projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Merge tag 'x86-apic-2024-03-10' of git://git.kernel.org/pub/scm/linux/kernel/git...
[linux-2.6-block.git] / net / netfilter / nf_flow_table_core.c
CommitLineData
09c434b8 1// SPDX-License-Identifier: GPL-2.0-only
ac2a6666
PNA		 2#include <linux/kernel.h>
3#include <linux/init.h>
4#include <linux/module.h>
5#include <linux/netfilter.h>
6#include <linux/rhashtable.h>
7#include <linux/netdevice.h>
4f3780c0
FF		 8#include <net/ip.h>
9#include <net/ip6_route.h>
c0ea1bcb 10#include <net/netfilter/nf_tables.h>
ac2a6666
PNA		 11#include <net/netfilter/nf_flow_table.h>
12#include <net/netfilter/nf_conntrack.h>
13#include <net/netfilter/nf_conntrack_core.h>
40d102cd 14#include <net/netfilter/nf_conntrack_l4proto.h>
ac2a6666
PNA		 15#include <net/netfilter/nf_conntrack_tuple.h>
16
84453a90
FF		 17static DEFINE_MUTEX(flowtable_lock);
18static LIST_HEAD(flowtables);
19
047b300e 20static void
458a1828 21flow_offload_fill_dir(struct flow_offload *flow,
047b300e
FF		 22 enum flow_offload_tuple_dir dir)
23{
24 struct flow_offload_tuple *ft = &flow->tuplehash[dir].tuple;
458a1828 25 struct nf_conntrack_tuple *ctt = &flow->ct->tuplehash[dir].tuple;
047b300e
FF		 26
27 ft->dir = dir;
28
29 switch (ctt->src.l3num) {
30 case NFPROTO_IPV4:
31 ft->src_v4 = ctt->src.u3.in;
32 ft->dst_v4 = ctt->dst.u3.in;
33 break;
34 case NFPROTO_IPV6:
35 ft->src_v6 = ctt->src.u3.in6;
36 ft->dst_v6 = ctt->dst.u3.in6;
37 break;
38 }
39
40 ft->l3proto = ctt->src.l3num;
41 ft->l4proto = ctt->dst.protonum;
4e8d9584
TM		 42
43 switch (ctt->dst.protonum) {
44 case IPPROTO_TCP:
45 case IPPROTO_UDP:
46 ft->src_port = ctt->src.u.tcp.port;
47 ft->dst_port = ctt->dst.u.tcp.port;
48 break;
49 }
047b300e
FF		 50}
51
f1363e05 52struct flow_offload *flow_offload_alloc(struct nf_conn *ct)
ac2a6666 53{
ac2a6666
PNA		 54 struct flow_offload *flow;
55
f02e7dc4 56 if (unlikely(nf_ct_is_dying(ct)))
ac2a6666
PNA		 57 return NULL;
58
62248df8
PNA		 59 flow = kzalloc(sizeof(*flow), GFP_ATOMIC);
60 if (!flow)
f02e7dc4 61 return NULL;
ac2a6666 62
f02e7dc4 63 refcount_inc(&ct->ct_general.use);
b32d2f34 64 flow->ct = ct;
ac2a6666 65
458a1828 66 flow_offload_fill_dir(flow, FLOW_OFFLOAD_DIR_ORIGINAL);
67 flow_offload_fill_dir(flow, FLOW_OFFLOAD_DIR_REPLY);
ac2a6666
PNA		 68
69 if (ct->status & IPS_SRC_NAT)
355a8b13 70 __set_bit(NF_FLOW_SNAT, &flow->flags);
df1e2025 71 if (ct->status & IPS_DST_NAT)
355a8b13 72 __set_bit(NF_FLOW_DNAT, &flow->flags);
ac2a6666
PNA		 73
74 return flow;
ac2a6666
PNA		 75}
76EXPORT_SYMBOL_GPL(flow_offload_alloc);
77
8b9229d1
PNA		 78static u32 flow_offload_dst_cookie(struct flow_offload_tuple *flow_tuple)
79{
80 const struct rt6_info *rt;
81
82 if (flow_tuple->l3proto == NFPROTO_IPV6) {
83 rt = (const struct rt6_info *)flow_tuple->dst_cache;
84 return rt6_get_cookie(rt);
85 }
86
87 return 0;
88}
89
9e0f0430
PNA		 90static struct dst_entry *nft_route_dst_fetch(struct nf_flow_route *route,
91 enum flow_offload_tuple_dir dir)
92{
93 struct dst_entry *dst = route->tuple[dir].dst;
94
95 route->tuple[dir].dst = NULL;
96
97 return dst;
98}
99
f1363e05 100static int flow_offload_fill_route(struct flow_offload *flow,
9e0f0430 101 struct nf_flow_route *route,
f1363e05
PNA		 102 enum flow_offload_tuple_dir dir)
103{
