projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
netfilter: nf_nat: register NAT helpers.
[linux-2.6-block.git] / net / openvswitch / conntrack.c
CommitLineData
7f8a436e
JS		 1/*
2 * Copyright (c) 2015 Nicira, Inc.
3 *
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
14#include <linux/module.h>
15#include <linux/openvswitch.h>
05752523
JR		 16#include <linux/tcp.h>
17#include <linux/udp.h>
18#include <linux/sctp.h>
11efd5cb 19#include <linux/static_key.h>
7f8a436e 20#include <net/ip.h>
11efd5cb 21#include <net/genetlink.h>
7f8a436e 22#include <net/netfilter/nf_conntrack_core.h>
11efd5cb 23#include <net/netfilter/nf_conntrack_count.h>
cae3a262 24#include <net/netfilter/nf_conntrack_helper.h>
c2ac6673 25#include <net/netfilter/nf_conntrack_labels.h>
05752523 26#include <net/netfilter/nf_conntrack_seqadj.h>
06bd2bdf 27#include <net/netfilter/nf_conntrack_timeout.h>
7f8a436e
JS		 28#include <net/netfilter/nf_conntrack_zones.h>
29#include <net/netfilter/ipv6/nf_defrag_ipv6.h>
70b095c8 30#include <net/ipv6_frag.h>
7f8a436e 31
4806e975 32#if IS_ENABLED(CONFIG_NF_NAT)
d2c5c103 33#include <net/netfilter/nf_nat.h>
05752523
JR		 34#endif
35
7f8a436e
JS		 36#include "datapath.h"
37#include "conntrack.h"
38#include "flow.h"
39#include "flow_netlink.h"
40
41struct ovs_ct_len_tbl {
05752523
JR		 42 int maxlen;
43 int minlen;
7f8a436e
JS		 44};
45
182e3042
JS		 46/* Metadata mark for masked write to conntrack mark */
47struct md_mark {
48 u32 value;
49 u32 mask;
50};
51
c2ac6673 52/* Metadata label for masked write to conntrack label. */
33db4125
JS		 53struct md_labels {
54 struct ovs_key_ct_labels value;
55 struct ovs_key_ct_labels mask;
c2ac6673
JS		 56};
57
05752523
JR		 58enum ovs_ct_nat {
59 OVS_CT_NAT = 1 << 0, /* NAT for committed connections only. */
60 OVS_CT_SRC_NAT = 1 << 1, /* Source NAT for NEW connections. */
61 OVS_CT_DST_NAT = 1 << 2, /* Destination NAT for NEW connections. */
62};
63
7f8a436e
JS		 64/* Conntrack action context for execution. */
65struct ovs_conntrack_info {
cae3a262 66 struct nf_conntrack_helper *helper;
7f8a436e
JS		 67 struct nf_conntrack_zone zone;
68 struct nf_conn *ct;
ab38a7b5 69 u8 commit : 1;
05752523 70 u8 nat : 3; /* enum ovs_ct_nat */
dd41d33f 71 u8 force : 1;
12064551 72 u8 have_eventmask : 1;
7f8a436e 73 u16 family;
12064551 74 u32 eventmask; /* Mask of 1 << IPCT_*. */
182e3042 75 struct md_mark mark;
33db4125 76 struct md_labels labels;
06bd2bdf 77 char timeout[CTNL_TIMEOUT_NAME_MAX];
4806e975 78#if IS_ENABLED(CONFIG_NF_NAT)
2eb0f624 79 struct nf_nat_range2 range; /* Only present for SRC NAT and DST NAT. */
05752523 80#endif
7f8a436e
JS		 81};
82
11efd5cb
YHW		 83#if IS_ENABLED(CONFIG_NETFILTER_CONNCOUNT)
84#define OVS_CT_LIMIT_UNLIMITED 0
85#define OVS_CT_LIMIT_DEFAULT OVS_CT_LIMIT_UNLIMITED
86#define CT_LIMIT_HASH_BUCKETS 512
87static DEFINE_STATIC_KEY_FALSE(ovs_ct_limit_enabled);
88
89struct ovs_ct_limit {
90 /* Elements in ovs_ct_limit_info->limits hash table */
91 struct hlist_node hlist_node;
92 struct rcu_head rcu;
93 u16 zone;
94 u32 limit;
95};
96
97struct ovs_ct_limit_info {
98 u32 default_limit;
99 struct hlist_head *limits;
100 struct nf_conncount_data *data;
101};
102
103static const struct nla_policy ct_limit_policy[OVS_CT_LIMIT_ATTR_MAX + 1] = {
104 [OVS_CT_LIMIT_ATTR_ZONE_LIMIT] = { .type = NLA_NESTED, },
105};
106#endif
107
09aa98ad
JR		 108static bool labels_nonzero(const struct ovs_key_ct_labels *labels);
109
2f3ab9f9
JS		 110static void __ovs_ct_free_action(struct ovs_conntrack_info *ct_info);
111
7f8a436e
JS		 112static u16 key_to_nfproto(const struct sw_flow_key *key)
113{
114 switch (ntohs(key->eth.type)) {
115 case ETH_P_IP:
116 return NFPROTO_IPV4;
117 case ETH_P_IPV6:
118 return NFPROTO_IPV6;
119 default:
120 return NFPROTO_UNSPEC;
121 }
122}
123
124/* Map SKB connection state into the values used by flow definition. */
125static u8 ovs_ct_get_state(enum ip_conntrack_info ctinfo)
126{
127 u8 ct_state = OVS_CS_F_TRACKED;
128
129 switch (ctinfo) {
130 case IP_CT_ESTABLISHED_REPLY:
131 case IP_CT_RELATED_REPLY:
7f8a436e
JS		 132 ct_state |= OVS_CS_F_REPLY_DIR;
133 break;
134 default:
135 break;
136 }
137
138 switch (ctinfo) {
139 case IP_CT_ESTABLISHED:
140 case IP_CT_ESTABLISHED_REPLY:
141 ct_state |= OVS_CS_F_ESTABLISHED;
142 break;
143 case IP_CT_RELATED:
144 case IP_CT_RELATED_REPLY:
145 ct_state |= OVS_CS_F_RELATED;
146 break;
147 case IP_CT_NEW:
7f8a436e
JS		 148 ct_state |= OVS_CS_F_NEW;
149 break;
150 default:
151 break;
152 }
153
154 return ct_state;
155}
156
0d5cdef8
JS		 157static u32 ovs_ct_get_mark(const struct nf_conn *ct)
158{
159#if IS_ENABLED(CONFIG_NF_CONNTRACK_MARK)
160 return ct ? ct->mark : 0;
161#else
162 return 0;
163#endif
164}
165
b87cec38
JR		 166/* Guard against conntrack labels max size shrinking below 128 bits. */
167#if NF_CT_LABELS_MAX_SIZE < 16
168#error NF_CT_LABELS_MAX_SIZE must be at least 16 bytes
169#endif
170
33db4125
JS		 171static void ovs_ct_get_labels(const struct nf_conn *ct,
172 struct ovs_key_ct_labels *labels)
c2ac6673
JS		 173{
174 struct nf_conn_labels *cl = ct ? nf_ct_labels_find(ct) : NULL;
175
b87cec38
JR		 176 if (cl)
177 memcpy(labels, cl->bits, OVS_CT_LABELS_LEN);
178 else
33db4125 179 memset(labels, 0, OVS_CT_LABELS_LEN);
c2ac6673
JS		 180}
181
9dd7f890
JR		 182static void __ovs_ct_update_key_orig_tp(struct sw_flow_key *key,
183 const struct nf_conntrack_tuple *orig,
184 u8 icmp_proto)
185{
316d4d78 186 key->ct_orig_proto = orig->dst.protonum;
9dd7f890
JR		 187 if (orig->dst.protonum == icmp_proto) {
188 key->ct.orig_tp.src = htons(orig->dst.u.icmp.type);
189 key->ct.orig_tp.dst = htons(orig->dst.u.icmp.code);
190 } else {
191 key->ct.orig_tp.src = orig->src.u.all;
192 key->ct.orig_tp.dst = orig->dst.u.all;
193 }
194}
195
7f8a436e 196static void __ovs_ct_update_key(struct sw_flow_key *key, u8 state,
182e3042
JS		 197 const struct nf_conntrack_zone *zone,
198 const struct nf_conn *ct)
7f8a436e 199{
316d4d78
JR		 200 key->ct_state = state;
201 key->ct_zone = zone->id;
0d5cdef8 202 key->ct.mark = ovs_ct_get_mark(ct);
33db4125 203 ovs_ct_get_labels(ct, &key->ct.labels);
9dd7f890
JR		 204
205 if (ct) {
206 const struct nf_conntrack_tuple *orig;
207
208 /* Use the master if we have one. */
209 if (ct->master)
210 ct = ct->master;
211 orig = &ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple;
212
213 /* IP version must match with the master connection. */
214 if (key->eth.type == htons(ETH_P_IP) &&
215 nf_ct_l3num(ct) == NFPROTO_IPV4) {
216 key->ipv4.ct_orig.src = orig->src.u3.ip;
217 key->ipv4.ct_orig.dst = orig->dst.u3.ip;
218 __ovs_ct_update_key_orig_tp(key, orig, IPPROTO_ICMP);
219 return;
220 } else if (key->eth.type == htons(ETH_P_IPV6) &&
221 !sw_flow_key_is_nd(key) &&
222 nf_ct_l3num(ct) == NFPROTO_IPV6) {
223 key->ipv6.ct_orig.src = orig->src.u3.in6;
224 key->ipv6.ct_orig.dst = orig->dst.u3.in6;
225 __ovs_ct_update_key_orig_tp(key, orig, NEXTHDR_ICMP);
226 return;
227 }
228 }
316d4d78 229 /* Clear 'ct_orig_proto' to mark the non-existence of conntrack
9dd7f890
JR		 230 * original direction key fields.
231 */
316d4d78 232 key->ct_orig_proto = 0;
7f8a436e
JS		 233}
234
5e17da63 235/* Update 'key' based on skb->_nfct. If 'post_ct' is true, then OVS has
05752523
JR		 236 * previously sent the packet to conntrack via the ct action. If
237 * 'keep_nat_flags' is true, the existing NAT flags retained, else they are
238 * initialized from the connection status.
7f8a436e
JS		 239 */
240static void ovs_ct_update_key(const struct sk_buff *skb,
d110986c 241 const struct ovs_conntrack_info *info,
05752523
JR		 242 struct sw_flow_key *key, bool post_ct,
243 bool keep_nat_flags)
7f8a436e
JS		 244{
245 const struct nf_conntrack_zone *zone = &nf_ct_zone_dflt;
246 enum ip_conntrack_info ctinfo;
247 struct nf_conn *ct;
248 u8 state = 0;
249
250 ct = nf_ct_get(skb, &ctinfo);
251 if (ct) {
252 state = ovs_ct_get_state(ctinfo);
9f13ded8 253 /* All unconfirmed entries are NEW connections. */
4f0909ee
JS		 254 if (!nf_ct_is_confirmed(ct))
255 state |= OVS_CS_F_NEW;
9f13ded8
JR		 256 /* OVS persists the related flag for the duration of the
257 * connection.
258 */
7f8a436e
JS		 259 if (ct->master)
260 state |= OVS_CS_F_RELATED;
05752523 261 if (keep_nat_flags) {
316d4d78 262 state |= key->ct_state & OVS_CS_F_NAT_MASK;
05752523
JR		 263 } else {
264 if (ct->status & IPS_SRC_NAT)
265 state |= OVS_CS_F_SRC_NAT;
266 if (ct->status & IPS_DST_NAT)
267 state |= OVS_CS_F_DST_NAT;
268 }
7f8a436e
JS		 269 zone = nf_ct_zone(ct);
270 } else if (post_ct) {
271 state = OVS_CS_F_TRACKED | OVS_CS_F_INVALID;
d110986c
JS		 272 if (info)
273 zone = &info->zone;
7f8a436e 274 }
182e3042 275 __ovs_ct_update_key(key, state, zone, ct);
7f8a436e
JS		 276}
277
9f13ded8
JR		 278/* This is called to initialize CT key fields possibly coming in from the local
279 * stack.
