, u32 ufid_flags)
37a1300c
BP		 880{
881 struct sk_buff *skb;
882 int retval;
d6569377 883
bc619e29
JS		 884 skb = ovs_flow_cmd_alloc_info(ovsl_dereference(flow->sf_acts),
885 &flow->id, info, always, ufid_flags);
a6ddcc9a 886 if (IS_ERR_OR_NULL(skb))
afad3556 887 return skb;
d6569377 888
2c622e5a 889 retval = ovs_flow_cmd_fill_info(flow, dp_ifindex, skb,
f71db6b1 890 info->snd_portid, info->snd_seq, 0,
bc619e29 891 cmd, ufid_flags);
37a1300c 892 BUG_ON(retval < 0);
d6569377 893 return skb;
064af421
BP		 894}
895
0c9fd022 896static int ovs_flow_cmd_new(struct sk_buff *skb, struct genl_info *info)
064af421 897{
a94ebc39 898 struct net *net = sock_net(skb->sk);
37a1300c 899 struct nlattr **a = info->attrs;
df2c07f4 900 struct ovs_header *ovs_header = info->userhdr;
bc619e29 901 struct sw_flow *flow = NULL, *new_flow;
a1c564be 902 struct sw_flow_mask mask;
37a1300c 903 struct sk_buff *reply;
9c52546b 904 struct datapath *dp;
0c9fd022 905 struct sw_flow_actions *acts;
a1c564be 906 struct sw_flow_match match;
bc619e29 907 u32 ufid_flags = ovs_nla_get_ufid_flags(a[OVS_FLOW_ATTR_UFID_FLAGS]);
bc4a05c6 908 int error;
9233cef7 909 bool log = !a[OVS_FLOW_ATTR_PROBE];
064af421 910
6740b721 911 /* Must have key and actions. */
37a1300c 912 error = -EINVAL;
a473df5b 913 if (!a[OVS_FLOW_ATTR_KEY]) {
7d16c847 914 OVS_NLERR(log, "Flow key attr not present in new flow.");
37a1300c 915 goto error;
a473df5b
JG		 916 }
917 if (!a[OVS_FLOW_ATTR_ACTIONS]) {
7d16c847 918 OVS_NLERR(log, "Flow actions attr not present in new flow.");
6740b721 919 goto error;
a473df5b 920 }
a1c564be 921
6740b721 922 /* Most of the time we need to allocate a new flow, do it before
af465b67
PS		 923 * locking.
924 */
6740b721
JR		 925 new_flow = ovs_flow_alloc();
926 if (IS_ERR(new_flow)) {
927 error = PTR_ERR(new_flow);
928 goto error;
929 }
930
931 /* Extract key. */
9b94fa6c 932 ovs_match_init(&match, &new_flow->key, false, &mask);
038e34ab 933 error = ovs_nla_get_match(net, &match, a[OVS_FLOW_ATTR_KEY],
9233cef7 934 a[OVS_FLOW_ATTR_MASK], log);
37a1300c 935 if (error)
6740b721 936 goto err_kfree_flow;
064af421 937
bc619e29
JS		 938 /* Extract flow identifier. */
939 error = ovs_nla_get_identifier(&new_flow->id, a[OVS_FLOW_ATTR_UFID],
1d334d4f 940 &new_flow->key, log);
bc619e29
JS		 941 if (error)
942 goto err_kfree_flow;
9b405f1a 943
1d334d4f 944 /* unmasked key is needed to match when ufid is not used. */
945 if (ovs_identifier_is_key(&new_flow->id))
946 match.key = new_flow->id.unmasked_key;
947
948 ovs_flow_mask_key(&new_flow->key, &new_flow->key, true, &mask);
949
6740b721 950 /* Validate actions. */
a94ebc39
JS		 951 error = ovs_nla_copy_actions(net, a[OVS_FLOW_ATTR_ACTIONS],
952 &new_flow->key, &acts, log);
0c9fd022 953 if (error) {
7d16c847 954 OVS_NLERR(log, "Flow actions may not be safe on all matching packets.");
4f67b12a 955 goto err_kfree_flow;
6740b721
JR		 956 }
957
bc619e29
JS		 958 reply = ovs_flow_cmd_alloc_info(acts, &new_flow->id, info, false,
959 ufid_flags);
6740b721
JR		 960 if (IS_ERR(reply)) {
961 error = PTR_ERR(reply);
962 goto err_kfree_acts;
37a1300c
BP		 963 }
964
cd2a59e9 965 ovs_lock();
a94ebc39 966 dp = get_dp(net, ovs_header->dp_ifindex);
6740b721
JR		 967 if (unlikely(!dp)) {
968 error = -ENODEV;
cd2a59e9 969 goto err_unlock_ovs;
6740b721 970 }
bc619e29 971
a1c564be 972 /* Check if this is a duplicate flow */
bc619e29
JS		 973 if (ovs_identifier_is_ufid(&new_flow->id))
974 flow = ovs_flow_tbl_lookup_ufid(&dp->table, &new_flow->id);
975 if (!flow)
1d334d4f 976 flow = ovs_flow_tbl_lookup(&dp->table, &new_flow->key);
6740b721
JR		 977 if (likely(!flow)) {
978 rcu_assign_pointer(new_flow->sf_acts, acts);
d6569377 979
d6569377 980 /* Put flow in bucket. */
6740b721
JR		 981 error = ovs_flow_tbl_insert(&dp->table, new_flow, &mask);
982 if (unlikely(error)) {
0585f7a8 983 acts = NULL;
6740b721
JR		 984 goto err_unlock_ovs;
985 }
986
987 if (unlikely(reply)) {
2c622e5a 988 error = ovs_flow_cmd_fill_info(new_flow,
6740b721
JR		 989 ovs_header->dp_ifindex,
990 reply, info->snd_portid,
991 info->snd_seq, 0,
bc619e29
JS		 992 OVS_FLOW_CMD_NEW,
993 ufid_flags);
6740b721 994 BUG_ON(error < 0);
0585f7a8 995 }
6740b721 996 ovs_unlock();
d6569377 997 } else {
0c9fd022
JR		 998 struct sw_flow_actions *old_acts;
999
d6569377
BP		 1000 /* Bail out if we're not allowed to modify an existing flow.
1001 * We accept NLM_F_CREATE in place of the intended NLM_F_EXCL
1002 * because Generic Netlink treats the latter as a dump
1003 * request. We also accept NLM_F_EXCL in case that bug ever
1004 * gets fixed.
1005 */
6740b721
JR		 1006 if (unlikely(info->nlhdr->nlmsg_flags & (NLM_F_CREATE
1007 | NLM_F_EXCL))) {
1008 error = -EEXIST;
cd2a59e9 1009 goto err_unlock_ovs;
6740b721 1010 }
bc619e29
JS		 1011 /* The flow identifier has to be the same for flow updates.
1012 * Look for any overlapping flow.
1013 */
1014 if (unlikely(!ovs_flow_cmp(flow, &match))) {
1015 if (ovs_identifier_is_key(&flow->id))
1016 flow = ovs_flow_tbl_lookup_exact(&dp->table,
1017 &match);
1018 else /* UFID matches but key is different */
1019 flow = NULL;
3440e4bc
AW		 1020 if (!flow) {
1021 error = -ENOENT;
1022 goto err_unlock_ovs;
1023 }
6740b721 1024 }
0c9fd022
JR		 1025 /* Update actions. */
1026 old_acts = ovsl_dereference(flow->sf_acts);
1027 rcu_assign_pointer(flow->sf_acts, acts);
0c9fd022 1028
6740b721 1029 if (unlikely(reply)) {
2c622e5a 1030 error = ovs_flow_cmd_fill_info(flow,
6740b721
JR		 1031 ovs_header->dp_ifindex,
1032 reply, info->snd_portid,
1033 info->snd_seq, 0,
bc619e29
JS		 1034 OVS_FLOW_CMD_NEW,
1035 ufid_flags);
6740b721
JR		 1036 BUG_ON(error < 0);
1037 }
1038 ovs_unlock();
0c9fd022 1039
e23775f2 1040 ovs_nla_free_flow_actions_rcu(old_acts);
6740b721 1041 ovs_flow_free(new_flow, false);
0c9fd022 1042 }
6740b721
JR		 1043
1044 if (reply)
cb25142c 1045 ovs_notify(&dp_flow_genl_family, &ovs_dp_flow_multicast_group, reply, info);
0c9fd022
JR		 1046 return 0;
1047
0c9fd022
JR		 1048err_unlock_ovs:
1049 ovs_unlock();
6740b721
JR		 1050 kfree_skb(reply);
1051err_kfree_acts:
e23775f2 1052 ovs_nla_free_flow_actions(acts);
6740b721
JR		 1053err_kfree_flow:
1054 ovs_flow_free(new_flow, false);
0c9fd022
JR		 1055error:
1056 return error;
1057}
1058
cc561abf 1059/* Factor out action copy to avoid "Wframe-larger-than=1024" warning. */
a94ebc39
JS		 1060static struct sw_flow_actions *get_flow_actions(struct net *net,
1061 const struct nlattr *a,
cc561abf 1062 const struct sw_flow_key *key,
9233cef7
JR		 1063 const struct sw_flow_mask *mask,
1064 bool log)
cc561abf
PS		 1065{
1066 struct sw_flow_actions *acts;
1067 struct sw_flow_key masked_key;
1068 int error;
1069
ad4adec2 1070 ovs_flow_mask_key(&masked_key, key, true, mask);
a94ebc39 1071 error = ovs_nla_copy_actions(net, a, &masked_key, &acts, log);
cc561abf 1072 if (error) {
9233cef7 1073 OVS_NLERR(log,
7d16c847 1074 "Actions may not be safe on all matching packets");
cc561abf
PS		 1075 return ERR_PTR(error);
1076 }
1077
1078 return acts;
1079}
1080
850c2a4d
TZ		 1081/* Factor out match-init and action-copy to avoid
1082 * "Wframe-larger-than=1024" warning. Because mask is only
1083 * used to get actions, we new a function to save some
1084 * stack space.
1085 *
1086 * If there are not key and action attrs, we return 0
1087 * directly. In the case, the caller will also not use the
1088 * match as before. If there is action attr, we try to get
1089 * actions and save them to *acts. Before returning from
1090 * the function, we reset the match->mask pointer. Because
1091 * we should not to return match object with dangling reference
1092 * to mask.