104 struct flow_offload_tuple *flow_tuple = &flow->tuplehash[dir].tuple;
9e0f0430 105 struct dst_entry *dst = nft_route_dst_fetch(route, dir);
4cd91f7c 106 int i, j = 0;
f1363e05 107
f1363e05
PNA		 108 switch (flow_tuple->l3proto) {
109 case NFPROTO_IPV4:
110 flow_tuple->mtu = ip_dst_mtu_maybe_forward(dst, true);
111 break;
112 case NFPROTO_IPV6:
427faee1 113 flow_tuple->mtu = ip6_dst_mtu_maybe_forward(dst, true);
f1363e05
PNA		 114 break;
115 }
116
c63a7cc4 117 flow_tuple->iifidx = route->tuple[dir].in.ifindex;
4cd91f7c
PNA		 118 for (i = route->tuple[dir].in.num_encaps - 1; i >= 0; i--) {
119 flow_tuple->encap[j].id = route->tuple[dir].in.encap[i].id;
120 flow_tuple->encap[j].proto = route->tuple[dir].in.encap[i].proto;
26267bf9
FF		 121 if (route->tuple[dir].in.ingress_vlans & BIT(i))
122 flow_tuple->in_vlan_ingress |= BIT(j);
4cd91f7c
PNA		 123 j++;
124 }
125 flow_tuple->encap_num = route->tuple[dir].in.num_encaps;
7a27f6ab
PNA		 126
127 switch (route->tuple[dir].xmit_type) {
128 case FLOW_OFFLOAD_XMIT_DIRECT:
129 memcpy(flow_tuple->out.h_dest, route->tuple[dir].out.h_dest,
130 ETH_ALEN);
131 memcpy(flow_tuple->out.h_source, route->tuple[dir].out.h_source,
132 ETH_ALEN);
133 flow_tuple->out.ifidx = route->tuple[dir].out.ifindex;
73f97025 134 flow_tuple->out.hw_ifidx = route->tuple[dir].out.hw_ifindex;
8762785f 135 dst_release(dst);
7a27f6ab
PNA		 136 break;
137 case FLOW_OFFLOAD_XMIT_XFRM:
138 case FLOW_OFFLOAD_XMIT_NEIGH:
7a27f6ab 139 flow_tuple->dst_cache = dst;
8b9229d1 140 flow_tuple->dst_cookie = flow_offload_dst_cookie(flow_tuple);
7a27f6ab 141 break;
78ed0a9b
RD		 142 default:
143 WARN_ON_ONCE(1);
144 break;
7a27f6ab 145 }
5139c0c0 146 flow_tuple->xmit_type = route->tuple[dir].xmit_type;
f1363e05
PNA		 147
148 return 0;
149}
150
7a27f6ab
PNA		 151static void nft_flow_dst_release(struct flow_offload *flow,
152 enum flow_offload_tuple_dir dir)
153{
154 if (flow->tuplehash[dir].tuple.xmit_type == FLOW_OFFLOAD_XMIT_NEIGH ||
155 flow->tuplehash[dir].tuple.xmit_type == FLOW_OFFLOAD_XMIT_XFRM)
156 dst_release(flow->tuplehash[dir].tuple.dst_cache);
157}
158
fa502c86 159void flow_offload_route_init(struct flow_offload *flow,
9e0f0430 160 struct nf_flow_route *route)
f1363e05 161{
fa502c86
PNA		 162 flow_offload_fill_route(flow, route, FLOW_OFFLOAD_DIR_ORIGINAL);
163 flow_offload_fill_route(flow, route, FLOW_OFFLOAD_DIR_REPLY);
f1363e05 164 flow->type = NF_FLOW_OFFLOAD_ROUTE;
f1363e05
PNA		 165}
166EXPORT_SYMBOL_GPL(flow_offload_route_init);
167
da5984e5
FF		 168static void flow_offload_fixup_tcp(struct ip_ct_tcp *tcp)
169{
da5984e5
FF		 170 tcp->seen[0].td_maxwin = 0;
171 tcp->seen[1].td_maxwin = 0;
172}
173
e5eaac2b 174static void flow_offload_fixup_ct(struct nf_conn *ct)
da5984e5 175{
1d91d2e1 176 struct net *net = nf_ct_net(ct);
1e5b2471 177 int l4num = nf_ct_protonum(ct);
4592ee7f 178 s32 timeout;
da5984e5 179
1d91d2e1
OS		 180 if (l4num == IPPROTO_TCP) {
181 struct nf_tcp_net *tn = nf_tcp_pernet(net);
182
e5eaac2b
PNA		 183 flow_offload_fixup_tcp(&ct->proto.tcp);
184
185 timeout = tn->timeouts[ct->proto.tcp.state];
4592ee7f 186 timeout -= tn->offload_timeout;
1d91d2e1
OS		 187 } else if (l4num == IPPROTO_UDP) {
188 struct nf_udp_net *tn = nf_udp_pernet(net);
0eb5acb1
VB		 189 enum udp_conntrack state =
190 test_bit(IPS_SEEN_REPLY_BIT, &ct->status) ?