280 */
7f8a436e
JS		 281void ovs_ct_fill_key(const struct sk_buff *skb, struct sw_flow_key *key)
282{
05752523 283 ovs_ct_update_key(skb, NULL, key, false, false);
7f8a436e
JS		 284}
285
9dd7f890
JR		 286#define IN6_ADDR_INITIALIZER(ADDR) \
287 { (ADDR).s6_addr32[0], (ADDR).s6_addr32[1], \
288 (ADDR).s6_addr32[2], (ADDR).s6_addr32[3] }
289
290int ovs_ct_put_key(const struct sw_flow_key *swkey,
291 const struct sw_flow_key *output, struct sk_buff *skb)
7f8a436e 292{
316d4d78 293 if (nla_put_u32(skb, OVS_KEY_ATTR_CT_STATE, output->ct_state))
7f8a436e
JS		 294 return -EMSGSIZE;
295
296 if (IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES) &&
316d4d78 297 nla_put_u16(skb, OVS_KEY_ATTR_CT_ZONE, output->ct_zone))
7f8a436e
JS		 298 return -EMSGSIZE;
299
182e3042 300 if (IS_ENABLED(CONFIG_NF_CONNTRACK_MARK) &&
9dd7f890 301 nla_put_u32(skb, OVS_KEY_ATTR_CT_MARK, output->ct.mark))
182e3042
JS		 302 return -EMSGSIZE;
303
9723e6ab 304 if (IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS) &&
9dd7f890
JR		 305 nla_put(skb, OVS_KEY_ATTR_CT_LABELS, sizeof(output->ct.labels),
306 &output->ct.labels))
c2ac6673
JS		 307 return -EMSGSIZE;
308
316d4d78 309 if (swkey->ct_orig_proto) {
9dd7f890
JR		 310 if (swkey->eth.type == htons(ETH_P_IP)) {
311 struct ovs_key_ct_tuple_ipv4 orig = {
312 output->ipv4.ct_orig.src,
313 output->ipv4.ct_orig.dst,
314 output->ct.orig_tp.src,
315 output->ct.orig_tp.dst,
316d4d78 316 output->ct_orig_proto,
9dd7f890
JR		 317 };
318 if (nla_put(skb, OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV4,
319 sizeof(orig), &orig))
320 return -EMSGSIZE;
321 } else if (swkey->eth.type == htons(ETH_P_IPV6)) {
322 struct ovs_key_ct_tuple_ipv6 orig = {
323 IN6_ADDR_INITIALIZER(output->ipv6.ct_orig.src),
324 IN6_ADDR_INITIALIZER(output->ipv6.ct_orig.dst),
325 output->ct.orig_tp.src,
326 output->ct.orig_tp.dst,
316d4d78 327 output->ct_orig_proto,
9dd7f890
JR		 328 };
329 if (nla_put(skb, OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6,
330 sizeof(orig), &orig))
331 return -EMSGSIZE;
332 }
333 }
334
182e3042
JS		 335 return 0;
336}
337
6ffcea79 338static int ovs_ct_set_mark(struct nf_conn *ct, struct sw_flow_key *key,
182e3042
JS		 339 u32 ct_mark, u32 mask)
340{
0d5cdef8 341#if IS_ENABLED(CONFIG_NF_CONNTRACK_MARK)
182e3042
JS		 342 u32 new_mark;
343
182e3042
JS		 344 new_mark = ct_mark | (ct->mark & ~(mask));
345 if (ct->mark != new_mark) {
346 ct->mark = new_mark;
193e3096
JR		 347 if (nf_ct_is_confirmed(ct))
348 nf_conntrack_event_cache(IPCT_MARK, ct);
182e3042
JS		 349 key->ct.mark = new_mark;
350 }
351
7f8a436e 352 return 0;
0d5cdef8
JS		 353#else
354 return -ENOTSUPP;
355#endif
7f8a436e
JS		 356}
357
6ffcea79 358static struct nf_conn_labels *ovs_ct_get_conn_labels(struct nf_conn *ct)
c2ac6673 359{
c2ac6673 360 struct nf_conn_labels *cl;
c2ac6673
JS		 361
362 cl = nf_ct_labels_find(ct);
363 if (!cl) {
364 nf_ct_labels_ext_add(ct);
365 cl = nf_ct_labels_find(ct);
366 }
6ffcea79
JR		 367
368 return cl;
369}
370
371/* Initialize labels for a new, yet to be committed conntrack entry. Note that
372 * since the new connection is not yet confirmed, and thus no-one else has
2317c6b5 373 * access to it's labels, we simply write them over.
6ffcea79
JR		 374 */
375static int ovs_ct_init_labels(struct nf_conn *ct, struct sw_flow_key *key,
376 const struct ovs_key_ct_labels *labels,
377 const struct ovs_key_ct_labels *mask)
378{
09aa98ad
JR		 379 struct nf_conn_labels *cl, *master_cl;
380 bool have_mask = labels_nonzero(mask);
381
382 /* Inherit master's labels to the related connection? */
383 master_cl = ct->master ? nf_ct_labels_find(ct->master) : NULL;
384
385 if (!master_cl && !have_mask)
386 return 0; /* Nothing to do. */
6ffcea79
JR		 387
388 cl = ovs_ct_get_conn_labels(ct);
b87cec38 389 if (!cl)
c2ac6673
JS		 390 return -ENOSPC;
391
09aa98ad
JR		 392 /* Inherit the master's labels, if any. */
393 if (master_cl)
394 *cl = *master_cl;
395
396 if (have_mask) {
397 u32 *dst = (u32 *)cl->bits;
398 int i;
399
400 for (i = 0; i < OVS_CT_LABELS_LEN_32; i++)
401 dst[i] = (dst[i] & ~mask->ct_labels_32[i]) |
402 (labels->ct_labels_32[i]
403 & mask->ct_labels_32[i]);
404 }
193e3096 405
2317c6b5 406 /* Labels are included in the IPCTNL_MSG_CT_NEW event only if the
abd0a4f2 407 * IPCT_LABEL bit is set in the event cache.
2317c6b5
JR		 408 */
409 nf_conntrack_event_cache(IPCT_LABEL, ct);
410
6ffcea79
JR		 411 memcpy(&key->ct.labels, cl->bits, OVS_CT_LABELS_LEN);
412
413 return 0;
414}
415
416static int ovs_ct_set_labels(struct nf_conn *ct, struct sw_flow_key *key,
417 const struct ovs_key_ct_labels *labels,
418 const struct ovs_key_ct_labels *mask)
419{
420 struct nf_conn_labels *cl;
421 int err;
422
423 cl = ovs_ct_get_conn_labels(ct);
424 if (!cl)
425 return -ENOSPC;
426
427 err = nf_connlabels_replace(ct, labels->ct_labels_32,
428 mask->ct_labels_32,
429 OVS_CT_LABELS_LEN_32);
430 if (err)
431 return err;
432
433 memcpy(&key->ct.labels, cl->bits, OVS_CT_LABELS_LEN);
c2ac6673 434
c2ac6673
JS		 435 return 0;
436}
437
cae3a262
JS		 438/* 'skb' should already be pulled to nh_ofs. */
439static int ovs_ct_helper(struct sk_buff *skb, u16 proto)
440{
441 const struct nf_conntrack_helper *helper;
442 const struct nf_conn_help *help;
443 enum ip_conntrack_info ctinfo;
444 unsigned int protoff;
445 struct nf_conn *ct;
05752523 446 int err;
cae3a262
JS		 447
448 ct = nf_ct_get(skb, &ctinfo);
449 if (!ct || ctinfo == IP_CT_RELATED_REPLY)
450 return NF_ACCEPT;
451
452 help = nfct_help(ct);
453 if (!help)
454 return NF_ACCEPT;
455
456 helper = rcu_dereference(help->helper);
457 if (!helper)
458 return NF_ACCEPT;
459
460 switch (proto) {
461 case NFPROTO_IPV4:
462 protoff = ip_hdrlen(skb);
463 break;
464 case NFPROTO_IPV6: {
465 u8 nexthdr = ipv6_hdr(skb)->nexthdr;
466 __be16 frag_off;
cc570605 467 int ofs;
cae3a262 468
cc570605
JS		 469 ofs = ipv6_skip_exthdr(skb, sizeof(struct ipv6hdr), &nexthdr,
470 &frag_off);
471 if (ofs < 0 || (frag_off & htons(~0x7)) != 0) {
cae3a262
JS		 472 pr_debug("proto header not found\n");
473 return NF_ACCEPT;
474 }
cc570605 475 protoff = ofs;
cae3a262
JS		 476 break;
477 }
478 default:
479 WARN_ONCE(1, "helper invoked on non-IP family!");
480 return NF_DROP;
481 }
482
05752523
JR		 483 err = helper->help(skb, protoff, ct, ctinfo);
484 if (err != NF_ACCEPT)
485 return err;
486
487 /* Adjust seqs after helper. This is needed due to some helpers (e.g.,
488 * FTP with NAT) adusting the TCP payload size when mangling IP
489 * addresses and/or port numbers in the text-based control connection.
490 */
491 if (test_bit(IPS_SEQ_ADJUST_BIT, &ct->status) &&
492 !nf_ct_seq_adjust(skb, ct, ctinfo, protoff))
493 return NF_DROP;
494 return NF_ACCEPT;
cae3a262
JS		 495}
496
74c16618
JS		 497/* Returns 0 on success, -EINPROGRESS if 'skb' is stolen, or other nonzero
498 * value if 'skb' is freed.
499 */
7f8a436e
JS		 500static int handle_fragments(struct net *net, struct sw_flow_key *key,
501 u16 zone, struct sk_buff *skb)
502{
503 struct ovs_skb_cb ovs_cb = *OVS_CB(skb);
daaa7d64 504 int err;
7f8a436e
JS		 505
506 if (key->eth.type == htons(ETH_P_IP)) {
507 enum ip_defrag_users user = IP_DEFRAG_CONNTRACK_IN + zone;
7f8a436e
JS		 508
509 memset(IPCB(skb), 0, sizeof(struct inet_skb_parm));
19bcf9f2 510 err = ip_defrag(net, skb, user);
7f8a436e
JS		 511 if (err)
512 return err;
513
514 ovs_cb.mru = IPCB(skb)->frag_max_size;
7f8a436e 515#if IS_ENABLED(CONFIG_NF_DEFRAG_IPV6)
74c16618 516 } else if (key->eth.type == htons(ETH_P_IPV6)) {
7f8a436e 517 enum ip6_defrag_users user = IP6_DEFRAG_CONNTRACK_IN + zone;
7f8a436e
JS		 518
519 memset(IP6CB(skb), 0, sizeof(struct inet6_skb_parm));
daaa7d64 520 err = nf_ct_frag6_gather(net, skb, user);
f92a80a9
DDP		 521 if (err) {
522 if (err != -EINPROGRESS)
523 kfree_skb(skb);
daaa7d64 524 return err;
f92a80a9 525 }
7f8a436e 526
daaa7d64 527 key->ip.proto = ipv6_hdr(skb)->nexthdr;
7f8a436e 528 ovs_cb.mru = IP6CB(skb)->frag_max_size;
7f8a436e
JS		 529#endif
530 } else {
74c16618 531 kfree_skb(skb);
7f8a436e
JS		 532 return -EPFNOSUPPORT;
533 }
534
535 key->ip.frag = OVS_FRAG_TYPE_NONE;
536 skb_clear_hash(skb);
537 skb->ignore_df = 1;
538 *OVS_CB(skb) = ovs_cb;
539
540 return 0;
541}
542
543static struct nf_conntrack_expect *
544ovs_ct_expect_find(struct net *net, const struct nf_conntrack_zone *zone,
545 u16 proto, const struct sk_buff *skb)
546{
547 struct nf_conntrack_tuple tuple;
cf5d7091 548 struct nf_conntrack_expect *exp;
7f8a436e 549
a31f1adc 550 if (!nf_ct_get_tuplepr(skb, skb_network_offset(skb), proto, net, &tuple))
7f8a436e 551 return NULL;
cf5d7091
JR		 552
553 exp = __nf_ct_expect_find(net, zone, &tuple);
554 if (exp) {
555 struct nf_conntrack_tuple_hash *h;
556
557 /* Delete existing conntrack entry, if it clashes with the
558 * expectation. This can happen since conntrack ALGs do not
559 * check for clashes between (new) expectations and existing
560 * conntrack entries. nf_conntrack_in() will check the
561 * expectations only if a conntrack entry can not be found,
562 * which can lead to OVS finding the expectation (here) in the
563 * init direction, but which will not be removed by the
564 * nf_conntrack_in() call, if a matching conntrack entry is
565 * found instead. In this case all init direction packets
566 * would be reported as new related packets, while reply
567 * direction packets would be reported as un-related
568 * established packets.