1093 * */
1094static int ovs_nla_init_match_and_action(struct net *net,
1095 struct sw_flow_match *match,
1096 struct sw_flow_key *key,
1097 struct nlattr **a,
1098 struct sw_flow_actions **acts,
1099 bool log)
1100{
1101 struct sw_flow_mask mask;
1102 int error = 0;
1103
1104 if (a[OVS_FLOW_ATTR_KEY]) {
1105 ovs_match_init(match, key, true, &mask);
1106 error = ovs_nla_get_match(net, match, a[OVS_FLOW_ATTR_KEY],
1107 a[OVS_FLOW_ATTR_MASK], log);
1108 if (error)
1109 goto error;
1110 }
1111
1112 if (a[OVS_FLOW_ATTR_ACTIONS]) {
1113 if (!a[OVS_FLOW_ATTR_KEY]) {
1114 OVS_NLERR(log,
1115 "Flow key attribute not present in set flow.");
497895ca
CJ		 1116 error = -EINVAL;
1117 goto error;
850c2a4d
TZ		 1118 }
1119
1120 *acts = get_flow_actions(net, a[OVS_FLOW_ATTR_ACTIONS], key,
1121 &mask, log);
1122 if (IS_ERR(*acts)) {
1123 error = PTR_ERR(*acts);
1124 goto error;
1125 }
1126 }
1127
1128 /* On success, error is 0. */
1129error:
1130 match->mask = NULL;
1131 return error;
1132}
1133
0c9fd022
JR		 1134static int ovs_flow_cmd_set(struct sk_buff *skb, struct genl_info *info)
1135{
a94ebc39 1136 struct net *net = sock_net(skb->sk);
0c9fd022
JR		 1137 struct nlattr **a = info->attrs;
1138 struct ovs_header *ovs_header = info->userhdr;
1d2a1b5f 1139 struct sw_flow_key key;
0c9fd022 1140 struct sw_flow *flow;
0c9fd022
JR		 1141 struct sk_buff *reply = NULL;
1142 struct datapath *dp;
6740b721 1143 struct sw_flow_actions *old_acts = NULL, *acts = NULL;
0c9fd022 1144 struct sw_flow_match match;
bc619e29
JS		 1145 struct sw_flow_id sfid;
1146 u32 ufid_flags = ovs_nla_get_ufid_flags(a[OVS_FLOW_ATTR_UFID_FLAGS]);
b24baa1a 1147 int error = 0;
9233cef7 1148 bool log = !a[OVS_FLOW_ATTR_PROBE];
bc619e29 1149 bool ufid_present;
0c9fd022 1150
bc619e29 1151 ufid_present = ovs_nla_get_ufid(&sfid, a[OVS_FLOW_ATTR_UFID], log);
850c2a4d 1152 if (!a[OVS_FLOW_ATTR_KEY] && !ufid_present) {
b24baa1a
PS		 1153 OVS_NLERR(log,
1154 "Flow set message rejected, Key attribute missing.");
850c2a4d 1155 return -EINVAL;
b24baa1a 1156 }
850c2a4d
TZ		 1157
1158 error = ovs_nla_init_match_and_action(net, &match, &key, a,
1159 &acts, log);
0c9fd022
JR		 1160 if (error)
1161 goto error;
d6569377 1162
850c2a4d 1163 if (acts) {
ff27161e 1164 /* Can allocate before locking if have acts. */
bc619e29
JS		 1165 reply = ovs_flow_cmd_alloc_info(acts, &sfid, info, false,
1166 ufid_flags);
6740b721
JR		 1167 if (IS_ERR(reply)) {
1168 error = PTR_ERR(reply);
1169 goto err_kfree_acts;
90b8c2f7 1170 }
0c9fd022
JR		 1171 }
1172
1173 ovs_lock();
a94ebc39 1174 dp = get_dp(net, ovs_header->dp_ifindex);
6740b721
JR		 1175 if (unlikely(!dp)) {
1176 error = -ENODEV;
0c9fd022 1177 goto err_unlock_ovs;
6740b721 1178 }
0c9fd022 1179 /* Check that the flow exists. */
bc619e29
JS		 1180 if (ufid_present)
1181 flow = ovs_flow_tbl_lookup_ufid(&dp->table, &sfid);
1182 else
1183 flow = ovs_flow_tbl_lookup_exact(&dp->table, &match);
6740b721
JR		 1184 if (unlikely(!flow)) {
1185 error = -ENOENT;
0c9fd022 1186 goto err_unlock_ovs;
6740b721 1187 }
3440e4bc 1188
0c9fd022 1189 /* Update actions, if present. */
6740b721 1190 if (likely(acts)) {
0c9fd022
JR		 1191 old_acts = ovsl_dereference(flow->sf_acts);
1192 rcu_assign_pointer(flow->sf_acts, acts);
6740b721
JR		 1193
1194 if (unlikely(reply)) {
2c622e5a 1195 error = ovs_flow_cmd_fill_info(flow,
6740b721
JR		 1196 ovs_header->dp_ifindex,
1197 reply, info->snd_portid,
1198 info->snd_seq, 0,
bc619e29
JS		 1199 OVS_FLOW_CMD_NEW,
1200 ufid_flags);
6740b721
JR		 1201 BUG_ON(error < 0);
1202 }
1203 } else {
1204 /* Could not alloc without acts before locking. */
7d16c847 1205 reply = ovs_flow_cmd_build_info(flow, ovs_header->dp_ifindex,
bc619e29
JS		 1206 info, OVS_FLOW_CMD_NEW, false,
1207 ufid_flags);
1208
6740b721
JR		 1209 if (unlikely(IS_ERR(reply))) {
1210 error = PTR_ERR(reply);
1211 goto err_unlock_ovs;
1212 }
9c52546b 1213 }
0c9fd022 1214
0c9fd022
JR		 1215 /* Clear stats. */
1216 if (a[OVS_FLOW_ATTR_CLEAR])
1217 ovs_flow_stats_clear(flow);
cd2a59e9 1218 ovs_unlock();
37a1300c 1219
6740b721 1220 if (reply)
cb25142c 1221 ovs_notify(&dp_flow_genl_family, &ovs_dp_flow_multicast_group, reply, info);
6740b721 1222 if (old_acts)
e23775f2 1223 ovs_nla_free_flow_actions_rcu(old_acts);
7d16c847 1224
d6569377 1225 return 0;
704a1e09 1226
cd2a59e9
PS		 1227err_unlock_ovs:
1228 ovs_unlock();
6740b721
JR		 1229 kfree_skb(reply);
1230err_kfree_acts:
e23775f2 1231 ovs_nla_free_flow_actions(acts);
37a1300c 1232error:
9c52546b 1233 return error;
704a1e09
BP		 1234}
1235
df2c07f4 1236static int ovs_flow_cmd_get(struct sk_buff *skb, struct genl_info *info)
704a1e09 1237{
37a1300c 1238 struct nlattr **a = info->attrs;
df2c07f4 1239 struct ovs_header *ovs_header = info->userhdr;
038e34ab 1240 struct net *net = sock_net(skb->sk);
37a1300c 1241 struct sw_flow_key key;
37a1300c 1242 struct sk_buff *reply;
704a1e09 1243 struct sw_flow *flow;
9c52546b 1244 struct datapath *dp;
a1c564be 1245 struct sw_flow_match match;
bc619e29
JS		 1246 struct sw_flow_id ufid;
1247 u32 ufid_flags = ovs_nla_get_ufid_flags(a[OVS_FLOW_ATTR_UFID_FLAGS]);
1248 int err = 0;
9233cef7 1249 bool log = !a[OVS_FLOW_ATTR_PROBE];
bc619e29 1250 bool ufid_present;
704a1e09 1251
bc619e29
JS		 1252 ufid_present = ovs_nla_get_ufid(&ufid, a[OVS_FLOW_ATTR_UFID], log);
1253 if (a[OVS_FLOW_ATTR_KEY]) {
9b94fa6c 1254 ovs_match_init(&match, &key, true, NULL);
038e34ab 1255 err = ovs_nla_get_match(net, &match, a[OVS_FLOW_ATTR_KEY], NULL,
bc619e29
JS		 1256 log);
1257 } else if (!ufid_present) {
9233cef7
JR		 1258 OVS_NLERR(log,
1259 "Flow get message rejected, Key attribute missing.");
bc619e29 1260 err = -EINVAL;
1b936472 1261 }
37a1300c
BP		 1262 if (err)
1263 return err;
704a1e09 1264
cd2a59e9 1265 ovs_lock();
2a4999f3 1266 dp = get_dp(sock_net(skb->sk), ovs_header->dp_ifindex);
cd2a59e9
PS		 1267 if (!dp) {
1268 err = -ENODEV;
1269 goto unlock;
1270 }
704a1e09 1271
bc619e29
JS		 1272 if (ufid_present)
1273 flow = ovs_flow_tbl_lookup_ufid(&dp->table, &ufid);
1274 else
1275 flow = ovs_flow_tbl_lookup_exact(&dp->table, &match);
3440e4bc 1276 if (!flow) {
cd2a59e9
PS		 1277 err = -ENOENT;
1278 goto unlock;
1279 }
d6569377 1280
7d16c847 1281 reply = ovs_flow_cmd_build_info(flow, ovs_header->dp_ifindex, info,
bc619e29 1282 OVS_FLOW_CMD_NEW, true, ufid_flags);
cd2a59e9
PS		 1283 if (IS_ERR(reply)) {
1284 err = PTR_ERR(reply);
1285 goto unlock;
1286 }
36956a7d 1287
cd2a59e9 1288 ovs_unlock();
37a1300c 1289 return genlmsg_reply(reply, info);
cd2a59e9
PS		 1290unlock:
1291 ovs_unlock();
1292 return err;
d6569377 1293}
9c52546b 1294
df2c07f4 1295static int ovs_flow_cmd_del(struct sk_buff *skb, struct genl_info *info)
d6569377 1296{
37a1300c 1297 struct nlattr **a = info->attrs;
df2c07f4 1298 struct ovs_header *ovs_header = info->userhdr;
038e34ab 1299 struct net *net = sock_net(skb->sk);
37a1300c 1300 struct sw_flow_key key;
37a1300c 1301 struct sk_buff *reply;