191 UDP_CT_REPLIED : UDP_CT_UNREPLIED;
1d91d2e1 192
0eb5acb1 193 timeout = tn->timeouts[state];
4592ee7f 194 timeout -= tn->offload_timeout;
1d91d2e1 195 } else {
da5984e5 196 return;
1d91d2e1 197 }
da5984e5 198
4592ee7f
FW		 199 if (timeout < 0)
200 timeout = 0;
201
802a7dc5
ED		 202 if (nf_flow_timeout_delta(READ_ONCE(ct->timeout)) > (__s32)timeout)
203 WRITE_ONCE(ct->timeout, nfct_time_stamp + timeout);
1e5b2471
PNA		 204}
205
f1363e05 206static void flow_offload_route_release(struct flow_offload *flow)
ac2a6666 207{
7a27f6ab
PNA		 208 nft_flow_dst_release(flow, FLOW_OFFLOAD_DIR_ORIGINAL);
209 nft_flow_dst_release(flow, FLOW_OFFLOAD_DIR_REPLY);
f1363e05
PNA		 210}
211
212void flow_offload_free(struct flow_offload *flow)
213{
214 switch (flow->type) {
215 case NF_FLOW_OFFLOAD_ROUTE:
216 flow_offload_route_release(flow);
217 break;
218 default:
219 break;
220 }
b32d2f34 221 nf_ct_put(flow->ct);
62248df8 222 kfree_rcu(flow, rcu_head);
ac2a6666
PNA		 223}
224EXPORT_SYMBOL_GPL(flow_offload_free);
225
a268de77
FF		 226static u32 flow_offload_hash(const void *data, u32 len, u32 seed)
227{
228 const struct flow_offload_tuple *tuple = data;
229
dbc859d9 230 return jhash(tuple, offsetof(struct flow_offload_tuple, __hash), seed);
a268de77
FF		 231}
232
233static u32 flow_offload_hash_obj(const void *data, u32 len, u32 seed)
234{
235 const struct flow_offload_tuple_rhash *tuplehash = data;
236
dbc859d9 237 return jhash(&tuplehash->tuple, offsetof(struct flow_offload_tuple, __hash), seed);
a268de77
FF		 238}
239
240static int flow_offload_hash_cmp(struct rhashtable_compare_arg *arg,
241 const void *ptr)
242{
243 const struct flow_offload_tuple *tuple = arg->key;
244 const struct flow_offload_tuple_rhash *x = ptr;
245
dbc859d9 246 if (memcmp(&x->tuple, tuple, offsetof(struct flow_offload_tuple, __hash)))
a268de77
FF		 247 return 1;
248
249 return 0;
250}
251
252static const struct rhashtable_params nf_flow_offload_rhash_params = {
253 .head_offset = offsetof(struct flow_offload_tuple_rhash, node),
254 .hashfn = flow_offload_hash,
255 .obj_hashfn = flow_offload_hash_obj,
256 .obj_cmpfn = flow_offload_hash_cmp,
257 .automatic_shrinking = true,
258};
259
1d91d2e1
OS		 260unsigned long flow_offload_get_timeout(struct flow_offload *flow)
261{
1d91d2e1
OS		 262 unsigned long timeout = NF_FLOW_TIMEOUT;
263 struct net *net = nf_ct_net(flow->ct);
264 int l4num = nf_ct_protonum(flow->ct);
265
1d91d2e1
OS		 266 if (l4num == IPPROTO_TCP) {
267 struct nf_tcp_net *tn = nf_tcp_pernet(net);
268
269 timeout = tn->offload_timeout;
270 } else if (l4num == IPPROTO_UDP) {
271 struct nf_udp_net *tn = nf_udp_pernet(net);
272
273 timeout = tn->offload_timeout;
274 }
275
276 return timeout;
277}
278
ac2a6666
PNA		 279int flow_offload_add(struct nf_flowtable *flow_table, struct flow_offload *flow)
280{
43c8f131
TY		 281 int err;
282