569 */
570 h = nf_conntrack_find_get(net, zone, &tuple);
571 if (h) {
572 struct nf_conn *ct = nf_ct_tuplehash_to_ctrack(h);
573
574 nf_ct_delete(ct, 0, 0);
575 nf_conntrack_put(&ct->ct_general);
576 }
577 }
578
579 return exp;
7f8a436e
JS		 580}
581
289f2253
JR		 582/* This replicates logic from nf_conntrack_core.c that is not exported. */
583static enum ip_conntrack_info
584ovs_ct_get_info(const struct nf_conntrack_tuple_hash *h)
585{
586 const struct nf_conn *ct = nf_ct_tuplehash_to_ctrack(h);
587
588 if (NF_CT_DIRECTION(h) == IP_CT_DIR_REPLY)
589 return IP_CT_ESTABLISHED_REPLY;
590 /* Once we've had two way comms, always ESTABLISHED. */
591 if (test_bit(IPS_SEEN_REPLY_BIT, &ct->status))
592 return IP_CT_ESTABLISHED;
593 if (test_bit(IPS_EXPECTED_BIT, &ct->status))
594 return IP_CT_RELATED;
595 return IP_CT_NEW;
596}
597
598/* Find an existing connection which this packet belongs to without
599 * re-attributing statistics or modifying the connection state. This allows an
5e17da63 600 * skb->_nfct lost due to an upcall to be recovered during actions execution.
289f2253
JR		 601 *
602 * Must be called with rcu_read_lock.
603 *
5e17da63
JR		 604 * On success, populates skb->_nfct and returns the connection. Returns NULL
605 * if there is no existing entry.
289f2253
JR		 606 */
607static struct nf_conn *
608ovs_ct_find_existing(struct net *net, const struct nf_conntrack_zone *zone,
9ff464db 609 u8 l3num, struct sk_buff *skb, bool natted)
289f2253 610{
289f2253
JR		 611 struct nf_conntrack_tuple tuple;
612 struct nf_conntrack_tuple_hash *h;
289f2253 613 struct nf_conn *ct;
289f2253 614
60e3be94
FW		 615 if (!nf_ct_get_tuplepr(skb, skb_network_offset(skb), l3num,
616 net, &tuple)) {
289f2253
JR		 617 pr_debug("ovs_ct_find_existing: Can't get tuple\n");
618 return NULL;
619 }
620
9ff464db
JR		 621 /* Must invert the tuple if skb has been transformed by NAT. */
622 if (natted) {
623 struct nf_conntrack_tuple inverse;
624
303e0c55 625 if (!nf_ct_invert_tuple(&inverse, &tuple)) {
9ff464db
JR		 626 pr_debug("ovs_ct_find_existing: Inversion failed!\n");
627 return NULL;
628 }
629 tuple = inverse;
630 }
631
289f2253
JR		 632 /* look for tuple match */
633 h = nf_conntrack_find_get(net, zone, &tuple);
634 if (!h)
635 return NULL; /* Not found. */
636
637 ct = nf_ct_tuplehash_to_ctrack(h);
638
9ff464db
JR		 639 /* Inverted packet tuple matches the reverse direction conntrack tuple,
640 * select the other tuplehash to get the right 'ctinfo' bits for this
641 * packet.
642 */
643 if (natted)
644 h = &ct->tuplehash[!h->tuple.dst.dir];
645
c74454fa 646 nf_ct_set(skb, ct, ovs_ct_get_info(h));
289f2253
JR		 647 return ct;
648}
649
8b97ac5b
GR		 650static
651struct nf_conn *ovs_ct_executed(struct net *net,
652 const struct sw_flow_key *key,
653 const struct ovs_conntrack_info *info,
654 struct sk_buff *skb,
655 bool *ct_executed)
656{
657 struct nf_conn *ct = NULL;
658
659 /* If no ct, check if we have evidence that an existing conntrack entry
660 * might be found for this skb. This happens when we lose a skb->_nfct
661 * due to an upcall, or if the direction is being forced. If the
662 * connection was not confirmed, it is not cached and needs to be run
663 * through conntrack again.
664 */
665 *ct_executed = (key->ct_state & OVS_CS_F_TRACKED) &&
666 !(key->ct_state & OVS_CS_F_INVALID) &&
667 (key->ct_zone == info->zone.id);
668
669 if (*ct_executed || (!key->ct_state && info->force)) {
670 ct = ovs_ct_find_existing(net, &info->zone, info->family, skb,
671 !!(key->ct_state &
672 OVS_CS_F_NAT_MASK));
673 }
674
675 return ct;
676}
677
5e17da63 678/* Determine whether skb->_nfct is equal to the result of conntrack lookup. */
289f2253
JR		 679static bool skb_nfct_cached(struct net *net,
680 const struct sw_flow_key *key,
681 const struct ovs_conntrack_info *info,
682 struct sk_buff *skb)
7f8a436e
JS		 683{
684 enum ip_conntrack_info ctinfo;
685 struct nf_conn *ct;
8b97ac5b 686 bool ct_executed = true;
7f8a436e
JS		 687
688 ct = nf_ct_get(skb, &ctinfo);
689 if (!ct)
8b97ac5b
GR		 690 ct = ovs_ct_executed(net, key, info, skb, &ct_executed);
691
692 if (ct)
693 nf_ct_get(skb, &ctinfo);
694 else
7f8a436e 695 return false;
8b97ac5b 696
7f8a436e
JS		 697 if (!net_eq(net, read_pnet(&ct->ct_net)))
698 return false;
699 if (!nf_ct_zone_equal_any(info->ct, nf_ct_zone(ct)))
700 return false;
cae3a262
JS		 701 if (info->helper) {
702 struct nf_conn_help *help;
703
704 help = nf_ct_ext_find(ct, NF_CT_EXT_HELPER);
705 if (help && rcu_access_pointer(help->helper) != info->helper)
706 return false;
707 }
dd41d33f
JR		 708 /* Force conntrack entry direction to the current packet? */
709 if (info->force && CTINFO2DIR(ctinfo) != IP_CT_DIR_ORIGINAL) {
710 /* Delete the conntrack entry if confirmed, else just release
711 * the reference.
712 */
713 if (nf_ct_is_confirmed(ct))
714 nf_ct_delete(ct, 0, 0);
b768b16d
JR		 715
716 nf_conntrack_put(&ct->ct_general);
dd41d33f
JR		 717 nf_ct_set(skb, NULL, 0);
718 return false;
719 }
7f8a436e 720
8b97ac5b 721 return ct_executed;
7f8a436e
JS		 722}
723
4806e975 724#if IS_ENABLED(CONFIG_NF_NAT)
05752523
JR		 725/* Modelled after nf_nat_ipv[46]_fn().
726 * range is only used for new, uninitialized NAT state.
727 * Returns either NF_ACCEPT or NF_DROP.
728 */
729static int ovs_ct_nat_execute(struct sk_buff *skb, struct nf_conn *ct,
730 enum ip_conntrack_info ctinfo,
2eb0f624 731 const struct nf_nat_range2 *range,
05752523
JR		 732 enum nf_nat_manip_type maniptype)
733{
734 int hooknum, nh_off, err = NF_ACCEPT;
735
736 nh_off = skb_network_offset(skb);
75f01a4c 737 skb_pull_rcsum(skb, nh_off);
05752523
JR		 738
739 /* See HOOK2MANIP(). */
740 if (maniptype == NF_NAT_MANIP_SRC)
741 hooknum = NF_INET_LOCAL_IN; /* Source NAT */
742 else
743 hooknum = NF_INET_LOCAL_OUT; /* Destination NAT */
744
745 switch (ctinfo) {
746 case IP_CT_RELATED:
747 case IP_CT_RELATED_REPLY:
3bf195ae 748 if (IS_ENABLED(CONFIG_NF_NAT) &&
99b7248e 749 skb->protocol == htons(ETH_P_IP) &&
05752523
JR		 750 ip_hdr(skb)->protocol == IPPROTO_ICMP) {
751 if (!nf_nat_icmp_reply_translation(skb, ct, ctinfo,
752 hooknum))
753 err = NF_DROP;
754 goto push;
3bf195ae 755 } else if (IS_ENABLED(CONFIG_IPV6) &&
99b7248e 756 skb->protocol == htons(ETH_P_IPV6)) {
05752523
JR		 757 __be16 frag_off;
758 u8 nexthdr = ipv6_hdr(skb)->nexthdr;
759 int hdrlen = ipv6_skip_exthdr(skb,
760 sizeof(struct ipv6hdr),
761 &nexthdr, &frag_off);
762
763 if (hdrlen >= 0 && nexthdr == IPPROTO_ICMPV6) {
764 if (!nf_nat_icmpv6_reply_translation(skb, ct,
765 ctinfo,
766 hooknum,
767 hdrlen))
768 err = NF_DROP;
769 goto push;
770 }
05752523
JR		 771 }
772 /* Non-ICMP, fall thru to initialize if needed. */
279badc2 773 /* fall through */
05752523
JR		 774 case IP_CT_NEW:
775 /* Seen it before? This can happen for loopback, retrans,
776 * or local packets.
777 */
778 if (!nf_nat_initialized(ct, maniptype)) {
779 /* Initialize according to the NAT action. */
780 err = (range && range->flags & NF_NAT_RANGE_MAP_IPS)
781 /* Action is set up to establish a new
782 * mapping.
783 */
784 ? nf_nat_setup_info(ct, range, maniptype)
785 : nf_nat_alloc_null_binding(ct, hooknum);
786 if (err != NF_ACCEPT)
787 goto push;
788 }
789 break;
790
791 case IP_CT_ESTABLISHED:
792 case IP_CT_ESTABLISHED_REPLY:
793 break;
794
795 default:
796 err = NF_DROP;
797 goto push;
798 }
799
800 err = nf_nat_packet(ct, ctinfo, hooknum, skb);
801push:
802 skb_push(skb, nh_off);
75f01a4c 803 skb_postpush_rcsum(skb, skb->data, nh_off);
05752523
JR		 804
805 return err;
806}
807
808static void ovs_nat_update_key(struct sw_flow_key *key,
809 const struct sk_buff *skb,
810 enum nf_nat_manip_type maniptype)
811{
812 if (maniptype == NF_NAT_MANIP_SRC) {
813 __be16 src;
814
316d4d78 815 key->ct_state |= OVS_CS_F_SRC_NAT;
05752523
JR		 816 if (key->eth.type == htons(ETH_P_IP))
817 key->ipv4.addr.src = ip_hdr(skb)->saddr;
818 else if (key->eth.type == htons(ETH_P_IPV6))
819 memcpy(&key->ipv6.addr.src, &ipv6_hdr(skb)->saddr,
820 sizeof(key->ipv6.addr.src));
821 else
822 return;
823
824 if (key->ip.proto == IPPROTO_UDP)
825 src = udp_hdr(skb)->source;
826 else if (key->ip.proto == IPPROTO_TCP)
827 src = tcp_hdr(skb)->source;
828 else if (key->ip.proto == IPPROTO_SCTP)
829 src = sctp_hdr(skb)->source;
830 else
831 return;
832
833 key->tp.src = src;
834 } else {
835 __be16 dst;
836
316d4d78 837 key->ct_state |= OVS_CS_F_DST_NAT;
05752523
JR		 838 if (key->eth.type == htons(ETH_P_IP))
839 key->ipv4.addr.dst = ip_hdr(skb)->daddr;
840 else if (key->eth.type == htons(ETH_P_IPV6))
841 memcpy(&key->ipv6.addr.dst, &ipv6_hdr(skb)->daddr,
842 sizeof(key->ipv6.addr.dst));
843 else
844 return;
845
846 if (key->ip.proto == IPPROTO_UDP)
847 dst = udp_hdr(skb)->dest;
848 else if (key->ip.proto == IPPROTO_TCP)
849 dst = tcp_hdr(skb)->dest;
850 else if (key->ip.proto == IPPROTO_SCTP)
851 dst = sctp_hdr(skb)->dest;
852 else
853 return;
854
855 key->tp.dst = dst;
856 }
857}
858
859/* Returns NF_DROP if the packet should be dropped, NF_ACCEPT otherwise. */
860static int ovs_ct_nat(struct net *net, struct sw_flow_key *key,
861 const struct ovs_conntrack_info *info,
862 struct sk_buff *skb, struct nf_conn *ct,
863 enum ip_conntrack_info ctinfo)
864{
865 enum nf_nat_manip_type maniptype;
866 int err;
867
05752523
JR		 868 /* Add NAT extension if not confirmed yet. */
869 if (!nf_ct_is_confirmed(ct) && !nf_ct_nat_ext_add(ct))
870 return NF_ACCEPT; /* Can't NAT. */
871
872 /* Determine NAT type.