bc619e29 1302 struct sw_flow *flow = NULL;
d6569377 1303 struct datapath *dp;
a1c564be 1304 struct sw_flow_match match;
bc619e29
JS		 1305 struct sw_flow_id ufid;
1306 u32 ufid_flags = ovs_nla_get_ufid_flags(a[OVS_FLOW_ATTR_UFID_FLAGS]);
d6569377 1307 int err;
9233cef7 1308 bool log = !a[OVS_FLOW_ATTR_PROBE];
bc619e29 1309 bool ufid_present;
36956a7d 1310
bc619e29
JS		 1311 ufid_present = ovs_nla_get_ufid(&ufid, a[OVS_FLOW_ATTR_UFID], log);
1312 if (a[OVS_FLOW_ATTR_KEY]) {
9b94fa6c 1313 ovs_match_init(&match, &key, true, NULL);
038e34ab
JS		 1314 err = ovs_nla_get_match(net, &match, a[OVS_FLOW_ATTR_KEY],
1315 NULL, log);
cde7f3ba
JR		 1316 if (unlikely(err))
1317 return err;
1318 }
1319
cd2a59e9 1320 ovs_lock();
2a4999f3 1321 dp = get_dp(sock_net(skb->sk), ovs_header->dp_ifindex);
cde7f3ba 1322 if (unlikely(!dp)) {
cd2a59e9
PS		 1323 err = -ENODEV;
1324 goto unlock;
1325 }
7d16c847 1326
bc619e29 1327 if (unlikely(!a[OVS_FLOW_ATTR_KEY] && !ufid_present)) {
994dc286 1328 err = ovs_flow_tbl_flush(&dp->table);
cd2a59e9
PS		 1329 goto unlock;
1330 }
7d16c847 1331
bc619e29
JS		 1332 if (ufid_present)
1333 flow = ovs_flow_tbl_lookup_ufid(&dp->table, &ufid);
1334 else
1335 flow = ovs_flow_tbl_lookup_exact(&dp->table, &match);
3440e4bc 1336 if (unlikely(!flow)) {
cd2a59e9
PS		 1337 err = -ENOENT;
1338 goto unlock;
1339 }
d6569377 1340
994dc286 1341 ovs_flow_tbl_remove(&dp->table, flow);
cde7f3ba 1342 ovs_unlock();
37a1300c 1343
46051cf8 1344 reply = ovs_flow_cmd_alloc_info(rcu_dereference_raw(flow->sf_acts),
bc619e29 1345 &flow->id, info, false, ufid_flags);
cde7f3ba
JR		 1346
1347 if (likely(reply)) {
1348 if (likely(!IS_ERR(reply))) {
7d16c847
PS		 1349 rcu_read_lock(); /*To keep RCU checker happy. */
1350 err = ovs_flow_cmd_fill_info(flow, ovs_header->dp_ifindex,
cde7f3ba
JR		 1351 reply, info->snd_portid,
1352 info->snd_seq, 0,
bc619e29
JS		 1353 OVS_FLOW_CMD_DEL,
1354 ufid_flags);
cde7f3ba
JR		 1355 rcu_read_unlock();
1356 BUG_ON(err < 0);
cb25142c 1357 ovs_notify(&dp_flow_genl_family, &ovs_dp_flow_multicast_group, reply, info);
cde7f3ba 1358 } else {
cb25142c
PS		 1359 genl_set_err(&dp_flow_genl_family, sock_net(skb->sk), 0,
1360 GROUP_ID(&ovs_dp_flow_multicast_group), PTR_ERR(reply));
1361
cde7f3ba 1362 }
afad3556 1363 }
37a1300c 1364
a1c564be 1365 ovs_flow_free(flow, true);
37a1300c 1366 return 0;
cd2a59e9
PS		 1367unlock:
1368 ovs_unlock();
1369 return err;
37a1300c
BP		 1370}
1371
df2c07f4 1372static int ovs_flow_cmd_dump(struct sk_buff *skb, struct netlink_callback *cb)
37a1300c 1373{
bc619e29 1374 struct nlattr *a[__OVS_FLOW_ATTR_MAX];
df2c07f4 1375 struct ovs_header *ovs_header = genlmsg_data(nlmsg_data(cb->nlh));
994dc286 1376 struct table_instance *ti;
37a1300c 1377 struct datapath *dp;
bc619e29
JS		 1378 u32 ufid_flags;
1379 int err;
1380
1381 err = genlmsg_parse(cb->nlh, &dp_flow_genl_family, a,
15702dc9 1382 OVS_FLOW_ATTR_MAX, flow_policy, NULL);
bc619e29
JS		 1383 if (err)
1384 return err;
1385 ufid_flags = ovs_nla_get_ufid_flags(a[OVS_FLOW_ATTR_UFID_FLAGS]);
37a1300c 1386
f44ccce1 1387 rcu_read_lock();
01ac0970 1388 dp = get_dp_rcu(sock_net(skb->sk), ovs_header->dp_ifindex);
cd2a59e9 1389 if (!dp) {
f44ccce1 1390 rcu_read_unlock();
37a1300c 1391 return -ENODEV;
cd2a59e9 1392 }
37a1300c 1393
994dc286 1394 ti = rcu_dereference(dp->table.ti);
37a1300c 1395 for (;;) {
37a1300c
BP		 1396 struct sw_flow *flow;
1397 u32 bucket, obj;
1398
1399 bucket = cb->args[0];
1400 obj = cb->args[1];
994dc286 1401 flow = ovs_flow_tbl_dump_next(ti, &bucket, &obj);
3544358a 1402 if (!flow)
37a1300c
BP		 1403 break;
1404
2c622e5a 1405 if (ovs_flow_cmd_fill_info(flow, ovs_header->dp_ifindex, skb,
28aea917 1406 NETLINK_CB(cb->skb).portid,
37a1300c 1407 cb->nlh->nlmsg_seq, NLM_F_MULTI,
bc619e29 1408 OVS_FLOW_CMD_NEW, ufid_flags) < 0)
37a1300c
BP		 1409 break;
1410
1411 cb->args[0] = bucket;
1412 cb->args[1] = obj;
1413 }
f44ccce1 1414 rcu_read_unlock();
37a1300c 1415 return skb->len;
704a1e09
BP		 1416}
1417
cb25142c
PS		 1418static const struct nla_policy flow_policy[OVS_FLOW_ATTR_MAX + 1] = {
1419 [OVS_FLOW_ATTR_KEY] = { .type = NLA_NESTED },
9233cef7 1420 [OVS_FLOW_ATTR_MASK] = { .type = NLA_NESTED },
cb25142c
PS		 1421 [OVS_FLOW_ATTR_ACTIONS] = { .type = NLA_NESTED },
1422 [OVS_FLOW_ATTR_CLEAR] = { .type = NLA_FLAG },
9233cef7 1423 [OVS_FLOW_ATTR_PROBE] = { .type = NLA_FLAG },
bc619e29
JS		 1424 [OVS_FLOW_ATTR_UFID] = { .type = NLA_UNSPEC, .len = 1 },
1425 [OVS_FLOW_ATTR_UFID_FLAGS] = { .type = NLA_U32 },
cb25142c
PS		 1426};
1427
18fd3a52 1428static struct genl_ops dp_flow_genl_ops[] = {
df2c07f4 1429 { .cmd = OVS_FLOW_CMD_NEW,
a6a8674d 1430 .flags = GENL_UNS_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
37a1300c 1431 .policy = flow_policy,
0c9fd022 1432 .doit = ovs_flow_cmd_new
37a1300c 1433 },
df2c07f4 1434 { .cmd = OVS_FLOW_CMD_DEL,
a6a8674d 1435 .flags = GENL_UNS_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
37a1300c 1436 .policy = flow_policy,
df2c07f4 1437 .doit = ovs_flow_cmd_del
37a1300c 1438 },
df2c07f4 1439 { .cmd = OVS_FLOW_CMD_GET,
37a1300c
BP		 1440 .flags = 0, /* OK for unprivileged users. */
1441 .policy = flow_policy,
df2c07f4
JP		 1442 .doit = ovs_flow_cmd_get,
1443 .dumpit = ovs_flow_cmd_dump
37a1300c 1444 },
df2c07f4 1445 { .cmd = OVS_FLOW_CMD_SET,
a6a8674d 1446 .flags = GENL_UNS_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
37a1300c 1447 .policy = flow_policy,
0c9fd022 1448 .doit = ovs_flow_cmd_set,
37a1300c
BP		 1449 },
1450};
1451
ba63fe26 1452static struct genl_family dp_flow_genl_family __ro_after_init = {
df2c07f4 1453 .hdrsize = sizeof(struct ovs_header),
cb25142c
PS		 1454 .name = OVS_FLOW_FAMILY,
1455 .version = OVS_FLOW_VERSION,
1456 .maxattr = OVS_FLOW_ATTR_MAX,
b3dcb73c 1457 .netnsok = true,
cb25142c
PS		 1458 .parallel_ops = true,
1459 .ops = dp_flow_genl_ops,
1460 .n_ops = ARRAY_SIZE(dp_flow_genl_ops),
1461 .mcgrps = &ovs_dp_flow_multicast_group,
1462 .n_mcgrps = 1,
ba63fe26 1463 .module = THIS_MODULE,
aaff4b55
BP		 1464};
1465
0afa2373
TG		 1466static size_t ovs_dp_cmd_msg_size(void)
1467{
1468 size_t msgsize = NLMSG_ALIGN(sizeof(struct ovs_header));
1469
1470 msgsize += nla_total_size(IFNAMSIZ);
91b37647
PS		 1471 msgsize += nla_total_size_64bit(sizeof(struct ovs_dp_stats));
1472 msgsize += nla_total_size_64bit(sizeof(struct ovs_dp_megaflow_stats));
300af20a 1473 msgsize += nla_total_size(sizeof(u32)); /* OVS_DP_ATTR_USER_FEATURES */
0afa2373
TG		 1474
1475 return msgsize;
1476}
1477
d637497c 1478/* Called with ovs_mutex. */
df2c07f4 1479static int ovs_dp_cmd_fill_info(struct datapath *dp, struct sk_buff *skb,
28aea917 1480 u32 portid, u32 seq, u32 flags, u8 cmd)
064af421 1481{
df2c07f4 1482 struct ovs_header *ovs_header;
e926dfe3 1483 struct ovs_dp_stats dp_stats;
4fa72a95 1484 struct ovs_dp_megaflow_stats dp_megaflow_stats;
064af421
BP		 1485 int err;
1486
28aea917 1487 ovs_header = genlmsg_put(skb, portid, seq, &dp_datapath_genl_family,