1d91d2e1 283 flow->timeout = nf_flowtable_time_stamp + flow_offload_get_timeout(flow);
daf61b02 284
43c8f131
TY		 285 err = rhashtable_insert_fast(&flow_table->rhashtable,
286 &flow->tuplehash[0].node,
287 nf_flow_offload_rhash_params);
288 if (err < 0)
289 return err;
290
291 err = rhashtable_insert_fast(&flow_table->rhashtable,
292 &flow->tuplehash[1].node,
293 nf_flow_offload_rhash_params);
294 if (err < 0) {
295 rhashtable_remove_fast(&flow_table->rhashtable,
296 &flow->tuplehash[0].node,
297 nf_flow_offload_rhash_params);
298 return err;
299 }
ac2a6666 300
4203b19c
RD		 301 nf_ct_offload_timeout(flow->ct);
302
f698fe40
PNA		 303 if (nf_flowtable_hw_offload(flow_table)) {
304 __set_bit(NF_FLOW_HW, &flow->flags);
c29f74e0 305 nf_flow_offload_add(flow_table, flow);
f698fe40 306 }
c29f74e0 307
ac2a6666
PNA		 308 return 0;
309}
310EXPORT_SYMBOL_GPL(flow_offload_add);
311
8b3646d6 312void flow_offload_refresh(struct nf_flowtable *flow_table,
41f2c7c3 313 struct flow_offload *flow, bool force)
8b3646d6 314{
32c3973d
PNA		 315 u32 timeout;
316
317 timeout = nf_flowtable_time_stamp + flow_offload_get_timeout(flow);
41f2c7c3 318 if (force || timeout - READ_ONCE(flow->timeout) > HZ)
32c3973d 319 WRITE_ONCE(flow->timeout, timeout);
396ef641
FF		 320 else
321 return;
8b3646d6 322
c07531c0 323 if (likely(!nf_flowtable_hw_offload(flow_table)))
8b3646d6
PB		 324 return;
325
326 nf_flow_offload_add(flow_table, flow);
327}
328EXPORT_SYMBOL_GPL(flow_offload_refresh);
329
3e68db2f
PNA		 330static inline bool nf_flow_has_expired(const struct flow_offload *flow)
331{
1e5b2471 332 return nf_flow_timeout_delta(flow->timeout) <= 0;
3e68db2f
PNA		 333}
334
0ff90b6c
FF		 335static void flow_offload_del(struct nf_flowtable *flow_table,
336 struct flow_offload *flow)
ac2a6666 337{
ac2a6666
PNA		 338 rhashtable_remove_fast(&flow_table->rhashtable,
339 &flow->tuplehash[FLOW_OFFLOAD_DIR_ORIGINAL].node,
a268de77 340 nf_flow_offload_rhash_params);
ac2a6666
PNA		 341 rhashtable_remove_fast(&flow_table->rhashtable,
342 &flow->tuplehash[FLOW_OFFLOAD_DIR_REPLY].node,
a268de77 343 nf_flow_offload_rhash_params);
0ff90b6c 344 flow_offload_free(flow);
ac2a6666 345}
ac2a6666 346
59c466dd
FF		 347void flow_offload_teardown(struct flow_offload *flow)
348{
e5eaac2b 349 clear_bit(IPS_OFFLOAD_BIT, &flow->ct->status);
355a8b13 350 set_bit(NF_FLOW_TEARDOWN, &flow->flags);
e5eaac2b 351 flow_offload_fixup_ct(flow->ct);
59c466dd
FF		 352}
353EXPORT_SYMBOL_GPL(flow_offload_teardown);
354
ac2a6666
PNA		 355struct flow_offload_tuple_rhash *
356flow_offload_lookup(struct nf_flowtable *flow_table,
357 struct flow_offload_tuple *tuple)
358{
ba03137f
FF		 359 struct flow_offload_tuple_rhash *tuplehash;
360 struct flow_offload *flow;
361 int dir;
362
a2d88182
TY		 363 tuplehash = rhashtable_lookup(&flow_table->rhashtable, tuple,
364 nf_flow_offload_rhash_params);
ba03137f