873 * Check if the NAT type can be deduced from the tracked connection.
5745b0be
JR		 874 * Make sure new expected connections (IP_CT_RELATED) are NATted only
875 * when committing.
05752523
JR		 876 */
877 if (info->nat & OVS_CT_NAT && ctinfo != IP_CT_NEW &&
878 ct->status & IPS_NAT_MASK &&
5745b0be 879 (ctinfo != IP_CT_RELATED || info->commit)) {
05752523
JR		 880 /* NAT an established or related connection like before. */
881 if (CTINFO2DIR(ctinfo) == IP_CT_DIR_REPLY)
882 /* This is the REPLY direction for a connection
883 * for which NAT was applied in the forward
884 * direction. Do the reverse NAT.
885 */
886 maniptype = ct->status & IPS_SRC_NAT
887 ? NF_NAT_MANIP_DST : NF_NAT_MANIP_SRC;
888 else
889 maniptype = ct->status & IPS_SRC_NAT
890 ? NF_NAT_MANIP_SRC : NF_NAT_MANIP_DST;
891 } else if (info->nat & OVS_CT_SRC_NAT) {
892 maniptype = NF_NAT_MANIP_SRC;
893 } else if (info->nat & OVS_CT_DST_NAT) {
894 maniptype = NF_NAT_MANIP_DST;
895 } else {
896 return NF_ACCEPT; /* Connection is not NATed. */
897 }
898 err = ovs_ct_nat_execute(skb, ct, ctinfo, &info->range, maniptype);
899
900 /* Mark NAT done if successful and update the flow key. */
901 if (err == NF_ACCEPT)
902 ovs_nat_update_key(key, skb, maniptype);
903
904 return err;
905}
4806e975 906#else /* !CONFIG_NF_NAT */
05752523
JR		 907static int ovs_ct_nat(struct net *net, struct sw_flow_key *key,
908 const struct ovs_conntrack_info *info,
909 struct sk_buff *skb, struct nf_conn *ct,
910 enum ip_conntrack_info ctinfo)
911{
912 return NF_ACCEPT;
913}
914#endif
915
9f13ded8 916/* Pass 'skb' through conntrack in 'net', using zone configured in 'info', if
394e910e
JR		 917 * not done already. Update key with new CT state after passing the packet
918 * through conntrack.
5e17da63 919 * Note that if the packet is deemed invalid by conntrack, skb->_nfct will be
9f13ded8
JR		 920 * set to NULL and 0 will be returned.
921 */
4f0909ee 922static int __ovs_ct_lookup(struct net *net, struct sw_flow_key *key,
7f8a436e
JS		 923 const struct ovs_conntrack_info *info,
924 struct sk_buff *skb)
925{
926 /* If we are recirculating packets to match on conntrack fields and
927 * committing with a separate conntrack action, then we don't need to
928 * actually run the packet through conntrack twice unless it's for a
929 * different zone.
930 */
28b6e0c1
JR		 931 bool cached = skb_nfct_cached(net, key, info, skb);
932 enum ip_conntrack_info ctinfo;
933 struct nf_conn *ct;
934
935 if (!cached) {
93e66024
FW		 936 struct nf_hook_state state = {
937 .hook = NF_INET_PRE_ROUTING,
938 .pf = info->family,
939 .net = net,
940 };
7f8a436e 941 struct nf_conn *tmpl = info->ct;
5b6b9293 942 int err;
7f8a436e
JS		 943
944 /* Associate skb with specified zone. */
945 if (tmpl) {
cb9c6836
FW		 946 if (skb_nfct(skb))
947 nf_conntrack_put(skb_nfct(skb));
7f8a436e 948 nf_conntrack_get(&tmpl->ct_general);
c74454fa 949 nf_ct_set(skb, tmpl, IP_CT_NEW);
7f8a436e
JS		 950 }
951
93e66024 952 err = nf_conntrack_in(skb, &state);
5b6b9293 953 if (err != NF_ACCEPT)
7f8a436e 954 return -ENOENT;
cae3a262 955
05752523
JR		 956 /* Clear CT state NAT flags to mark that we have not yet done
957 * NAT after the nf_conntrack_in() call. We can actually clear
958 * the whole state, as it will be re-initialized below.
959 */
316d4d78 960 key->ct_state = 0;
05752523
JR		 961
962 /* Update the key, but keep the NAT flags. */
963 ovs_ct_update_key(skb, info, key, true, true);
28b6e0c1 964 }
394e910e 965
28b6e0c1 966 ct = nf_ct_get(skb, &ctinfo);
05752523
JR		 967 if (ct) {
968 /* Packets starting a new connection must be NATted before the
969 * helper, so that the helper knows about the NAT. We enforce
970 * this by delaying both NAT and helper calls for unconfirmed
971 * connections until the committing CT action. For later
972 * packets NAT and Helper may be called in either order.
973 *
974 * NAT will be done only if the CT action has NAT, and only
975 * once per packet (per zone), as guarded by the NAT bits in
316d4d78 976 * the key->ct_state.
05752523 977 */
316d4d78 978 if (info->nat && !(key->ct_state & OVS_CS_F_NAT_MASK) &&
05752523
JR		 979 (nf_ct_is_confirmed(ct) || info->commit) &&
980 ovs_ct_nat(net, key, info, skb, ct, ctinfo) != NF_ACCEPT) {
981 return -EINVAL;
982 }
983
16ec3d4f
JS		 984 /* Userspace may decide to perform a ct lookup without a helper
985 * specified followed by a (recirculate and) commit with one.
986 * Therefore, for unconfirmed connections which we will commit,
987 * we need to attach the helper here.
988 */
989 if (!nf_ct_is_confirmed(ct) && info->commit &&
990 info->helper && !nfct_help(ct)) {
991 int err = __nf_ct_try_assign_helper(ct, info->ct,
992 GFP_ATOMIC);
993 if (err)
994 return err;
fa7e428c
FL		 995
996 /* helper installed, add seqadj if NAT is required */
997 if (info->nat && !nfct_seqadj(ct)) {
998 if (!nfct_seqadj_ext_add(ct))
999 return -EINVAL;
1000 }
16ec3d4f
JS		 1001 }
1002
05752523
JR		 1003 /* Call the helper only if:
1004 * - nf_conntrack_in() was executed above ("!cached") for a
1005 * confirmed connection, or
1006 * - When committing an unconfirmed connection.
1007 */
1008 if ((nf_ct_is_confirmed(ct) ? !cached : info->commit) &&
1009 ovs_ct_helper(skb, info->family) != NF_ACCEPT) {
1010 return -EINVAL;
1011 }
7f8a436e
JS		 1012 }
1013
1014 return 0;
1015}
1016
1017/* Lookup connection and read fields into key. */
1018static int ovs_ct_lookup(struct net *net, struct sw_flow_key *key,
1019 const struct ovs_conntrack_info *info,
1020 struct sk_buff *skb)
1021{
1022 struct nf_conntrack_expect *exp;
1023
9f13ded8
JR		 1024 /* If we pass an expected packet through nf_conntrack_in() the
1025 * expectation is typically removed, but the packet could still be
1026 * lost in upcall processing. To prevent this from happening we
1027 * perform an explicit expectation lookup. Expected connections are
1028 * always new, and will be passed through conntrack only when they are
1029 * committed, as it is OK to remove the expectation at that time.
1030 */
7f8a436e
JS		 1031 exp = ovs_ct_expect_find(net, &info->zone, info->family, skb);
1032 if (exp) {
1033 u8 state;
1034
05752523
JR		 1035 /* NOTE: New connections are NATted and Helped only when
1036 * committed, so we are not calling into NAT here.
1037 */
7f8a436e 1038 state = OVS_CS_F_TRACKED | OVS_CS_F_NEW | OVS_CS_F_RELATED;
182e3042 1039 __ovs_ct_update_key(key, state, &info->zone, exp->master);
d913d3a7
SG		 1040 } else {
1041 struct nf_conn *ct;
1042 int err;
1043
1044 err = __ovs_ct_lookup(net, key, info, skb);
1045 if (err)
1046 return err;
1047
cb9c6836 1048 ct = (struct nf_conn *)skb_nfct(skb);
d913d3a7
SG		 1049 if (ct)
1050 nf_ct_deliver_cached_events(ct);
1051 }
7f8a436e
JS		 1052
1053 return 0;
1054}
1055
33db4125 1056static bool labels_nonzero(const struct ovs_key_ct_labels *labels)
c2ac6673
JS		 1057{
1058 size_t i;
1059
cb80d58f
JR		 1060 for (i = 0; i < OVS_CT_LABELS_LEN_32; i++)
1061 if (labels->ct_labels_32[i])
c2ac6673
JS		 1062 return true;
1063
1064 return false;
1065}
1066
11efd5cb
YHW		 1067#if IS_ENABLED(CONFIG_NETFILTER_CONNCOUNT)
1068static struct hlist_head *ct_limit_hash_bucket(
1069 const struct ovs_ct_limit_info *info, u16 zone)
1070{
1071 return &info->limits[zone & (CT_LIMIT_HASH_BUCKETS - 1)];
1072}
1073
1074/* Call with ovs_mutex */
1075static void ct_limit_set(const struct ovs_ct_limit_info *info,
1076 struct ovs_ct_limit *new_ct_limit)
1077{
1078 struct ovs_ct_limit *ct_limit;
1079 struct hlist_head *head;
1080
1081 head = ct_limit_hash_bucket(info, new_ct_limit->zone);
1082 hlist_for_each_entry_rcu(ct_limit, head, hlist_node) {
1083 if (ct_limit->zone == new_ct_limit->zone) {
1084 hlist_replace_rcu(&ct_limit->hlist_node,
1085 &new_ct_limit->hlist_node);
1086 kfree_rcu(ct_limit, rcu);
1087 return;
1088 }
1089 }
1090
1091 hlist_add_head_rcu(&new_ct_limit->hlist_node, head);
1092}
1093
1094/* Call with ovs_mutex */
1095static void ct_limit_del(const struct ovs_ct_limit_info *info, u16 zone)
1096{
1097 struct ovs_ct_limit *ct_limit;
1098 struct hlist_head *head;
1099 struct hlist_node *n;
1100
1101 head = ct_limit_hash_bucket(info, zone);
1102 hlist_for_each_entry_safe(ct_limit, n, head, hlist_node) {
1103 if (ct_limit->zone == zone) {
1104 hlist_del_rcu(&ct_limit->hlist_node);
1105 kfree_rcu(ct_limit, rcu);
1106 return;
1107 }
1108 }
1109}
1110
1111/* Call with RCU read lock */
1112static u32 ct_limit_get(const struct ovs_ct_limit_info *info, u16 zone)
1113{
1114 struct ovs_ct_limit *ct_limit;
1115 struct hlist_head *head;
1116
1117 head = ct_limit_hash_bucket(info, zone);
1118 hlist_for_each_entry_rcu(ct_limit, head, hlist_node) {
1119 if (ct_limit->zone == zone)
1120 return ct_limit->limit;
1121 }
1122
1123 return info->default_limit;
1124}
1125
1126static int ovs_ct_check_limit(struct net *net,
1127 const struct ovs_conntrack_info *info,
1128 const struct nf_conntrack_tuple *tuple)
1129{
1130 struct ovs_net *ovs_net = net_generic(net, ovs_net_id);
1131 const struct ovs_ct_limit_info *ct_limit_info = ovs_net->ct_limit_info;
1132 u32 per_zone_limit, connections;
1133 u32 conncount_key;
1134
1135 conncount_key = info->zone.id;
1136
1137 per_zone_limit = ct_limit_get(ct_limit_info, info->zone.id);
1138 if (per_zone_limit == OVS_CT_LIMIT_UNLIMITED)
1139 return 0;
1140
1141 connections = nf_conncount_count(net, ct_limit_info->data,
1142 &conncount_key, tuple, &info->zone);
1143 if (connections > per_zone_limit)
1144 return -ENOMEM;
1145
1146 return 0;
1147}
1148#endif
1149
7d904c7b
JR		 1150/* Lookup connection and confirm if unconfirmed. */
1151static int ovs_ct_commit(struct net *net, struct sw_flow_key *key,
1152 const struct ovs_conntrack_info *info,
1153 struct sk_buff *skb)
1154{
6ffcea79
JR		 1155 enum ip_conntrack_info ctinfo;
1156 struct nf_conn *ct;
7d904c7b
JR		 1157 int err;
1158
1159 err = __ovs_ct_lookup(net, key, info, skb);
1160 if (err)
1161 return err;
1162
6ffcea79
JR		 1163 /* The connection could be invalid, in which case this is a no-op.*/
1164 ct = nf_ct_get(skb, &ctinfo);
1165 if (!ct)
1166 return 0;
1167
11efd5cb
YHW		 1168#if IS_ENABLED(CONFIG_NETFILTER_CONNCOUNT)
1169 if (static_branch_unlikely(&ovs_ct_limit_enabled)) {
1170 if (!nf_ct_is_confirmed(ct)) {
1171 err = ovs_ct_check_limit(net, info,
1172 &ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple);
1173 if (err) {
1174 net_warn_ratelimited("openvswitch: zone: %u "
43d0e960 1175 "exceeds conntrack limit\n",
11efd5cb
YHW		 1176 info->zone.id);
1177 return err;
1178 }
1179 }
1180 }
1181#endif
1182
12064551
JR		 1183 /* Set the conntrack event mask if given. NEW and DELETE events have
1184 * their own groups, but the NFNLGRP_CONNTRACK_UPDATE group listener
1185 * typically would receive many kinds of updates. Setting the event
1186 * mask allows those events to be filtered. The set event mask will
1187 * remain in effect for the lifetime of the connection unless changed
1188 * by a further CT action with both the commit flag and the eventmask
1189 * option. */
1190 if (info->have_eventmask) {
1191 struct nf_conntrack_ecache *cache = nf_ct_ecache_find(ct);
1192
1193 if (cache)
1194 cache->ctmask = info->eventmask;
1195 }
1196
7d904c7b
JR		 1197 /* Apply changes before confirming the connection so that the initial
1198 * conntrack NEW netlink event carries the values given in the CT
1199 * action.