aaff4b55 1488 flags, cmd);
df2c07f4 1489 if (!ovs_header)
aaff4b55 1490 goto error;
064af421 1491
b063d9f0 1492 ovs_header->dp_ifindex = get_dpifindex(dp);
064af421 1493
850b6b3b 1494 err = nla_put_string(skb, OVS_DP_ATTR_NAME, ovs_dp_name(dp));
064af421 1495 if (err)
d6569377 1496 goto nla_put_failure;
064af421 1497
4fa72a95 1498 get_dp_stats(dp, &dp_stats, &dp_megaflow_stats);
91b37647
PS		 1499 if (nla_put_64bit(skb, OVS_DP_ATTR_STATS, sizeof(struct ovs_dp_stats),
1500 &dp_stats, OVS_DP_ATTR_PAD))
4fa72a95
AZ		 1501 goto nla_put_failure;
1502
91b37647
PS		 1503 if (nla_put_64bit(skb, OVS_DP_ATTR_MEGAFLOW_STATS,
1504 sizeof(struct ovs_dp_megaflow_stats),
1505 &dp_megaflow_stats, OVS_DP_ATTR_PAD))
c3cc8c03 1506 goto nla_put_failure;
d6569377 1507
c58cc9a4
TG		 1508 if (nla_put_u32(skb, OVS_DP_ATTR_USER_FEATURES, dp->user_features))
1509 goto nla_put_failure;
1510
23b48dc1
TG		 1511 genlmsg_end(skb, ovs_header);
1512 return 0;
d6569377
BP		 1513
1514nla_put_failure:
df2c07f4 1515 genlmsg_cancel(skb, ovs_header);
aaff4b55
BP		 1516error:
1517 return -EMSGSIZE;
d6569377
BP		 1518}
1519
40c08cda 1520static struct sk_buff *ovs_dp_cmd_alloc_info(void)
d6569377 1521{
40c08cda 1522 return genlmsg_new(ovs_dp_cmd_msg_size(), GFP_KERNEL);
aaff4b55 1523}
9dca7bd5 1524
aa917006 1525/* Called with rcu_read_lock or ovs_mutex. */
2a4999f3 1526static struct datapath *lookup_datapath(struct net *net,
f1f60b85 1527 const struct ovs_header *ovs_header,
6455100f 1528 struct nlattr *a[OVS_DP_ATTR_MAX + 1])
d6569377 1529{
254f2dc8
BP		 1530 struct datapath *dp;
1531
df2c07f4 1532 if (!a[OVS_DP_ATTR_NAME])
2a4999f3 1533 dp = get_dp(net, ovs_header->dp_ifindex);
254f2dc8 1534 else {
d6569377 1535 struct vport *vport;
d6569377 1536
2a4999f3 1537 vport = ovs_vport_locate(net, nla_data(a[OVS_DP_ATTR_NAME]));
df2c07f4 1538 dp = vport && vport->port_no == OVSP_LOCAL ? vport->dp : NULL;
d6569377 1539 }
254f2dc8 1540 return dp ? dp : ERR_PTR(-ENODEV);
d6569377
BP		 1541}
1542
94358dcf
TG		 1543static void ovs_dp_reset_user_features(struct sk_buff *skb, struct genl_info *info)
1544{
1545 struct datapath *dp;
1546
1547 dp = lookup_datapath(sock_net(skb->sk), info->userhdr, info->attrs);
09350a3d 1548 if (IS_ERR(dp))
94358dcf
TG		 1549 return;
1550
1551 WARN(dp->user_features, "Dropping previously announced user features\n");
1552 dp->user_features = 0;
1553}
1554
f1f60b85 1555static void ovs_dp_change(struct datapath *dp, struct nlattr *a[])
c58cc9a4
TG		 1556{
1557 if (a[OVS_DP_ATTR_USER_FEATURES])
1558 dp->user_features = nla_get_u32(a[OVS_DP_ATTR_USER_FEATURES]);
1559}
1560
df2c07f4 1561static int ovs_dp_cmd_new(struct sk_buff *skb, struct genl_info *info)
d6569377 1562{
aaff4b55 1563 struct nlattr **a = info->attrs;
d6569377 1564 struct vport_parms parms;
aaff4b55 1565 struct sk_buff *reply;
d6569377
BP		 1566 struct datapath *dp;
1567 struct vport *vport;
2a4999f3 1568 struct ovs_net *ovs_net;
95b1d73a 1569 int err, i;
d6569377 1570
d6569377 1571 err = -EINVAL;
ea36840f 1572 if (!a[OVS_DP_ATTR_NAME] || !a[OVS_DP_ATTR_UPCALL_PID])
aaff4b55
BP		 1573 goto err;
1574
40c08cda 1575 reply = ovs_dp_cmd_alloc_info();
d81eef1b
JR		 1576 if (!reply)
1577 return -ENOMEM;
d6569377 1578
d6569377
BP		 1579 err = -ENOMEM;
1580 dp = kzalloc(sizeof(*dp), GFP_KERNEL);
1581 if (dp == NULL)
d81eef1b 1582 goto err_free_reply;
2a4999f3 1583
c0cddcec 1584 ovs_dp_set_net(dp, sock_net(skb->sk));
0ceaa66c 1585
d6569377 1586 /* Allocate table. */
994dc286
PS		 1587 err = ovs_flow_tbl_init(&dp->table);
1588 if (err)
d6569377
BP		 1589 goto err_free_dp;
1590
08fb1bbd 1591 dp->stats_percpu = netdev_alloc_pcpu_stats(struct dp_stats_percpu);
99769a40
JG		 1592 if (!dp->stats_percpu) {
1593 err = -ENOMEM;
1594 goto err_destroy_table;
1595 }
1596
95b1d73a
PS		 1597 dp->ports = kmalloc(DP_VPORT_HASH_BUCKETS * sizeof(struct hlist_head),
1598 GFP_KERNEL);
1599 if (!dp->ports) {
1600 err = -ENOMEM;
1601 goto err_destroy_percpu;
1602 }
1603
1604 for (i = 0; i < DP_VPORT_HASH_BUCKETS; i++)
1605 INIT_HLIST_HEAD(&dp->ports[i]);
1606
1cb57039
AZ		 1607 err = ovs_meters_init(dp);
1608 if (err)
1609 goto err_destroy_ports_array;
1610
d6569377 1611 /* Set up our datapath device. */
df2c07f4
JP		 1612 parms.name = nla_data(a[OVS_DP_ATTR_NAME]);
1613 parms.type = OVS_VPORT_TYPE_INTERNAL;
d6569377
BP		 1614 parms.options = NULL;
1615 parms.dp = dp;
df2c07f4 1616 parms.port_no = OVSP_LOCAL;
beb1c69a 1617 parms.upcall_portids = a[OVS_DP_ATTR_UPCALL_PID];
b063d9f0 1618
c58cc9a4
TG		 1619 ovs_dp_change(dp, a);
1620
d81eef1b
JR		 1621 /* So far only local changes have been made, now need the lock. */
1622 ovs_lock();
1623
d6569377
BP		 1624 vport = new_vport(&parms);
1625 if (IS_ERR(vport)) {
1626 err = PTR_ERR(vport);
1627 if (err == -EBUSY)
1628 err = -EEXIST;
1629
94358dcf
TG		 1630 if (err == -EEXIST) {
1631 /* An outdated user space instance that does not understand
1632 * the concept of user_features has attempted to create a new
1633 * datapath and is likely to reuse it. Drop all user features.
1634 */
1635 if (info->genlhdr->version < OVS_DP_VER_FEATURES)
1636 ovs_dp_reset_user_features(skb, info);
1637 }
1638
1cb57039 1639 goto err_destroy_meters;
d6569377 1640 }
d6569377 1641
d81eef1b
JR		 1642 err = ovs_dp_cmd_fill_info(dp, reply, info->snd_portid,
1643 info->snd_seq, 0, OVS_DP_CMD_NEW);
1644 BUG_ON(err < 0);
aaff4b55 1645
2a4999f3 1646 ovs_net = net_generic(ovs_dp_get_net(dp), ovs_net_id);
fb93e9aa 1647 list_add_tail_rcu(&dp->list_node, &ovs_net->dps);
a0fb56c1 1648
cd2a59e9 1649 ovs_unlock();
d6569377 1650
cb25142c 1651 ovs_notify(&dp_datapath_genl_family, &ovs_dp_datapath_multicast_group, reply, info);
d6569377
BP		 1652 return 0;
1653
1cb57039 1654err_destroy_meters:
d81eef1b 1655 ovs_unlock();
1cb57039
AZ		 1656 ovs_meters_exit(dp);
1657err_destroy_ports_array:
95b1d73a 1658 kfree(dp->ports);
99769a40
JG		 1659err_destroy_percpu:
1660 free_percpu(dp->stats_percpu);
d6569377 1661err_destroy_table:
e379e4d1 1662 ovs_flow_tbl_destroy(&dp->table);
d6569377 1663err_free_dp:
d6569377 1664 kfree(dp);
d81eef1b
JR		 1665err_free_reply:
1666 kfree_skb(reply);
d6569377 1667err:
064af421
BP		 1668 return err;
1669}
1670
cd2a59e9 1671/* Called with ovs_mutex. */
2a4999f3 1672static void __dp_destroy(struct datapath *dp)
44e05eca 1673{
95b1d73a 1674 int i;
44e05eca 1675
95b1d73a
PS		 1676 for (i = 0; i < DP_VPORT_HASH_BUCKETS; i++) {
1677 struct vport *vport;
f8dfbcb7 1678 struct hlist_node *n;
95b1d73a 1679
f8dfbcb7 1680 hlist_for_each_entry_safe(vport, n, &dp->ports[i], dp_hash_node)
95b1d73a
PS		 1681 if (vport->port_no != OVSP_LOCAL)
1682 ovs_dp_detach_port(vport);
1683 }
ed099e92 1684
fb93e9aa 1685 list_del_rcu(&dp->list_node);
ed099e92 1686
cd2a59e9 1687 /* OVSP_LOCAL is datapath internal port. We need to make sure that
d103f479
AZ		 1688 * all ports in datapath are destroyed first before freeing datapath.