FF		 365 if (!tuplehash)
366 return NULL;
367
368 dir = tuplehash->tuple.dir;
369 flow = container_of(tuplehash, struct flow_offload, tuplehash[dir]);
355a8b13 370 if (test_bit(NF_FLOW_TEARDOWN, &flow->flags))
ba03137f
FF		 371 return NULL;
372
b32d2f34 373 if (unlikely(nf_ct_is_dying(flow->ct)))
8cd2bc98
TY		 374 return NULL;
375
ba03137f 376 return tuplehash;
ac2a6666
PNA		 377}
378EXPORT_SYMBOL_GPL(flow_offload_lookup);
379
49de9c09
TY		 380static int
381nf_flow_table_iterate(struct nf_flowtable *flow_table,
217cff36
PNA		 382 void (*iter)(struct nf_flowtable *flowtable,
383 struct flow_offload *flow, void *data),
49de9c09 384 void *data)
ac2a6666
PNA		 385{
386 struct flow_offload_tuple_rhash *tuplehash;
387 struct rhashtable_iter hti;
388 struct flow_offload *flow;
0de22baa 389 int err = 0;
ac2a6666 390
0de22baa 391 rhashtable_walk_enter(&flow_table->rhashtable, &hti);
ac2a6666
PNA		 392 rhashtable_walk_start(&hti);
393
394 while ((tuplehash = rhashtable_walk_next(&hti))) {
395 if (IS_ERR(tuplehash)) {
0de22baa
TY		 396 if (PTR_ERR(tuplehash) != -EAGAIN) {
397 err = PTR_ERR(tuplehash);
398 break;
399 }
ac2a6666
PNA		 400 continue;
401 }
402 if (tuplehash->tuple.dir)
403 continue;
404
405 flow = container_of(tuplehash, struct flow_offload, tuplehash[0]);
406
217cff36 407 iter(flow_table, flow, data);
ac2a6666 408 }
ac2a6666
PNA		 409 rhashtable_walk_stop(&hti);
410 rhashtable_walk_exit(&hti);
411
412 return err;
413}
ac2a6666 414
735795f6
PNA		 415static bool nf_flow_custom_gc(struct nf_flowtable *flow_table,
416 const struct flow_offload *flow)
417{
418 return flow_table->type->gc && flow_table->type->gc(flow);
419}
420
217cff36
PNA		 421static void nf_flow_offload_gc_step(struct nf_flowtable *flow_table,
422 struct flow_offload *flow, void *data)
ac2a6666 423{
8b9229d1 424 if (nf_flow_has_expired(flow) ||
41f2c7c3 425 nf_ct_is_dying(flow->ct) ||
735795f6 426 nf_flow_custom_gc(flow_table, flow))
e5eaac2b 427 flow_offload_teardown(flow);
9ed81c8e
PNA		 428
429 if (test_bit(NF_FLOW_TEARDOWN, &flow->flags)) {
355a8b13
PNA		 430 if (test_bit(NF_FLOW_HW, &flow->flags)) {
431 if (!test_bit(NF_FLOW_HW_DYING, &flow->flags))
c29f74e0 432 nf_flow_offload_del(flow_table, flow);
355a8b13 433 else if (test_bit(NF_FLOW_HW_DEAD, &flow->flags))
c29f74e0
PNA		 434 flow_offload_del(flow_table, flow);
435 } else {
436 flow_offload_del(flow_table, flow);
437 }
355a8b13 438 } else if (test_bit(NF_FLOW_HW, &flow->flags)) {
79b9b685 439 nf_flow_offload_stats(flow_table, flow);
c29f74e0 440 }
b408c5b0
PNA		 441}
442
759eebbc
PNA		 443void nf_flow_table_gc_run(struct nf_flowtable *flow_table)
444{
445 nf_flow_table_iterate(flow_table, nf_flow_offload_gc_step, NULL);
446}
447
a268de77 448static void nf_flow_offload_work_gc(struct work_struct *work)
b408c5b0
PNA		 449{
450 struct nf_flowtable *flow_table;
451