1200 */
1201 if (info->mark.mask) {
6ffcea79 1202 err = ovs_ct_set_mark(ct, key, info->mark.value,
7d904c7b
JR		 1203 info->mark.mask);
1204 if (err)
1205 return err;
1206 }
09aa98ad
JR		 1207 if (!nf_ct_is_confirmed(ct)) {
1208 err = ovs_ct_init_labels(ct, key, &info->labels.value,
1209 &info->labels.mask);
1210 if (err)
1211 return err;
a277d516
AB		 1212 } else if (IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS) &&
1213 labels_nonzero(&info->labels.mask)) {
09aa98ad
JR		 1214 err = ovs_ct_set_labels(ct, key, &info->labels.value,
1215 &info->labels.mask);
7d904c7b
JR		 1216 if (err)
1217 return err;
1218 }
1219 /* This will take care of sending queued events even if the connection
1220 * is already confirmed.
1221 */
1222 if (nf_conntrack_confirm(skb) != NF_ACCEPT)
1223 return -EINVAL;
1224
1225 return 0;
1226}
1227
9382fe71
ES		 1228/* Trim the skb to the length specified by the IP/IPv6 header,
1229 * removing any trailing lower-layer padding. This prepares the skb
1230 * for higher-layer processing that assumes skb->len excludes padding
1231 * (such as nf_ip_checksum). The caller needs to pull the skb to the
1232 * network header, and ensure ip_hdr/ipv6_hdr points to valid data.
1233 */
1234static int ovs_skb_network_trim(struct sk_buff *skb)
1235{
1236 unsigned int len;
1237 int err;
1238
1239 switch (skb->protocol) {
1240 case htons(ETH_P_IP):
1241 len = ntohs(ip_hdr(skb)->tot_len);
1242 break;
1243 case htons(ETH_P_IPV6):
1244 len = sizeof(struct ipv6hdr)
1245 + ntohs(ipv6_hdr(skb)->payload_len);
1246 break;
1247 default:
1248 len = skb->len;
1249 }
1250
1251 err = pskb_trim_rcsum(skb, len);
1252 if (err)
1253 kfree_skb(skb);
1254
1255 return err;
1256}
1257
74c16618
JS		 1258/* Returns 0 on success, -EINPROGRESS if 'skb' is stolen, or other nonzero
1259 * value if 'skb' is freed.
1260 */
7f8a436e
JS		 1261int ovs_ct_execute(struct net *net, struct sk_buff *skb,
1262 struct sw_flow_key *key,
1263 const struct ovs_conntrack_info *info)
1264{
1265 int nh_ofs;
1266 int err;
1267
1268 /* The conntrack module expects to be working at L3. */
1269 nh_ofs = skb_network_offset(skb);
75f01a4c 1270 skb_pull_rcsum(skb, nh_ofs);
7f8a436e 1271
9382fe71
ES		 1272 err = ovs_skb_network_trim(skb);
1273 if (err)
1274 return err;
1275
7f8a436e
JS		 1276 if (key->ip.frag != OVS_FRAG_TYPE_NONE) {
1277 err = handle_fragments(net, key, info->zone.id, skb);
1278 if (err)
1279 return err;
1280 }
1281
ab38a7b5 1282 if (info->commit)
7d904c7b 1283 err = ovs_ct_commit(net, key, info, skb);
7f8a436e
JS		 1284 else
1285 err = ovs_ct_lookup(net, key, info, skb);
1286
1287 skb_push(skb, nh_ofs);
75f01a4c 1288 skb_postpush_rcsum(skb, skb->data, nh_ofs);
74c16618
JS		 1289 if (err)
1290 kfree_skb(skb);
7f8a436e
JS		 1291 return err;
1292}
1293
b8226962
EG		 1294int ovs_ct_clear(struct sk_buff *skb, struct sw_flow_key *key)
1295{
1296 if (skb_nfct(skb)) {
1297 nf_conntrack_put(skb_nfct(skb));
1298 nf_ct_set(skb, NULL, IP_CT_UNTRACKED);
1299 ovs_ct_fill_key(skb, key);
1300 }
1301
1302 return 0;
1303}
1304
cae3a262
JS		 1305static int ovs_ct_add_helper(struct ovs_conntrack_info *info, const char *name,
1306 const struct sw_flow_key *key, bool log)
1307{
1308 struct nf_conntrack_helper *helper;
1309 struct nf_conn_help *help;
1310
1311 helper = nf_conntrack_helper_try_module_get(name, info->family,
1312 key->ip.proto);
1313 if (!helper) {
1314 OVS_NLERR(log, "Unknown helper \"%s\"", name);
1315 return -EINVAL;
1316 }
1317
440534d3 1318 help = nf_ct_helper_ext_add(info->ct, GFP_KERNEL);
cae3a262 1319 if (!help) {
d91fc59c 1320 nf_conntrack_helper_put(helper);
cae3a262
JS		 1321 return -ENOMEM;
1322 }
1323
1324 rcu_assign_pointer(help->helper, helper);
1325 info->helper = helper;
17c357ef
FL		 1326
1327 if (info->nat)
1328 request_module("ip_nat_%s", name);
1329
cae3a262
JS		 1330 return 0;
1331}
1332
4806e975 1333#if IS_ENABLED(CONFIG_NF_NAT)
05752523
JR		 1334static int parse_nat(const struct nlattr *attr,
1335 struct ovs_conntrack_info *info, bool log)
1336{
1337 struct nlattr *a;
1338 int rem;
1339 bool have_ip_max = false;
1340 bool have_proto_max = false;
1341 bool ip_vers = (info->family == NFPROTO_IPV6);
1342
1343 nla_for_each_nested(a, attr, rem) {
1344 static const int ovs_nat_attr_lens[OVS_NAT_ATTR_MAX + 1][2] = {
1345 [OVS_NAT_ATTR_SRC] = {0, 0},
1346 [OVS_NAT_ATTR_DST] = {0, 0},
1347 [OVS_NAT_ATTR_IP_MIN] = {sizeof(struct in_addr),
1348 sizeof(struct in6_addr)},
1349 [OVS_NAT_ATTR_IP_MAX] = {sizeof(struct in_addr),
1350 sizeof(struct in6_addr)},
1351 [OVS_NAT_ATTR_PROTO_MIN] = {sizeof(u16), sizeof(u16)},
1352 [OVS_NAT_ATTR_PROTO_MAX] = {sizeof(u16), sizeof(u16)},
1353 [OVS_NAT_ATTR_PERSISTENT] = {0, 0},
1354 [OVS_NAT_ATTR_PROTO_HASH] = {0, 0},
1355 [OVS_NAT_ATTR_PROTO_RANDOM] = {0, 0},
1356 };
1357 int type = nla_type(a);
1358
1359 if (type > OVS_NAT_ATTR_MAX) {
0ed80da5 1360 OVS_NLERR(log, "Unknown NAT attribute (type=%d, max=%d)",
05752523
JR		 1361 type, OVS_NAT_ATTR_MAX);
1362 return -EINVAL;
1363 }
1364
1365 if (nla_len(a) != ovs_nat_attr_lens[type][ip_vers]) {
0ed80da5 1366 OVS_NLERR(log, "NAT attribute type %d has unexpected length (%d != %d)",
05752523
JR		 1367 type, nla_len(a),
1368 ovs_nat_attr_lens[type][ip_vers]);
1369 return -EINVAL;
1370 }
1371
1372 switch (type) {
1373 case OVS_NAT_ATTR_SRC:
1374 case OVS_NAT_ATTR_DST:
1375 if (info->nat) {
0ed80da5 1376 OVS_NLERR(log, "Only one type of NAT may be specified");
05752523
JR		 1377 return -ERANGE;
1378 }
1379 info->nat |= OVS_CT_NAT;
1380 info->nat |= ((type == OVS_NAT_ATTR_SRC)
1381 ? OVS_CT_SRC_NAT : OVS_CT_DST_NAT);
1382 break;
1383
1384 case OVS_NAT_ATTR_IP_MIN:
ac71b46e
HY		 1385 nla_memcpy(&info->range.min_addr, a,
1386 sizeof(info->range.min_addr));
05752523
JR		 1387 info->range.flags |= NF_NAT_RANGE_MAP_IPS;
1388 break;
1389
1390 case OVS_NAT_ATTR_IP_MAX:
1391 have_ip_max = true;
1392 nla_memcpy(&info->range.max_addr, a,
1393 sizeof(info->range.max_addr));
1394 info->range.flags |= NF_NAT_RANGE_MAP_IPS;
1395 break;
1396
1397 case OVS_NAT_ATTR_PROTO_MIN:
1398 info->range.min_proto.all = htons(nla_get_u16(a));
1399 info->range.flags |= NF_NAT_RANGE_PROTO_SPECIFIED;
1400 break;
1401
1402 case OVS_NAT_ATTR_PROTO_MAX:
1403 have_proto_max = true;
1404 info->range.max_proto.all = htons(nla_get_u16(a));
1405 info->range.flags |= NF_NAT_RANGE_PROTO_SPECIFIED;
1406 break;
1407
1408 case OVS_NAT_ATTR_PERSISTENT:
1409 info->range.flags |= NF_NAT_RANGE_PERSISTENT;
1410 break;
1411
1412 case OVS_NAT_ATTR_PROTO_HASH:
1413 info->range.flags |= NF_NAT_RANGE_PROTO_RANDOM;
1414 break;
1415
1416 case OVS_NAT_ATTR_PROTO_RANDOM:
1417 info->range.flags |= NF_NAT_RANGE_PROTO_RANDOM_FULLY;
1418 break;
1419
1420 default:
0ed80da5 1421 OVS_NLERR(log, "Unknown nat attribute (%d)", type);
05752523
JR		 1422 return -EINVAL;
1423 }
1424 }
1425
1426 if (rem > 0) {
0ed80da5 1427 OVS_NLERR(log, "NAT attribute has %d unknown bytes", rem);
05752523
JR		 1428 return -EINVAL;
1429 }
1430 if (!info->nat) {
1431 /* Do not allow flags if no type is given. */
1432 if (info->range.flags) {
1433 OVS_NLERR(log,
e0b10844 1434 "NAT flags may be given only when NAT range (SRC or DST) is also specified."