1689 */
cd2a59e9 1690 ovs_dp_detach_port(ovs_vport_ovsl(dp, OVSP_LOCAL));
99620d2c 1691
d103f479 1692 /* RCU destroy the flow table */
ed099e92 1693 call_rcu(&dp->rcu, destroy_dp_rcu);
2a4999f3
PS		 1694}
1695
1696static int ovs_dp_cmd_del(struct sk_buff *skb, struct genl_info *info)
1697{
1698 struct sk_buff *reply;
1699 struct datapath *dp;
1700 int err;
1701
40c08cda 1702 reply = ovs_dp_cmd_alloc_info();
d81eef1b
JR		 1703 if (!reply)
1704 return -ENOMEM;
1705
cd2a59e9 1706 ovs_lock();
2a4999f3
PS		 1707 dp = lookup_datapath(sock_net(skb->sk), info->userhdr, info->attrs);
1708 err = PTR_ERR(dp);
1709 if (IS_ERR(dp))
d81eef1b 1710 goto err_unlock_free;
2a4999f3 1711
d81eef1b
JR		 1712 err = ovs_dp_cmd_fill_info(dp, reply, info->snd_portid,
1713 info->snd_seq, 0, OVS_DP_CMD_DEL);
1714 BUG_ON(err < 0);
2a4999f3
PS		 1715
1716 __dp_destroy(dp);
d81eef1b 1717 ovs_unlock();
7d16c847 1718
cb25142c 1719 ovs_notify(&dp_datapath_genl_family, &ovs_dp_datapath_multicast_group, reply, info);
99620d2c 1720 return 0;
d81eef1b
JR		 1721
1722err_unlock_free:
cd2a59e9 1723 ovs_unlock();
d81eef1b 1724 kfree_skb(reply);
cd2a59e9 1725 return err;
44e05eca
BP		 1726}
1727
df2c07f4 1728static int ovs_dp_cmd_set(struct sk_buff *skb, struct genl_info *info)
064af421 1729{
aaff4b55 1730 struct sk_buff *reply;
d6569377 1731 struct datapath *dp;
d6569377 1732 int err;
064af421 1733
40c08cda 1734 reply = ovs_dp_cmd_alloc_info();
d81eef1b
JR		 1735 if (!reply)
1736 return -ENOMEM;
1737
cd2a59e9 1738 ovs_lock();
2a4999f3 1739 dp = lookup_datapath(sock_net(skb->sk), info->userhdr, info->attrs);
cd2a59e9 1740 err = PTR_ERR(dp);
d6569377 1741 if (IS_ERR(dp))
d81eef1b 1742 goto err_unlock_free;
38c6ecbc 1743
c58cc9a4
TG		 1744 ovs_dp_change(dp, info->attrs);
1745
d81eef1b
JR		 1746 err = ovs_dp_cmd_fill_info(dp, reply, info->snd_portid,
1747 info->snd_seq, 0, OVS_DP_CMD_NEW);
1748 BUG_ON(err < 0);
a0fb56c1 1749
cd2a59e9 1750 ovs_unlock();
7d16c847 1751
cb25142c 1752 ovs_notify(&dp_datapath_genl_family, &ovs_dp_datapath_multicast_group, reply, info);
aaff4b55 1753 return 0;
d81eef1b
JR		 1754
1755err_unlock_free:
cd2a59e9 1756 ovs_unlock();
d81eef1b 1757 kfree_skb(reply);
cd2a59e9 1758 return err;
064af421
BP		 1759}
1760
df2c07f4 1761static int ovs_dp_cmd_get(struct sk_buff *skb, struct genl_info *info)
1dcf111b 1762{
aaff4b55 1763 struct sk_buff *reply;
d6569377 1764 struct datapath *dp;
d6569377 1765 int err;
1dcf111b 1766
40c08cda 1767 reply = ovs_dp_cmd_alloc_info();
d81eef1b
JR		 1768 if (!reply)
1769 return -ENOMEM;
1770
d637497c 1771 ovs_lock();
2a4999f3 1772 dp = lookup_datapath(sock_net(skb->sk), info->userhdr, info->attrs);
cd2a59e9
PS		 1773 if (IS_ERR(dp)) {
1774 err = PTR_ERR(dp);
d81eef1b 1775 goto err_unlock_free;
cd2a59e9 1776 }
d81eef1b
JR		 1777 err = ovs_dp_cmd_fill_info(dp, reply, info->snd_portid,
1778 info->snd_seq, 0, OVS_DP_CMD_NEW);
1779 BUG_ON(err < 0);
d637497c 1780 ovs_unlock();
aaff4b55
BP		 1781
1782 return genlmsg_reply(reply, info);
cd2a59e9 1783
d81eef1b 1784err_unlock_free:
d637497c 1785 ovs_unlock();
d81eef1b 1786 kfree_skb(reply);
cd2a59e9 1787 return err;
1dcf111b
JP		 1788}
1789
df2c07f4 1790static int ovs_dp_cmd_dump(struct sk_buff *skb, struct netlink_callback *cb)
a7786963 1791{
2a4999f3 1792 struct ovs_net *ovs_net = net_generic(sock_net(skb->sk), ovs_net_id);
254f2dc8
BP		 1793 struct datapath *dp;
1794 int skip = cb->args[0];
1795 int i = 0;
a7786963 1796
d637497c
PS		 1797 ovs_lock();
1798 list_for_each_entry(dp, &ovs_net->dps, list_node) {
a2bab2f0 1799 if (i >= skip &&
28aea917 1800 ovs_dp_cmd_fill_info(dp, skb, NETLINK_CB(cb->skb).portid,
aaff4b55 1801 cb->nlh->nlmsg_seq, NLM_F_MULTI,
df2c07f4 1802 OVS_DP_CMD_NEW) < 0)
aaff4b55 1803 break;
254f2dc8 1804 i++;
a7786963 1805 }
d637497c 1806 ovs_unlock();
aaff4b55 1807
254f2dc8
BP		 1808 cb->args[0] = i;
1809
aaff4b55 1810 return skb->len;
c19e6535
BP		 1811}
1812
cb25142c
PS		 1813static const struct nla_policy datapath_policy[OVS_DP_ATTR_MAX + 1] = {
1814 [OVS_DP_ATTR_NAME] = { .type = NLA_NUL_STRING, .len = IFNAMSIZ - 1 },
1815 [OVS_DP_ATTR_UPCALL_PID] = { .type = NLA_U32 },
1816 [OVS_DP_ATTR_USER_FEATURES] = { .type = NLA_U32 },
1817};
1818
18fd3a52 1819static struct genl_ops dp_datapath_genl_ops[] = {
df2c07f4 1820 { .cmd = OVS_DP_CMD_NEW,
a6a8674d 1821 .flags = GENL_UNS_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
aaff4b55 1822 .policy = datapath_policy,
df2c07f4 1823 .doit = ovs_dp_cmd_new
aaff4b55 1824 },
df2c07f4 1825 { .cmd = OVS_DP_CMD_DEL,
a6a8674d 1826 .flags = GENL_UNS_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
aaff4b55 1827 .policy = datapath_policy,
df2c07f4 1828 .doit = ovs_dp_cmd_del
aaff4b55 1829 },
df2c07f4 1830 { .cmd = OVS_DP_CMD_GET,
aaff4b55
BP		 1831 .flags = 0, /* OK for unprivileged users. */
1832 .policy = datapath_policy,
df2c07f4
JP		 1833 .doit = ovs_dp_cmd_get,
1834 .dumpit = ovs_dp_cmd_dump
aaff4b55 1835 },
df2c07f4 1836 { .cmd = OVS_DP_CMD_SET,
a6a8674d 1837 .flags = GENL_UNS_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
aaff4b55 1838 .policy = datapath_policy,
df2c07f4 1839 .doit = ovs_dp_cmd_set,
aaff4b55
BP		 1840 },
1841};
1842
ba63fe26 1843static struct genl_family dp_datapath_genl_family __ro_after_init = {
df2c07f4 1844 .hdrsize = sizeof(struct ovs_header),
cb25142c
PS		 1845 .name = OVS_DATAPATH_FAMILY,
1846 .version = OVS_DATAPATH_VERSION,
1847 .maxattr = OVS_DP_ATTR_MAX,
b3dcb73c 1848 .netnsok = true,
cb25142c
PS		 1849 .parallel_ops = true,
1850 .ops = dp_datapath_genl_ops,
1851 .n_ops = ARRAY_SIZE(dp_datapath_genl_ops),
1852 .mcgrps = &ovs_dp_datapath_multicast_group,
1853 .n_mcgrps = 1,
ba63fe26 1854 .module = THIS_MODULE,
f0fef760
BP		 1855};
1856
cd2a59e9 1857/* Called with ovs_mutex or RCU read lock. */
df2c07f4 1858static int ovs_vport_cmd_fill_info(struct vport *vport, struct sk_buff *skb,
b147f2e9
JB		 1859 struct net *net, u32 portid, u32 seq,
1860 u32 flags, u8 cmd)
064af421 1861{
df2c07f4 1862 struct ovs_header *ovs_header;
e926dfe3 1863 struct ovs_vport_stats vport_stats;
c19e6535
BP		 1864 int err;
1865
28aea917 1866 ovs_header = genlmsg_put(skb, portid, seq, &dp_vport_genl_family,
f0fef760 1867 flags, cmd);
df2c07f4 1868 if (!ovs_header)
f0fef760 1869 return -EMSGSIZE;
c19e6535 1870
99769a40 1871 ovs_header->dp_ifindex = get_dpifindex(vport->dp);
c19e6535 1872
c3cc8c03
DM		 1873 if (nla_put_u32(skb, OVS_VPORT_ATTR_PORT_NO, vport->port_no) ||
1874 nla_put_u32(skb, OVS_VPORT_ATTR_TYPE, vport->ops->type) ||
e23775f2 1875 nla_put_string(skb, OVS_VPORT_ATTR_NAME,
b147f2e9
JB		 1876 ovs_vport_name(vport)) ||
1877 nla_put_u32(skb, OVS_VPORT_ATTR_IFINDEX, vport->dev->ifindex))
c3cc8c03 1878 goto nla_put_failure;
c19e6535 1879
b147f2e9
JB		 1880#ifdef HAVE_PEERNET2ID_ALLOC
1881 if (!net_eq(net, dev_net(vport->dev))) {
1882 int id = peernet2id_alloc(net, dev_net(vport->dev));
1883