05752523
JR		 1435 );
1436 return -EINVAL;
1437 }
1438 info->nat = OVS_CT_NAT; /* NAT existing connections. */
1439 } else if (!info->commit) {
1440 OVS_NLERR(log,
e0b10844 1441 "NAT attributes may be specified only when CT COMMIT flag is also specified."
05752523
JR		 1442 );
1443 return -EINVAL;
1444 }
1445 /* Allow missing IP_MAX. */
1446 if (info->range.flags & NF_NAT_RANGE_MAP_IPS && !have_ip_max) {
1447 memcpy(&info->range.max_addr, &info->range.min_addr,
1448 sizeof(info->range.max_addr));
1449 }
1450 /* Allow missing PROTO_MAX. */
1451 if (info->range.flags & NF_NAT_RANGE_PROTO_SPECIFIED &&
1452 !have_proto_max) {
1453 info->range.max_proto.all = info->range.min_proto.all;
1454 }
1455 return 0;
1456}
1457#endif
1458
7f8a436e 1459static const struct ovs_ct_len_tbl ovs_ct_attr_lens[OVS_CT_ATTR_MAX + 1] = {
ab38a7b5 1460 [OVS_CT_ATTR_COMMIT] = { .minlen = 0, .maxlen = 0 },
dd41d33f 1461 [OVS_CT_ATTR_FORCE_COMMIT] = { .minlen = 0, .maxlen = 0 },
7f8a436e
JS		 1462 [OVS_CT_ATTR_ZONE] = { .minlen = sizeof(u16),
1463 .maxlen = sizeof(u16) },
182e3042
JS		 1464 [OVS_CT_ATTR_MARK] = { .minlen = sizeof(struct md_mark),
1465 .maxlen = sizeof(struct md_mark) },
33db4125
JS		 1466 [OVS_CT_ATTR_LABELS] = { .minlen = sizeof(struct md_labels),
1467 .maxlen = sizeof(struct md_labels) },
cae3a262 1468 [OVS_CT_ATTR_HELPER] = { .minlen = 1,
05752523 1469 .maxlen = NF_CT_HELPER_NAME_LEN },
4806e975 1470#if IS_ENABLED(CONFIG_NF_NAT)
05752523
JR		 1471 /* NAT length is checked when parsing the nested attributes. */
1472 [OVS_CT_ATTR_NAT] = { .minlen = 0, .maxlen = INT_MAX },
1473#endif
12064551
JR		 1474 [OVS_CT_ATTR_EVENTMASK] = { .minlen = sizeof(u32),
1475 .maxlen = sizeof(u32) },
06bd2bdf
YHW		 1476 [OVS_CT_ATTR_TIMEOUT] = { .minlen = 1,
1477 .maxlen = CTNL_TIMEOUT_NAME_MAX },
7f8a436e
JS		 1478};
1479
1480static int parse_ct(const struct nlattr *attr, struct ovs_conntrack_info *info,
cae3a262 1481 const char **helper, bool log)
7f8a436e
JS		 1482{
1483 struct nlattr *a;
1484 int rem;
1485
1486 nla_for_each_nested(a, attr, rem) {
1487 int type = nla_type(a);
69ec932e
LZ		 1488 int maxlen;
1489 int minlen;
7f8a436e
JS		 1490
1491 if (type > OVS_CT_ATTR_MAX) {
1492 OVS_NLERR(log,
1493 "Unknown conntrack attr (type=%d, max=%d)",
1494 type, OVS_CT_ATTR_MAX);
1495 return -EINVAL;
1496 }
69ec932e
LZ		 1497
1498 maxlen = ovs_ct_attr_lens[type].maxlen;
1499 minlen = ovs_ct_attr_lens[type].minlen;
7f8a436e
JS		 1500 if (nla_len(a) < minlen || nla_len(a) > maxlen) {
1501 OVS_NLERR(log,
1502 "Conntrack attr type has unexpected length (type=%d, length=%d, expected=%d)",
1503 type, nla_len(a), maxlen);
1504 return -EINVAL;
1505 }
1506
1507 switch (type) {
dd41d33f
JR		 1508 case OVS_CT_ATTR_FORCE_COMMIT:
1509 info->force = true;
1510 /* fall through. */
ab38a7b5
JS		 1511 case OVS_CT_ATTR_COMMIT:
1512 info->commit = true;
7f8a436e
JS		 1513 break;
1514#ifdef CONFIG_NF_CONNTRACK_ZONES
1515 case OVS_CT_ATTR_ZONE:
1516 info->zone.id = nla_get_u16(a);
1517 break;
182e3042
JS		 1518#endif
1519#ifdef CONFIG_NF_CONNTRACK_MARK
1520 case OVS_CT_ATTR_MARK: {
1521 struct md_mark *mark = nla_data(a);
1522
e754ec69
JS		 1523 if (!mark->mask) {
1524 OVS_NLERR(log, "ct_mark mask cannot be 0");
1525 return -EINVAL;
1526 }
182e3042
JS		 1527 info->mark = *mark;
1528 break;
1529 }
c2ac6673
JS		 1530#endif
1531#ifdef CONFIG_NF_CONNTRACK_LABELS
33db4125
JS		 1532 case OVS_CT_ATTR_LABELS: {
1533 struct md_labels *labels = nla_data(a);
c2ac6673 1534
e754ec69
JS		 1535 if (!labels_nonzero(&labels->mask)) {
1536 OVS_NLERR(log, "ct_labels mask cannot be 0");
1537 return -EINVAL;
1538 }
33db4125 1539 info->labels = *labels;
c2ac6673
JS		 1540 break;
1541 }
7f8a436e 1542#endif
cae3a262
JS		 1543 case OVS_CT_ATTR_HELPER:
1544 *helper = nla_data(a);
1545 if (!memchr(*helper, '\0', nla_len(a))) {
1546 OVS_NLERR(log, "Invalid conntrack helper");
1547 return -EINVAL;
1548 }
1549 break;
4806e975 1550#if IS_ENABLED(CONFIG_NF_NAT)
05752523
JR		 1551 case OVS_CT_ATTR_NAT: {
1552 int err = parse_nat(a, info, log);
1553
1554 if (err)
1555 return err;
1556 break;
1557 }
1558#endif
12064551
JR		 1559 case OVS_CT_ATTR_EVENTMASK:
1560 info->have_eventmask = true;
1561 info->eventmask = nla_get_u32(a);
1562 break;
06bd2bdf
YHW		 1563#ifdef CONFIG_NF_CONNTRACK_TIMEOUT
1564 case OVS_CT_ATTR_TIMEOUT:
1565 memcpy(info->timeout, nla_data(a), nla_len(a));
1566 if (!memchr(info->timeout, '\0', nla_len(a))) {
1567 OVS_NLERR(log, "Invalid conntrack helper");
1568 return -EINVAL;
1569 }
1570 break;
1571#endif
12064551 1572
7f8a436e
JS		 1573 default:
1574 OVS_NLERR(log, "Unknown conntrack attr (%d)",
1575 type);
1576 return -EINVAL;
1577 }
1578 }
1579
7d904c7b
JR		 1580#ifdef CONFIG_NF_CONNTRACK_MARK
1581 if (!info->commit && info->mark.mask) {
1582 OVS_NLERR(log,
1583 "Setting conntrack mark requires 'commit' flag.");
1584 return -EINVAL;
1585 }
1586#endif
1587#ifdef CONFIG_NF_CONNTRACK_LABELS
1588 if (!info->commit && labels_nonzero(&info->labels.mask)) {
1589 OVS_NLERR(log,
1590 "Setting conntrack labels requires 'commit' flag.");
1591 return -EINVAL;
1592 }
1593#endif
7f8a436e
JS		 1594 if (rem > 0) {
1595 OVS_NLERR(log, "Conntrack attr has %d unknown bytes", rem);
1596 return -EINVAL;
1597 }
1598
1599 return 0;
1600}
1601
c2ac6673 1602bool ovs_ct_verify(struct net *net, enum ovs_key_attr attr)
7f8a436e
JS		 1603{
1604 if (attr == OVS_KEY_ATTR_CT_STATE)
1605 return true;
1606 if (IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES) &&
1607 attr == OVS_KEY_ATTR_CT_ZONE)
1608 return true;
182e3042
JS		 1609 if (IS_ENABLED(CONFIG_NF_CONNTRACK_MARK) &&
1610 attr == OVS_KEY_ATTR_CT_MARK)
1611 return true;
c2ac6673 1612 if (IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS) &&
33db4125 1613 attr == OVS_KEY_ATTR_CT_LABELS) {
c2ac6673
JS		 1614 struct ovs_net *ovs_net = net_generic(net, ovs_net_id);
1615
1616 return ovs_net->xt_label;
1617 }
7f8a436e
JS		 1618
1619 return false;
1620}
1621
1622int ovs_ct_copy_action(struct net *net, const struct nlattr *attr,
1623 const struct sw_flow_key *key,
1624 struct sw_flow_actions **sfa, bool log)
1625{
1626 struct ovs_conntrack_info ct_info;
cae3a262 1627 const char *helper = NULL;
7f8a436e
JS		 1628 u16 family;
1629 int err;
1630
1631 family = key_to_nfproto(key);
1632 if (family == NFPROTO_UNSPEC) {
1633 OVS_NLERR(log, "ct family unspecified");
1634 return -EINVAL;
1635 }
1636
1637 memset(&ct_info, 0, sizeof(ct_info));
1638 ct_info.family = family;
1639
1640 nf_ct_zone_init(&ct_info.zone, NF_CT_DEFAULT_ZONE_ID,
1641 NF_CT_DEFAULT_ZONE_DIR, 0);
1642
cae3a262 1643 err = parse_ct(attr, &ct_info, &helper, log);
7f8a436e
JS		 1644 if (err)
1645 return err;
1646
1647 /* Set up template for tracking connections in specific zones. */
1648 ct_info.ct = nf_ct_tmpl_alloc(net, &ct_info.zone, GFP_KERNEL);
1649 if (!ct_info.ct) {
1650 OVS_NLERR(log, "Failed to allocate conntrack template");
1651 return -ENOMEM;
1652 }
06bd2bdf
YHW		 1653
1654 if (ct_info.timeout[0]) {
1655 if (nf_ct_set_timeout(net, ct_info.ct, family, key->ip.proto,
1656 ct_info.timeout))
1657 pr_info_ratelimited("Failed to associated timeout "
1658 "policy `%s'\n", ct_info.timeout);
1659 }
1660
cae3a262
JS		 1661 if (helper) {
1662 err = ovs_ct_add_helper(&ct_info, helper, key, log);
1663 if (err)
1664 goto err_free_ct;
1665 }
7f8a436e
JS		 1666
1667 err = ovs_nla_add_action(sfa, OVS_ACTION_ATTR_CT, &ct_info,
1668 sizeof(ct_info), log);
1669 if (err)
1670 goto err_free_ct;
1671
7f6d6558
FL		 1672 __set_bit(IPS_CONFIRMED_BIT, &ct_info.ct->status);
1673 nf_conntrack_get(&ct_info.ct->ct_general);
7f8a436e
JS		 1674 return 0;
1675err_free_ct:
2f3ab9f9 1676 __ovs_ct_free_action(&ct_info);
7f8a436e
JS		 1677 return err;
1678}
1679
4806e975 1680#if IS_ENABLED(CONFIG_NF_NAT)
05752523
JR		 1681static bool ovs_ct_nat_to_attr(const struct ovs_conntrack_info *info,
1682 struct sk_buff *skb)
1683{
1684 struct nlattr *start;
1685
ae0be8de 1686 start = nla_nest_start_noflag(skb, OVS_CT_ATTR_NAT);
05752523
JR		 1687 if (!start)
1688 return false;
1689
1690 if (info->nat & OVS_CT_SRC_NAT) {
1691 if (nla_put_flag(skb, OVS_NAT_ATTR_SRC))
1692 return false;
1693 } else if (info->nat & OVS_CT_DST_NAT) {
1694 if (nla_put_flag(skb, OVS_NAT_ATTR_DST))
1695 return false;
1696 } else {
1697 goto out;
1698 }
1699
1700 if (info->range.flags & NF_NAT_RANGE_MAP_IPS) {
3bf195ae 1701 if (IS_ENABLED(CONFIG_NF_NAT) &&
99b7248e 1702 info->family == NFPROTO_IPV4) {
05752523