1884 if (nla_put_s32(skb, OVS_VPORT_ATTR_NETNSID, id))
1885 goto nla_put_failure;
1886 }
1887
1888#endif
850b6b3b 1889 ovs_vport_get_stats(vport, &vport_stats);
91b37647
PS		 1890 if (nla_put_64bit(skb, OVS_VPORT_ATTR_STATS,
1891 sizeof(struct ovs_vport_stats), &vport_stats,
1892 OVS_VPORT_ATTR_PAD))
c3cc8c03 1893 goto nla_put_failure;
c19e6535 1894
beb1c69a
AW		 1895 if (ovs_vport_get_upcall_portids(vport, skb))
1896 goto nla_put_failure;
1897
850b6b3b 1898 err = ovs_vport_get_options(vport, skb);
f0fef760
BP		 1899 if (err == -EMSGSIZE)
1900 goto error;
c19e6535 1901
23b48dc1
TG		 1902 genlmsg_end(skb, ovs_header);
1903 return 0;
c19e6535
BP		 1904
1905nla_put_failure:
1906 err = -EMSGSIZE;
f0fef760 1907error:
df2c07f4 1908 genlmsg_cancel(skb, ovs_header);
f0fef760 1909 return err;
064af421
BP		 1910}
1911
d81eef1b
JR		 1912static struct sk_buff *ovs_vport_cmd_alloc_info(void)
1913{
1914 return nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
1915}
1916
1917/* Called with ovs_mutex, only via ovs_dp_notify_wq(). */
b147f2e9
JB		 1918struct sk_buff *ovs_vport_cmd_build_info(struct vport *vport, struct net *net,
1919 u32 portid, u32 seq, u8 cmd)
064af421 1920{
c19e6535 1921 struct sk_buff *skb;
f0fef760 1922 int retval;
c19e6535 1923
f0fef760 1924 skb = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_ATOMIC);
c19e6535
BP		 1925 if (!skb)
1926 return ERR_PTR(-ENOMEM);
1927
b147f2e9 1928 retval = ovs_vport_cmd_fill_info(vport, skb, net, portid, seq, 0, cmd);
c25ea534
JG		 1929 BUG_ON(retval < 0);
1930
c19e6535 1931 return skb;
f0fef760 1932}
c19e6535 1933
cd2a59e9 1934/* Called with ovs_mutex or RCU read lock. */
2a4999f3 1935static struct vport *lookup_vport(struct net *net,
f1f60b85 1936 const struct ovs_header *ovs_header,
df2c07f4 1937 struct nlattr *a[OVS_VPORT_ATTR_MAX + 1])
c19e6535
BP		 1938{
1939 struct datapath *dp;
1940 struct vport *vport;
1941
b147f2e9
JB		 1942 if (a[OVS_VPORT_ATTR_IFINDEX])
1943 return ERR_PTR(-EOPNOTSUPP);
df2c07f4 1944 if (a[OVS_VPORT_ATTR_NAME]) {
2a4999f3 1945 vport = ovs_vport_locate(net, nla_data(a[OVS_VPORT_ATTR_NAME]));
ed099e92 1946 if (!vport)
c19e6535 1947 return ERR_PTR(-ENODEV);
24ce832d
BP		 1948 if (ovs_header->dp_ifindex &&
1949 ovs_header->dp_ifindex != get_dpifindex(vport->dp))
1950 return ERR_PTR(-ENODEV);
c19e6535 1951 return vport;
df2c07f4
JP		 1952 } else if (a[OVS_VPORT_ATTR_PORT_NO]) {
1953 u32 port_no = nla_get_u32(a[OVS_VPORT_ATTR_PORT_NO]);
c19e6535
BP		 1954
1955 if (port_no >= DP_MAX_PORTS)
f0fef760 1956 return ERR_PTR(-EFBIG);
c19e6535 1957
2a4999f3 1958 dp = get_dp(net, ovs_header->dp_ifindex);
c19e6535
BP		 1959 if (!dp)
1960 return ERR_PTR(-ENODEV);
f2459fe7 1961
cd2a59e9 1962 vport = ovs_vport_ovsl_rcu(dp, port_no);
ed099e92 1963 if (!vport)
17535c57 1964 return ERR_PTR(-ENODEV);
c19e6535
BP		 1965 return vport;
1966 } else
1967 return ERR_PTR(-EINVAL);
b147f2e9 1968
064af421
BP		 1969}
1970
8ce37339
PS		 1971/* Called with ovs_mutex */
1972static void update_headroom(struct datapath *dp)
1973{
1974 unsigned dev_headroom, max_headroom = 0;
1975 struct net_device *dev;
1976 struct vport *vport;
1977 int i;
1978
1979 for (i = 0; i < DP_VPORT_HASH_BUCKETS; i++) {
1980 hlist_for_each_entry_rcu(vport, &dp->ports[i], dp_hash_node) {
1981 dev = vport->dev;
1982 dev_headroom = netdev_get_fwd_headroom(dev);
1983 if (dev_headroom > max_headroom)
1984 max_headroom = dev_headroom;
1985 }
1986 }
1987
1988 dp->max_headroom = max_headroom;
1989 for (i = 0; i < DP_VPORT_HASH_BUCKETS; i++)
1990 hlist_for_each_entry_rcu(vport, &dp->ports[i], dp_hash_node)
1991 netdev_set_rx_headroom(vport->dev, max_headroom);
1992}
1993
df2c07f4 1994static int ovs_vport_cmd_new(struct sk_buff *skb, struct genl_info *info)
c19e6535 1995{
f0fef760 1996 struct nlattr **a = info->attrs;
df2c07f4 1997 struct ovs_header *ovs_header = info->userhdr;
c19e6535 1998 struct vport_parms parms;
ed099e92 1999 struct sk_buff *reply;
c19e6535 2000 struct vport *vport;
c19e6535 2001 struct datapath *dp;
b0ec0f27 2002 u32 port_no;
c19e6535 2003 int err;
b0ec0f27 2004
ea36840f
BP		 2005 if (!a[OVS_VPORT_ATTR_NAME] || !a[OVS_VPORT_ATTR_TYPE] ||
2006 !a[OVS_VPORT_ATTR_UPCALL_PID])
d81eef1b 2007 return -EINVAL;
b147f2e9
JB		 2008 if (a[OVS_VPORT_ATTR_IFINDEX])
2009 return -EOPNOTSUPP;
d81eef1b
JR		 2010
2011 port_no = a[OVS_VPORT_ATTR_PORT_NO]
2012 ? nla_get_u32(a[OVS_VPORT_ATTR_PORT_NO]) : 0;
2013 if (port_no >= DP_MAX_PORTS)
2014 return -EFBIG;
2015
2016 reply = ovs_vport_cmd_alloc_info();
2017 if (!reply)
2018 return -ENOMEM;
f0fef760 2019
cd2a59e9 2020 ovs_lock();
5a38795f 2021restart:
2a4999f3 2022 dp = get_dp(sock_net(skb->sk), ovs_header->dp_ifindex);
c19e6535
BP		 2023 err = -ENODEV;
2024 if (!dp)
d81eef1b 2025 goto exit_unlock_free;
c19e6535 2026
d81eef1b 2027 if (port_no) {
cd2a59e9 2028 vport = ovs_vport_ovsl(dp, port_no);
c19e6535
BP		 2029 err = -EBUSY;
2030 if (vport)
d81eef1b 2031 goto exit_unlock_free;
c19e6535
BP		 2032 } else {
2033 for (port_no = 1; ; port_no++) {
2034 if (port_no >= DP_MAX_PORTS) {
2035 err = -EFBIG;
d81eef1b 2036 goto exit_unlock_free;
c19e6535 2037 }
cd2a59e9 2038 vport = ovs_vport_ovsl(dp, port_no);
c19e6535
BP		 2039 if (!vport)
2040 break;
51d4d598 2041 }
064af421 2042 }
b0ec0f27 2043
df2c07f4
JP		 2044 parms.name = nla_data(a[OVS_VPORT_ATTR_NAME]);
2045 parms.type = nla_get_u32(a[OVS_VPORT_ATTR_TYPE]);
2046 parms.options = a[OVS_VPORT_ATTR_OPTIONS];
c19e6535
BP		 2047 parms.dp = dp;
2048 parms.port_no = port_no;
beb1c69a 2049 parms.upcall_portids = a[OVS_VPORT_ATTR_UPCALL_PID];
c19e6535
BP		 2050
2051 vport = new_vport(&parms);
2052 err = PTR_ERR(vport);
5a38795f
TG		 2053 if (IS_ERR(vport)) {
2054 if (err == -EAGAIN)
2055 goto restart;
d81eef1b 2056 goto exit_unlock_free;
5a38795f 2057 }
c19e6535 2058
b147f2e9
JB		 2059 err = ovs_vport_cmd_fill_info(vport, reply, genl_info_net(info),
2060 info->snd_portid, info->snd_seq, 0,
2061 OVS_VPORT_CMD_NEW);
d81eef1b 2062 BUG_ON(err < 0);
8ce37339
PS		 2063
2064 if (netdev_get_fwd_headroom(vport->dev) > dp->max_headroom)
2065 update_headroom(dp);
2066 else
2067 netdev_set_rx_headroom(vport->dev, dp->max_headroom);
2068
d81eef1b 2069 ovs_unlock();
e297c6b7 2070
cb25142c 2071 ovs_notify(&dp_vport_genl_family, &ovs_dp_vport_multicast_group, reply, info);
d81eef1b 2072 return 0;
c19e6535 2073
d81eef1b 2074exit_unlock_free:
cd2a59e9 2075 ovs_unlock();
d81eef1b 2076 kfree_skb(reply);
c19e6535 2077 return err;
44e05eca
BP		 2078}
2079
df2c07f4 2080static int ovs_vport_cmd_set(struct sk_buff *skb, struct genl_info *info)
44e05eca 2081{
f0fef760
BP		 2082 struct nlattr **a = info->attrs;
2083 struct sk_buff *reply;
c19e6535 2084 struct vport *vport;
c19e6535 2085 int err;
44e05eca 2086
d81eef1b
JR		 2087 reply = ovs_vport_cmd_alloc_info();
2088 if (!reply)
2089 return -ENOMEM;
2090
cd2a59e9 2091 ovs_lock();
2a4999f3 2092 vport = lookup_vport(sock_net(skb->sk), info->userhdr, a);
c19e6535
BP		 2093 err = PTR_ERR(vport);
2094 if (IS_ERR(vport))
d81eef1b 2095 goto exit_unlock_free;
44e05eca 2096
6455100f 2097 if (a[OVS_VPORT_ATTR_TYPE] &&
17ec1d04 2098 nla_get_u32(a[OVS_VPORT_ATTR_TYPE]) != vport->ops->type) {
4879d4c7 2099 err = -EINVAL;
d81eef1b 2100 goto exit_unlock_free;
c25ea534
JG		 2101 }
2102
17ec1d04 2103 if (a[OVS_VPORT_ATTR_OPTIONS]) {
850b6b3b 2104 err = ovs_vport_set_options(vport, a[OVS_VPORT_ATTR_OPTIONS]);
17ec1d04 2105 if (err)
d81eef1b 2106 goto exit_unlock_free;
17ec1d04 2107 }
1fc7083d 2108
beb1c69a 2109 if (a[OVS_VPORT_ATTR_UPCALL_PID]) {
7d16c847
PS		 2110 struct nlattr *ids = a[OVS_VPORT_ATTR_UPCALL_PID];
2111
2112 err = ovs_vport_set_upcall_portids(vport, ids);
beb1c69a
AW		 2113 if (err)
2114 goto exit_unlock_free;
2115 }
c19e6535 2116
b147f2e9
JB		 2117 err = ovs_vport_cmd_fill_info(vport, reply, genl_info_net(info),
2118 info->snd_portid, info->snd_seq, 0,
2119 OVS_VPORT_CMD_NEW);
c25ea534 2120 BUG_ON(err < 0);
cd2a59e9 2121 ovs_unlock();
d81eef1b 2122
cb25142c 2123 ovs_notify(&dp_vport_genl_family, &ovs_dp_vport_multicast_group, reply, info);
c25ea534
JG		 2124 return 0;
2125
d81eef1b 2126exit_unlock_free:
cd2a59e9 2127 ovs_unlock();
d81eef1b 2128 kfree_skb(reply);
c19e6535 2129 return err;
064af421
BP		 2130}
2131
df2c07f4 2132static int ovs_vport_cmd_del(struct sk_buff *skb, struct genl_info *info)
7c40efc9 2133{
8ce37339 2134 bool must_update_headroom = false;
f0fef760
BP		 2135 struct nlattr **a = info->attrs;
2136 struct sk_buff *reply;
8ce37339 2137 struct datapath *dp;
c19e6535 2138 struct vport *vport;
c19e6535
BP		 2139 int err;
2140
d81eef1b
JR		 2141 reply = ovs_vport_cmd_alloc_info();
2142 if (!reply)
2143 return -ENOMEM;
2144
cd2a59e9 2145 ovs_lock();
2a4999f3 2146 vport = lookup_vport(sock_net(skb->sk), info->userhdr, a);
c19e6535 2147 err = PTR_ERR(vport);
f0fef760 2148 if (IS_ERR(vport))
d81eef1b 2149 goto exit_unlock_free;
c19e6535 2150
df2c07f4 2151 if (vport->port_no == OVSP_LOCAL) {
f0fef760 2152 err = -EINVAL;
d81eef1b 2153 goto exit_unlock_free;
f0fef760
BP		 2154 }
2155
b147f2e9
JB		 2156 err = ovs_vport_cmd_fill_info(vport, reply, genl_info_net(info),
2157 info->snd_portid, info->snd_seq, 0,
2158 OVS_VPORT_CMD_DEL);
d81eef1b 2159 BUG_ON(err < 0);
8ce37339
PS		 2160
2161 /* the vport deletion may trigger dp headroom update */
2162 dp = vport->dp;
2163 if (netdev_get_fwd_headroom(vport->dev) == dp->max_headroom)
2164 must_update_headroom = true;
2165 netdev_reset_rx_headroom(vport->dev);
850b6b3b 2166 ovs_dp_detach_port(vport);
8ce37339
PS		 2167
2168 if (must_update_headroom)
2169 update_headroom(dp);
2170
d81eef1b 2171 ovs_unlock();
f0fef760 2172
cb25142c 2173 ovs_notify(&dp_vport_genl_family, &ovs_dp_vport_multicast_group, reply, info);
d81eef1b 2174 return 0;
f0fef760 2175
d81eef1b 2176exit_unlock_free:
cd2a59e9 2177 ovs_unlock();
d81eef1b 2178 kfree_skb(reply);
c19e6535 2179 return err;
7c40efc9
BP		 2180}
2181
df2c07f4 2182static int ovs_vport_cmd_get(struct sk_buff *skb, struct genl_info *info)
7c40efc9 2183{
f0fef760 2184 struct nlattr **a = info->attrs;
df2c07f4 2185 struct ovs_header *ovs_header = info->userhdr;
ed099e92 2186 struct sk_buff *reply;
c19e6535 2187 struct vport *vport;
c19e6535
BP		 2188 int err;
2189
d81eef1b
JR		 2190 reply = ovs_vport_cmd_alloc_info();
2191 if (!reply)
2192 return -ENOMEM;
2193
ed099e92 2194 rcu_read_lock();
2a4999f3 2195 vport = lookup_vport(sock_net(skb->sk), ovs_header, a);
c19e6535
BP		 2196 err = PTR_ERR(vport);
2197 if (IS_ERR(vport))
d81eef1b 2198 goto exit_unlock_free;
b147f2e9
JB		 2199 err = ovs_vport_cmd_fill_info(vport, reply, genl_info_net(info),
2200 info->snd_portid, info->snd_seq, 0,
2201 OVS_VPORT_CMD_NEW);
d81eef1b 2202 BUG_ON(err < 0);
df2fa9b5
JG		 2203 rcu_read_unlock();
2204
2205 return genlmsg_reply(reply, info);
ed099e92 2206
d81eef1b 2207exit_unlock_free:
ed099e92 2208 rcu_read_unlock();
d81eef1b 2209 kfree_skb(reply);
c19e6535
BP		 2210 return err;
2211}
2212
df2c07f4 2213static int ovs_vport_cmd_dump(struct sk_buff *skb, struct netlink_callback *cb)
c19e6535 2214{
df2c07f4 2215 struct ovs_header *ovs_header = genlmsg_data(nlmsg_data(cb->nlh));
c19e6535 2216 struct datapath *dp;
95b1d73a
PS		 2217 int bucket = cb->args[0], skip = cb->args[1];
2218 int i, j = 0;
c19e6535 2219
03fc2881 2220 rcu_read_lock();
01ac0970 2221 dp = get_dp_rcu(sock_net(skb->sk), ovs_header->dp_ifindex);
03fc2881
JR		 2222 if (!dp) {
2223 rcu_read_unlock();
f0fef760 2224 return -ENODEV;
03fc2881 2225 }
95b1d73a 2226 for (i = bucket; i < DP_VPORT_HASH_BUCKETS; i++) {
ed099e92 2227 struct vport *vport;
95b1d73a
PS		 2228
2229 j = 0;
f8dfbcb7 2230 hlist_for_each_entry_rcu(vport, &dp->ports[i], dp_hash_node) {
95b1d73a
PS		 2231 if (j >= skip &&
2232 ovs_vport_cmd_fill_info(vport, skb,
b147f2e9 2233 sock_net(skb->sk),
28aea917 2234 NETLINK_CB(cb->skb).portid,
95b1d73a
PS		 2235 cb->nlh->nlmsg_seq,
2236 NLM_F_MULTI,
2237 OVS_VPORT_CMD_NEW) < 0)
2238 goto out;
2239
2240 j++;
2241 }
2242 skip = 0;
c19e6535 2243 }
95b1d73a 2244out:
ed099e92 2245 rcu_read_unlock();
c19e6535 2246
95b1d73a
PS		 2247 cb->args[0] = i;
2248 cb->args[1] = j;
f0fef760 2249
95b1d73a 2250 return skb->len;
7c40efc9
BP		 2251}
2252
cb25142c
PS		 2253static const struct nla_policy vport_policy[OVS_VPORT_ATTR_MAX + 1] = {
2254 [OVS_VPORT_ATTR_NAME] = { .type = NLA_NUL_STRING, .len = IFNAMSIZ - 1 },
2255 [OVS_VPORT_ATTR_STATS] = { .len = sizeof(struct ovs_vport_stats) },
2256 [OVS_VPORT_ATTR_PORT_NO] = { .type = NLA_U32 },
2257 [OVS_VPORT_ATTR_TYPE] = { .type = NLA_U32 },
2258 [OVS_VPORT_ATTR_UPCALL_PID] = { .type = NLA_U32 },
2259 [OVS_VPORT_ATTR_OPTIONS] = { .type = NLA_NESTED },
b147f2e9
JB		 2260 [OVS_VPORT_ATTR_IFINDEX] = { .type = NLA_U32 },
2261 [OVS_VPORT_ATTR_NETNSID] = { .type = NLA_S32 },
cb25142c
PS		 2262};
2263
18fd3a52 2264static struct genl_ops dp_vport_genl_ops[] = {
df2c07f4 2265 { .cmd = OVS_VPORT_CMD_NEW,
a6a8674d 2266 .flags = GENL_UNS_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
f0fef760 2267 .policy = vport_policy,
df2c07f4 2268 .doit = ovs_vport_cmd_new
f0fef760 2269 },
df2c07f4 2270 { .cmd = OVS_VPORT_CMD_DEL,
a6a8674d 2271 .flags = GENL_UNS_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
f0fef760 2272 .policy = vport_policy,
df2c07f4 2273 .doit = ovs_vport_cmd_del
f0fef760 2274 },
df2c07f4 2275 { .cmd = OVS_VPORT_CMD_GET,
f0fef760
BP		 2276 .flags = 0, /* OK for unprivileged users. */
2277 .policy = vport_policy,
df2c07f4
JP		 2278 .doit = ovs_vport_cmd_get,
2279 .dumpit = ovs_vport_cmd_dump
f0fef760 2280 },
df2c07f4 2281 { .cmd = OVS_VPORT_CMD_SET,
a6a8674d 2282 .flags = GENL_UNS_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
f0fef760 2283 .policy = vport_policy,
df2c07f4 2284 .doit = ovs_vport_cmd_set,
f0fef760
BP		 2285 },
2286};
2287