JR		 1703 if (nla_put_in_addr(skb, OVS_NAT_ATTR_IP_MIN,
1704 info->range.min_addr.ip) ||
1705 (info->range.max_addr.ip
1706 != info->range.min_addr.ip &&
1707 (nla_put_in_addr(skb, OVS_NAT_ATTR_IP_MAX,
1708 info->range.max_addr.ip))))
1709 return false;
3bf195ae 1710 } else if (IS_ENABLED(CONFIG_IPV6) &&
99b7248e 1711 info->family == NFPROTO_IPV6) {
05752523
JR		 1712 if (nla_put_in6_addr(skb, OVS_NAT_ATTR_IP_MIN,
1713 &info->range.min_addr.in6) ||
1714 (memcmp(&info->range.max_addr.in6,
1715 &info->range.min_addr.in6,
1716 sizeof(info->range.max_addr.in6)) &&
1717 (nla_put_in6_addr(skb, OVS_NAT_ATTR_IP_MAX,
1718 &info->range.max_addr.in6))))
1719 return false;
05752523
JR		 1720 } else {
1721 return false;
1722 }
1723 }
1724 if (info->range.flags & NF_NAT_RANGE_PROTO_SPECIFIED &&
1725 (nla_put_u16(skb, OVS_NAT_ATTR_PROTO_MIN,
1726 ntohs(info->range.min_proto.all)) ||
1727 (info->range.max_proto.all != info->range.min_proto.all &&
1728 nla_put_u16(skb, OVS_NAT_ATTR_PROTO_MAX,
1729 ntohs(info->range.max_proto.all)))))
1730 return false;
1731
1732 if (info->range.flags & NF_NAT_RANGE_PERSISTENT &&
1733 nla_put_flag(skb, OVS_NAT_ATTR_PERSISTENT))
1734 return false;
1735 if (info->range.flags & NF_NAT_RANGE_PROTO_RANDOM &&
1736 nla_put_flag(skb, OVS_NAT_ATTR_PROTO_HASH))
1737 return false;
1738 if (info->range.flags & NF_NAT_RANGE_PROTO_RANDOM_FULLY &&
1739 nla_put_flag(skb, OVS_NAT_ATTR_PROTO_RANDOM))
1740 return false;
1741out:
1742 nla_nest_end(skb, start);
1743
1744 return true;
1745}
1746#endif
1747
7f8a436e
JS		 1748int ovs_ct_action_to_attr(const struct ovs_conntrack_info *ct_info,
1749 struct sk_buff *skb)
1750{
1751 struct nlattr *start;
1752
ae0be8de 1753 start = nla_nest_start_noflag(skb, OVS_ACTION_ATTR_CT);
7f8a436e
JS		 1754 if (!start)
1755 return -EMSGSIZE;
1756
dd41d33f
JR		 1757 if (ct_info->commit && nla_put_flag(skb, ct_info->force
1758 ? OVS_CT_ATTR_FORCE_COMMIT
1759 : OVS_CT_ATTR_COMMIT))
7f8a436e
JS		 1760 return -EMSGSIZE;
1761 if (IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES) &&
1762 nla_put_u16(skb, OVS_CT_ATTR_ZONE, ct_info->zone.id))
1763 return -EMSGSIZE;
e754ec69 1764 if (IS_ENABLED(CONFIG_NF_CONNTRACK_MARK) && ct_info->mark.mask &&
182e3042
JS		 1765 nla_put(skb, OVS_CT_ATTR_MARK, sizeof(ct_info->mark),
1766 &ct_info->mark))
1767 return -EMSGSIZE;
c2ac6673 1768 if (IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS) &&
e754ec69 1769 labels_nonzero(&ct_info->labels.mask) &&
33db4125
JS		 1770 nla_put(skb, OVS_CT_ATTR_LABELS, sizeof(ct_info->labels),
1771 &ct_info->labels))
c2ac6673 1772 return -EMSGSIZE;
cae3a262
JS		 1773 if (ct_info->helper) {
1774 if (nla_put_string(skb, OVS_CT_ATTR_HELPER,
1775 ct_info->helper->name))
1776 return -EMSGSIZE;
1777 }
12064551
JR		 1778 if (ct_info->have_eventmask &&
1779 nla_put_u32(skb, OVS_CT_ATTR_EVENTMASK, ct_info->eventmask))
1780 return -EMSGSIZE;
06bd2bdf
YHW		 1781 if (ct_info->timeout[0]) {
1782 if (nla_put_string(skb, OVS_CT_ATTR_TIMEOUT, ct_info->timeout))
1783 return -EMSGSIZE;
1784 }
12064551 1785
4806e975 1786#if IS_ENABLED(CONFIG_NF_NAT)
05752523
JR		 1787 if (ct_info->nat && !ovs_ct_nat_to_attr(ct_info, skb))
1788 return -EMSGSIZE;
1789#endif
7f8a436e
JS		 1790 nla_nest_end(skb, start);
1791
1792 return 0;
1793}
1794
1795void ovs_ct_free_action(const struct nlattr *a)
1796{
1797 struct ovs_conntrack_info *ct_info = nla_data(a);
1798
2f3ab9f9
JS		 1799 __ovs_ct_free_action(ct_info);
1800}
1801
1802static void __ovs_ct_free_action(struct ovs_conntrack_info *ct_info)
1803{
cae3a262 1804 if (ct_info->helper)
d91fc59c 1805 nf_conntrack_helper_put(ct_info->helper);
06bd2bdf 1806 if (ct_info->ct) {
06bd2bdf
YHW		 1807 if (ct_info->timeout[0])
1808 nf_ct_destroy_timeout(ct_info->ct);
6d670497 1809 nf_ct_tmpl_free(ct_info->ct);
06bd2bdf 1810 }
7f8a436e 1811}
c2ac6673 1812
11efd5cb
YHW		 1813#if IS_ENABLED(CONFIG_NETFILTER_CONNCOUNT)
1814static int ovs_ct_limit_init(struct net *net, struct ovs_net *ovs_net)
1815{
1816 int i, err;
1817
1818 ovs_net->ct_limit_info = kmalloc(sizeof(*ovs_net->ct_limit_info),
1819 GFP_KERNEL);
1820 if (!ovs_net->ct_limit_info)
1821 return -ENOMEM;
1822
1823 ovs_net->ct_limit_info->default_limit = OVS_CT_LIMIT_DEFAULT;
1824 ovs_net->ct_limit_info->limits =
1825 kmalloc_array(CT_LIMIT_HASH_BUCKETS, sizeof(struct hlist_head),
1826 GFP_KERNEL);
1827 if (!ovs_net->ct_limit_info->limits) {
1828 kfree(ovs_net->ct_limit_info);
1829 return -ENOMEM;
1830 }
1831
1832 for (i = 0; i < CT_LIMIT_HASH_BUCKETS; i++)
1833 INIT_HLIST_HEAD(&ovs_net->ct_limit_info->limits[i]);
1834
1835 ovs_net->ct_limit_info->data =
1836 nf_conncount_init(net, NFPROTO_INET, sizeof(u32));
1837
1838 if (IS_ERR(ovs_net->ct_limit_info->data)) {
1839 err = PTR_ERR(ovs_net->ct_limit_info->data);
1840 kfree(ovs_net->ct_limit_info->limits);
1841 kfree(ovs_net->ct_limit_info);
1842 pr_err("openvswitch: failed to init nf_conncount %d\n", err);
1843 return err;
1844 }
1845 return 0;
1846}
1847
1848static void ovs_ct_limit_exit(struct net *net, struct ovs_net *ovs_net)
1849{
1850 const struct ovs_ct_limit_info *info = ovs_net->ct_limit_info;
1851 int i;
1852
1853 nf_conncount_destroy(net, NFPROTO_INET, info->data);
1854 for (i = 0; i < CT_LIMIT_HASH_BUCKETS; ++i) {
1855 struct hlist_head *head = &info->limits[i];
1856 struct ovs_ct_limit *ct_limit;
1857
1858 hlist_for_each_entry_rcu(ct_limit, head, hlist_node)
1859 kfree_rcu(ct_limit, rcu);
1860 }
1861 kfree(ovs_net->ct_limit_info->limits);
1862 kfree(ovs_net->ct_limit_info);
1863}
1864
1865static struct sk_buff *
1866ovs_ct_limit_cmd_reply_start(struct genl_info *info, u8 cmd,
1867 struct ovs_header **ovs_reply_header)
1868{
1869 struct ovs_header *ovs_header = info->userhdr;
1870 struct sk_buff *skb;
1871
1872 skb = genlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
1873 if (!skb)
1874 return ERR_PTR(-ENOMEM);
1875
1876 *ovs_reply_header = genlmsg_put(skb, info->snd_portid,
1877 info->snd_seq,
1878 &dp_ct_limit_genl_family, 0, cmd);
1879
1880 if (!*ovs_reply_header) {
1881 nlmsg_free(skb);
1882 return ERR_PTR(-EMSGSIZE);
1883 }
1884 (*ovs_reply_header)->dp_ifindex = ovs_header->dp_ifindex;
1885
1886 return skb;
1887}
1888
1889static bool check_zone_id(int zone_id, u16 *pzone)
1890{
1891 if (zone_id >= 0 && zone_id <= 65535) {
1892 *pzone = (u16)zone_id;
1893 return true;
1894 }
1895 return false;
1896}
1897
1898static int ovs_ct_limit_set_zone_limit(struct nlattr *nla_zone_limit,
1899 struct ovs_ct_limit_info *info)
1900{
1901 struct ovs_zone_limit *zone_limit;
1902 int rem;
1903 u16 zone;
1904
1905 rem = NLA_ALIGN(nla_len(nla_zone_limit));
1906 zone_limit = (struct ovs_zone_limit *)nla_data(nla_zone_limit);
1907
1908 while (rem >= sizeof(*zone_limit)) {
1909 if (unlikely(zone_limit->zone_id ==
1910 OVS_ZONE_LIMIT_DEFAULT_ZONE)) {
1911 ovs_lock();
1912 info->default_limit = zone_limit->limit;
1913 ovs_unlock();
1914 } else if (unlikely(!check_zone_id(
1915 zone_limit->zone_id, &zone))) {
1916 OVS_NLERR(true, "zone id is out of range");
1917 } else {
1918 struct ovs_ct_limit *ct_limit;
1919
1920 ct_limit = kmalloc(sizeof(*ct_limit), GFP_KERNEL);
1921 if (!ct_limit)
1922 return -ENOMEM;
1923
1924 ct_limit->zone = zone;
1925 ct_limit->limit = zone_limit->limit;
1926
1927 ovs_lock();
1928 ct_limit_set(info, ct_limit);
1929 ovs_unlock();
1930 }
1931 rem -= NLA_ALIGN(sizeof(*zone_limit));
1932 zone_limit = (struct ovs_zone_limit *)((u8 *)zone_limit +
1933 NLA_ALIGN(sizeof(*zone_limit)));
1934 }
1935
1936 if (rem)
1937 OVS_NLERR(true, "set zone limit has %d unknown bytes", rem);
1938
1939 return 0;
1940}
1941
1942static int ovs_ct_limit_del_zone_limit(struct nlattr *nla_zone_limit,
1943 struct ovs_ct_limit_info *info)
1944{
1945 struct ovs_zone_limit *zone_limit;
1946 int rem;
1947 u16 zone;
1948
1949 rem = NLA_ALIGN(nla_len(nla_zone_limit));
1950 zone_limit = (struct ovs_zone_limit *)nla_data(nla_zone_limit);
1951
1952 while (rem >= sizeof(*zone_limit)) {
1953 if (unlikely(zone_limit->zone_id ==
1954 OVS_ZONE_LIMIT_DEFAULT_ZONE)) {
1955 ovs_lock();
1956 info->default_limit = OVS_CT_LIMIT_DEFAULT;
1957 ovs_unlock();
1958 } else if (unlikely(!check_zone_id(
1959 zone_limit->zone_id, &zone))) {
1960 OVS_NLERR(true, "zone id is out of range");
1961 } else {
1962 ovs_lock();
1963 ct_limit_del(info, zone);
1964 ovs_unlock();
1965 }
1966 rem -= NLA_ALIGN(sizeof(*zone_limit));
1967 zone_limit = (struct ovs_zone_limit *)((u8 *)zone_limit +
1968 NLA_ALIGN(sizeof(*zone_limit)));
1969 }
1970