ba63fe26 2288struct genl_family dp_vport_genl_family __ro_after_init = {
cb25142c
PS		 2289 .hdrsize = sizeof(struct ovs_header),
2290 .name = OVS_VPORT_FAMILY,
2291 .version = OVS_VPORT_VERSION,
2292 .maxattr = OVS_VPORT_ATTR_MAX,
2293 .netnsok = true,
2294 .parallel_ops = true,
2295 .ops = dp_vport_genl_ops,
2296 .n_ops = ARRAY_SIZE(dp_vport_genl_ops),
2297 .mcgrps = &ovs_dp_vport_multicast_group,
2298 .n_mcgrps = 1,
ba63fe26 2299 .module = THIS_MODULE,
982b8810 2300};
ed099e92 2301
18fd3a52 2302static struct genl_family *dp_genl_families[] = {
cb25142c
PS		 2303 &dp_datapath_genl_family,
2304 &dp_vport_genl_family,
2305 &dp_flow_genl_family,
2306 &dp_packet_genl_family,
1cb57039 2307 &dp_meter_genl_family,
982b8810 2308};
ed099e92 2309
982b8810
BP		 2310static void dp_unregister_genl(int n_families)
2311{
2312 int i;
ed099e92 2313
b867ca75 2314 for (i = 0; i < n_families; i++)
cb25142c 2315 genl_unregister_family(dp_genl_families[i]);
ed099e92
BP		 2316}
2317
ba63fe26 2318static int __init dp_register_genl(void)
064af421 2319{
982b8810
BP		 2320 int err;
2321 int i;
064af421 2322
982b8810 2323 for (i = 0; i < ARRAY_SIZE(dp_genl_families); i++) {
064af421 2324
cb25142c 2325 err = genl_register_family(dp_genl_families[i]);
982b8810
BP		 2326 if (err)
2327 goto error;
982b8810 2328 }
9cc8b4e4 2329
982b8810 2330 return 0;
064af421
BP		 2331
2332error:
cb25142c 2333 dp_unregister_genl(i);
982b8810 2334 return err;
064af421
BP		 2335}
2336
2a4999f3
PS		 2337static int __net_init ovs_init_net(struct net *net)
2338{
2339 struct ovs_net *ovs_net = net_generic(net, ovs_net_id);
2340
2341 INIT_LIST_HEAD(&ovs_net->dps);
cd2a59e9 2342 INIT_WORK(&ovs_net->dp_notify_work, ovs_dp_notify_wq);
038e34ab 2343 ovs_ct_init(net);
7f4a5d68 2344 ovs_netns_frags_init(net);
2345 ovs_netns_frags6_init(net);
2a4999f3
PS		 2346 return 0;
2347}
2348
cabd5516
PS		 2349static void __net_exit list_vports_from_net(struct net *net, struct net *dnet,
2350 struct list_head *head)
2a4999f3
PS		 2351{
2352 struct ovs_net *ovs_net = net_generic(net, ovs_net_id);
cabd5516
PS		 2353 struct datapath *dp;
2354
2355 list_for_each_entry(dp, &ovs_net->dps, list_node) {
2356 int i;
2357
2358 for (i = 0; i < DP_VPORT_HASH_BUCKETS; i++) {
2359 struct vport *vport;
2360
2361 hlist_for_each_entry(vport, &dp->ports[i], dp_hash_node) {
cabd5516
PS		 2362
2363 if (vport->ops->type != OVS_VPORT_TYPE_INTERNAL)
2364 continue;
2365
e23775f2 2366 if (dev_net(vport->dev) == dnet)
cabd5516
PS		 2367 list_add(&vport->detach_list, head);
2368 }
2369 }
2370 }
2371}
2372
2373static void __net_exit ovs_exit_net(struct net *dnet)
2374{
2375 struct datapath *dp, *dp_next;
2376 struct ovs_net *ovs_net = net_generic(dnet, ovs_net_id);
2377 struct vport *vport, *vport_next;
2378 struct net *net;
2379 LIST_HEAD(head);
2a4999f3 2380
7f4a5d68 2381 ovs_netns_frags6_exit(dnet);
2382 ovs_netns_frags_exit(dnet);
038e34ab 2383 ovs_ct_exit(dnet);
cd2a59e9
PS		 2384 ovs_lock();
2385 list_for_each_entry_safe(dp, dp_next, &ovs_net->dps, list_node)
2386 __dp_destroy(dp);
cabd5516 2387
3d10a0c8
YHW		 2388#ifdef HAVE_NET_RWSEM
2389 down_read(&net_rwsem);
2390#else
cabd5516 2391 rtnl_lock();
3d10a0c8 2392#endif
cabd5516
PS		 2393 for_each_net(net)
2394 list_vports_from_net(net, dnet, &head);
3d10a0c8
YHW		 2395#ifdef HAVE_NET_RWSEM
2396 up_read(&net_rwsem);
2397#else
cabd5516 2398 rtnl_unlock();
3d10a0c8 2399#endif
cabd5516
PS		 2400
2401 /* Detach all vports from given namespace. */
2402 list_for_each_entry_safe(vport, vport_next, &head, detach_list) {
2403 list_del(&vport->detach_list);
2404 ovs_dp_detach_port(vport);
2405 }
2406
cd2a59e9
PS		 2407 ovs_unlock();
2408
2409 cancel_work_sync(&ovs_net->dp_notify_work);
2a4999f3
PS		 2410}
2411
2412static struct pernet_operations ovs_net_ops = {
2413 .init = ovs_init_net,
2414 .exit = ovs_exit_net,
2415 .id = &ovs_net_id,
2416 .size = sizeof(struct ovs_net),
2417};
2418
22d24ebf
BP		 2419static int __init dp_init(void)
2420{
2421 int err;
2422
f3d85db3 2423 BUILD_BUG_ON(sizeof(struct ovs_skb_cb) > FIELD_SIZEOF(struct sk_buff, cb));
22d24ebf 2424
26bfaeaa 2425 pr_info("Open vSwitch switching datapath %s\n", VERSION);
064af421 2426
907c26a8 2427 ovs_nsh_init();
595e069a
JS		 2428 err = action_fifos_init();
2429 if (err)
7f4a5d68 2430 goto error;
595e069a 2431
5282e284 2432 err = ovs_internal_dev_rtnl_link_register();
2c8c4fb7
AZ		 2433 if (err)
2434 goto error_action_fifos_exit;
2435
5282e284
TG		 2436 err = ovs_flow_init();
2437 if (err)
2438 goto error_unreg_rtnl_link;
2439
850b6b3b 2440 err = ovs_vport_init();
064af421
BP		 2441 if (err)
2442 goto error_flow_exit;
2443
2a4999f3 2444 err = register_pernet_device(&ovs_net_ops);
f2459fe7
JG		 2445 if (err)
2446 goto error_vport_exit;
2447
7f4a5d68 2448 err = compat_init();
2a4999f3
PS		 2449 if (err)
2450 goto error_netns_exit;
2451
7f4a5d68 2452 err = register_netdevice_notifier(&ovs_dp_device_notifier);
2453 if (err)
2454 goto error_compat_exit;
2455
5a38795f
TG		 2456 err = ovs_netdev_init();
2457 if (err)
2458 goto error_unreg_notifier;
2459
982b8810
BP		 2460 err = dp_register_genl();
2461 if (err < 0)
5a38795f 2462 goto error_unreg_netdev;
982b8810 2463
064af421
BP		 2464 return 0;
2465
5a38795f
TG		 2466error_unreg_netdev:
2467 ovs_netdev_exit();
064af421 2468error_unreg_notifier:
850b6b3b 2469 unregister_netdevice_notifier(&ovs_dp_device_notifier);
7f4a5d68 2470error_compat_exit:
2471 compat_exit();
2a4999f3
PS		 2472error_netns_exit:
2473 unregister_pernet_device(&ovs_net_ops);
f2459fe7 2474error_vport_exit:
850b6b3b 2475 ovs_vport_exit();
064af421 2476error_flow_exit:
850b6b3b 2477 ovs_flow_exit();
5282e284
TG		 2478error_unreg_rtnl_link:
2479 ovs_internal_dev_rtnl_link_unregister();
2c8c4fb7
AZ		 2480error_action_fifos_exit:
2481 action_fifos_exit();
064af421 2482error:
907c26a8 2483 ovs_nsh_cleanup();
064af421
BP		 2484 return err;
2485}
2486
2487static void dp_cleanup(void)
2488{
982b8810 2489 dp_unregister_genl(ARRAY_SIZE(dp_genl_families));
5a38795f 2490 ovs_netdev_exit();
850b6b3b 2491 unregister_netdevice_notifier(&ovs_dp_device_notifier);
7f4a5d68 2492 compat_exit();
2a4999f3
PS		 2493 unregister_pernet_device(&ovs_net_ops);
2494 rcu_barrier();
850b6b3b
JG		 2495 ovs_vport_exit();
2496 ovs_flow_exit();
5282e284 2497 ovs_internal_dev_rtnl_link_unregister();
2c8c4fb7 2498 action_fifos_exit();
907c26a8 2499 ovs_nsh_cleanup();
064af421
BP		 2500}
2501
2502module_init(dp_init);
2503module_exit(dp_cleanup);
2504
2505MODULE_DESCRIPTION("Open vSwitch switching datapath");
2506MODULE_LICENSE("GPL");
3d0666d2 2507MODULE_VERSION(VERSION);
75e2077e
TLSC		 2508MODULE_ALIAS_GENL_FAMILY(OVS_DATAPATH_FAMILY);
2509MODULE_ALIAS_GENL_FAMILY(OVS_VPORT_FAMILY);
2510MODULE_ALIAS_GENL_FAMILY(OVS_FLOW_FAMILY);
2511MODULE_ALIAS_GENL_FAMILY(OVS_PACKET_FAMILY);
1cb57039 2512MODULE_ALIAS_GENL_FAMILY(OVS_METER_FAMILY);
openvswitch mirror