1971 if (rem)
1972 OVS_NLERR(true, "del zone limit has %d unknown bytes", rem);
1973
1974 return 0;
1975}
1976
1977static int ovs_ct_limit_get_default_limit(struct ovs_ct_limit_info *info,
1978 struct sk_buff *reply)
1979{
1980 struct ovs_zone_limit zone_limit;
1981 int err;
1982
1983 zone_limit.zone_id = OVS_ZONE_LIMIT_DEFAULT_ZONE;
1984 zone_limit.limit = info->default_limit;
1985 err = nla_put_nohdr(reply, sizeof(zone_limit), &zone_limit);
1986 if (err)
1987 return err;
1988
1989 return 0;
1990}
1991
1992static int __ovs_ct_limit_get_zone_limit(struct net *net,
1993 struct nf_conncount_data *data,
1994 u16 zone_id, u32 limit,
1995 struct sk_buff *reply)
1996{
1997 struct nf_conntrack_zone ct_zone;
1998 struct ovs_zone_limit zone_limit;
1999 u32 conncount_key = zone_id;
2000
2001 zone_limit.zone_id = zone_id;
2002 zone_limit.limit = limit;
2003 nf_ct_zone_init(&ct_zone, zone_id, NF_CT_DEFAULT_ZONE_DIR, 0);
2004
2005 zone_limit.count = nf_conncount_count(net, data, &conncount_key, NULL,
2006 &ct_zone);
2007 return nla_put_nohdr(reply, sizeof(zone_limit), &zone_limit);
2008}
2009
2010static int ovs_ct_limit_get_zone_limit(struct net *net,
2011 struct nlattr *nla_zone_limit,
2012 struct ovs_ct_limit_info *info,
2013 struct sk_buff *reply)
2014{
2015 struct ovs_zone_limit *zone_limit;
2016 int rem, err;
2017 u32 limit;
2018 u16 zone;
2019
2020 rem = NLA_ALIGN(nla_len(nla_zone_limit));
2021 zone_limit = (struct ovs_zone_limit *)nla_data(nla_zone_limit);
2022
2023 while (rem >= sizeof(*zone_limit)) {
2024 if (unlikely(zone_limit->zone_id ==
2025 OVS_ZONE_LIMIT_DEFAULT_ZONE)) {
2026 err = ovs_ct_limit_get_default_limit(info, reply);
2027 if (err)
2028 return err;
2029 } else if (unlikely(!check_zone_id(zone_limit->zone_id,
2030 &zone))) {
2031 OVS_NLERR(true, "zone id is out of range");
2032 } else {
2033 rcu_read_lock();
2034 limit = ct_limit_get(info, zone);
2035 rcu_read_unlock();
2036
2037 err = __ovs_ct_limit_get_zone_limit(
2038 net, info->data, zone, limit, reply);
2039 if (err)
2040 return err;
2041 }
2042 rem -= NLA_ALIGN(sizeof(*zone_limit));
2043 zone_limit = (struct ovs_zone_limit *)((u8 *)zone_limit +
2044 NLA_ALIGN(sizeof(*zone_limit)));
2045 }
2046
2047 if (rem)
2048 OVS_NLERR(true, "get zone limit has %d unknown bytes", rem);
2049
2050 return 0;
2051}
2052
2053static int ovs_ct_limit_get_all_zone_limit(struct net *net,
2054 struct ovs_ct_limit_info *info,
2055 struct sk_buff *reply)
2056{
2057 struct ovs_ct_limit *ct_limit;
2058 struct hlist_head *head;
2059 int i, err = 0;
2060
2061 err = ovs_ct_limit_get_default_limit(info, reply);
2062 if (err)
2063 return err;
2064
2065 rcu_read_lock();
2066 for (i = 0; i < CT_LIMIT_HASH_BUCKETS; ++i) {
2067 head = &info->limits[i];
2068 hlist_for_each_entry_rcu(ct_limit, head, hlist_node) {
2069 err = __ovs_ct_limit_get_zone_limit(net, info->data,
2070 ct_limit->zone, ct_limit->limit, reply);
2071 if (err)
2072 goto exit_err;
2073 }
2074 }
2075
2076exit_err:
2077 rcu_read_unlock();
2078 return err;
2079}
2080
2081static int ovs_ct_limit_cmd_set(struct sk_buff *skb, struct genl_info *info)
2082{
2083 struct nlattr **a = info->attrs;
2084 struct sk_buff *reply;
2085 struct ovs_header *ovs_reply_header;
2086 struct ovs_net *ovs_net = net_generic(sock_net(skb->sk), ovs_net_id);
2087 struct ovs_ct_limit_info *ct_limit_info = ovs_net->ct_limit_info;
2088 int err;
2089
2090 reply = ovs_ct_limit_cmd_reply_start(info, OVS_CT_LIMIT_CMD_SET,
2091 &ovs_reply_header);
2092 if (IS_ERR(reply))
2093 return PTR_ERR(reply);
2094
2095 if (!a[OVS_CT_LIMIT_ATTR_ZONE_LIMIT]) {
2096 err = -EINVAL;
2097 goto exit_err;
2098 }
2099
2100 err = ovs_ct_limit_set_zone_limit(a[OVS_CT_LIMIT_ATTR_ZONE_LIMIT],
2101 ct_limit_info);
2102 if (err)
2103 goto exit_err;
2104
2105 static_branch_enable(&ovs_ct_limit_enabled);
2106
2107 genlmsg_end(reply, ovs_reply_header);
2108 return genlmsg_reply(reply, info);
2109
2110exit_err:
2111 nlmsg_free(reply);
2112 return err;
2113}
2114
2115static int ovs_ct_limit_cmd_del(struct sk_buff *skb, struct genl_info *info)
2116{
2117 struct nlattr **a = info->attrs;
2118 struct sk_buff *reply;
2119 struct ovs_header *ovs_reply_header;
2120 struct ovs_net *ovs_net = net_generic(sock_net(skb->sk), ovs_net_id);
2121 struct ovs_ct_limit_info *ct_limit_info = ovs_net->ct_limit_info;
2122 int err;
2123
2124 reply = ovs_ct_limit_cmd_reply_start(info, OVS_CT_LIMIT_CMD_DEL,
2125 &ovs_reply_header);
2126 if (IS_ERR(reply))
2127 return PTR_ERR(reply);
2128
2129 if (!a[OVS_CT_LIMIT_ATTR_ZONE_LIMIT]) {
2130 err = -EINVAL;
2131 goto exit_err;
2132 }
2133
2134 err = ovs_ct_limit_del_zone_limit(a[OVS_CT_LIMIT_ATTR_ZONE_LIMIT],
2135 ct_limit_info);
2136 if (err)
2137 goto exit_err;
2138
2139 genlmsg_end(reply, ovs_reply_header);
2140 return genlmsg_reply(reply, info);
2141
2142exit_err:
2143 nlmsg_free(reply);
2144 return err;
2145}
2146
2147static int ovs_ct_limit_cmd_get(struct sk_buff *skb, struct genl_info *info)
2148{
2149 struct nlattr **a = info->attrs;
2150 struct nlattr *nla_reply;
2151 struct sk_buff *reply;
2152 struct ovs_header *ovs_reply_header;
2153 struct net *net = sock_net(skb->sk);
2154 struct ovs_net *ovs_net = net_generic(net, ovs_net_id);
2155 struct ovs_ct_limit_info *ct_limit_info = ovs_net->ct_limit_info;
2156 int err;
2157
2158 reply = ovs_ct_limit_cmd_reply_start(info, OVS_CT_LIMIT_CMD_GET,
2159 &ovs_reply_header);
2160 if (IS_ERR(reply))
2161 return PTR_ERR(reply);
2162
ae0be8de 2163 nla_reply = nla_nest_start_noflag(reply, OVS_CT_LIMIT_ATTR_ZONE_LIMIT);
11efd5cb
YHW		 2164
2165 if (a[OVS_CT_LIMIT_ATTR_ZONE_LIMIT]) {
2166 err = ovs_ct_limit_get_zone_limit(
2167 net, a[OVS_CT_LIMIT_ATTR_ZONE_LIMIT], ct_limit_info,
2168 reply);
2169 if (err)
2170 goto exit_err;
2171 } else {
2172 err = ovs_ct_limit_get_all_zone_limit(net, ct_limit_info,
2173 reply);
2174 if (err)
2175 goto exit_err;
2176 }
2177
2178 nla_nest_end(reply, nla_reply);
2179 genlmsg_end(reply, ovs_reply_header);
2180 return genlmsg_reply(reply, info);
2181
2182exit_err:
2183 nlmsg_free(reply);
2184 return err;
2185}
2186
2187static struct genl_ops ct_limit_genl_ops[] = {
2188 { .cmd = OVS_CT_LIMIT_CMD_SET,
ef6243ac 2189 .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
11efd5cb
YHW		 2190 .flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN
2191 * privilege. */
11efd5cb
YHW		 2192 .doit = ovs_ct_limit_cmd_set,
2193 },
2194 { .cmd = OVS_CT_LIMIT_CMD_DEL,
ef6243ac 2195 .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
11efd5cb
YHW		 2196 .flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN
2197 * privilege. */
11efd5cb
YHW		 2198 .doit = ovs_ct_limit_cmd_del,
2199 },
2200 { .cmd = OVS_CT_LIMIT_CMD_GET,
ef6243ac 2201 .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
11efd5cb 2202 .flags = 0, /* OK for unprivileged users. */
11efd5cb
YHW		 2203 .doit = ovs_ct_limit_cmd_get,
2204 },
2205};
2206
2207static const struct genl_multicast_group ovs_ct_limit_multicast_group = {
2208 .name = OVS_CT_LIMIT_MCGROUP,
2209};
2210
2211struct genl_family dp_ct_limit_genl_family __ro_after_init = {
2212 .hdrsize = sizeof(struct ovs_header),
2213 .name = OVS_CT_LIMIT_FAMILY,
2214 .version = OVS_CT_LIMIT_VERSION,
2215 .maxattr = OVS_CT_LIMIT_ATTR_MAX,
3b0f31f2 2216 .policy = ct_limit_policy,
11efd5cb
YHW		 2217 .netnsok = true,
2218 .parallel_ops = true,
2219 .ops = ct_limit_genl_ops,
2220 .n_ops = ARRAY_SIZE(ct_limit_genl_ops),
2221 .mcgrps = &ovs_ct_limit_multicast_group,
2222 .n_mcgrps = 1,
2223 .module = THIS_MODULE,
2224};
2225#endif
2226
2227int ovs_ct_init(struct net *net)
c2ac6673 2228{
33db4125 2229 unsigned int n_bits = sizeof(struct ovs_key_ct_labels) * BITS_PER_BYTE;
c2ac6673
JS		 2230 struct ovs_net *ovs_net = net_generic(net, ovs_net_id);
2231
adff6c65 2232 if (nf_connlabels_get(net, n_bits - 1)) {
c2ac6673
JS		 2233 ovs_net->xt_label = false;
2234 OVS_NLERR(true, "Failed to set connlabel length");
2235 } else {
2236 ovs_net->xt_label = true;
2237 }
11efd5cb
YHW		 2238
2239#if IS_ENABLED(CONFIG_NETFILTER_CONNCOUNT)
2240 return ovs_ct_limit_init(net, ovs_net);
2241#else
2242 return 0;
2243#endif
c2ac6673
JS		 2244}
2245
2246void ovs_ct_exit(struct net *net)
2247{
2248 struct ovs_net *ovs_net = net_generic(net, ovs_net_id);
2249
11efd5cb
YHW		 2250#if IS_ENABLED(CONFIG_NETFILTER_CONNCOUNT)
2251 ovs_ct_limit_exit(net, ovs_net);
2252#endif
2253
c2ac6673
JS		 2254 if (ovs_net->xt_label)
2255 nf_connlabels_put(net);
2256}
Linux 6.x block layer and io_uring tree(s)