452 flow_table = container_of(work, struct nf_flowtable, gc_work.work);
759eebbc 453 nf_flow_table_gc_run(flow_table);
ac2a6666
PNA		 454 queue_delayed_work(system_power_efficient_wq, &flow_table->gc_work, HZ);
455}
ac2a6666 456
f4401262
PNA		 457static void nf_flow_nat_port_tcp(struct sk_buff *skb, unsigned int thoff,
458 __be16 port, __be16 new_port)
ac2a6666
PNA		 459{
460 struct tcphdr *tcph;
461
ac2a6666 462 tcph = (void *)(skb_network_header(skb) + thoff);
8d6bca15 463 inet_proto_csum_replace2(&tcph->check, skb, port, new_port, false);
ac2a6666
PNA		 464}
465
f4401262
PNA		 466static void nf_flow_nat_port_udp(struct sk_buff *skb, unsigned int thoff,
467 __be16 port, __be16 new_port)
ac2a6666
PNA		 468{
469 struct udphdr *udph;
470
ac2a6666
PNA		 471 udph = (void *)(skb_network_header(skb) + thoff);
472 if (udph->check || skb->ip_summed == CHECKSUM_PARTIAL) {
473 inet_proto_csum_replace2(&udph->check, skb, port,
8d6bca15 474 new_port, false);
ac2a6666
PNA		 475 if (!udph->check)
476 udph->check = CSUM_MANGLED_0;
477 }
ac2a6666
PNA		 478}
479
f4401262
PNA		 480static void nf_flow_nat_port(struct sk_buff *skb, unsigned int thoff,
481 u8 protocol, __be16 port, __be16 new_port)
ac2a6666
PNA		 482{
483 switch (protocol) {
484 case IPPROTO_TCP:
f4401262 485 nf_flow_nat_port_tcp(skb, thoff, port, new_port);
ac2a6666
PNA		 486 break;
487 case IPPROTO_UDP:
f4401262 488 nf_flow_nat_port_udp(skb, thoff, port, new_port);
ac2a6666
PNA		 489 break;
490 }
ac2a6666
PNA		 491}
492
f4401262
PNA		 493void nf_flow_snat_port(const struct flow_offload *flow,
494 struct sk_buff *skb, unsigned int thoff,
495 u8 protocol, enum flow_offload_tuple_dir dir)
ac2a6666
PNA		 496{
497 struct flow_ports *hdr;
498 __be16 port, new_port;
499
ac2a6666
PNA		 500 hdr = (void *)(skb_network_header(skb) + thoff);
501
502 switch (dir) {
503 case FLOW_OFFLOAD_DIR_ORIGINAL:
504 port = hdr->source;
505 new_port = flow->tuplehash[FLOW_OFFLOAD_DIR_REPLY].tuple.dst_port;
506 hdr->source = new_port;
507 break;
508 case FLOW_OFFLOAD_DIR_REPLY:
509 port = hdr->dest;
510 new_port = flow->tuplehash[FLOW_OFFLOAD_DIR_ORIGINAL].tuple.src_port;
511 hdr->dest = new_port;
512 break;
ac2a6666
PNA		 513 }
514
f4401262 515 nf_flow_nat_port(skb, thoff, protocol, port, new_port);
ac2a6666
PNA		 516}
517EXPORT_SYMBOL_GPL(nf_flow_snat_port);
518
f4401262
PNA		 519void nf_flow_dnat_port(const struct flow_offload *flow, struct sk_buff *skb,
520 unsigned int thoff, u8 protocol,
521 enum flow_offload_tuple_dir dir)
ac2a6666
PNA		 522{
523 struct flow_ports *hdr;
524 __be16 port, new_port;
525
ac2a6666
PNA		 526 hdr = (void *)(skb_network_header(skb) + thoff);
527
528 switch (dir) {
529 case FLOW_OFFLOAD_DIR_ORIGINAL:
530 port = hdr->dest;
531 new_port = flow->tuplehash[FLOW_OFFLOAD_DIR_REPLY].tuple.src_port;
532 hdr->dest = new_port;
533 break;
534 case FLOW_OFFLOAD_DIR_REPLY:
535 port = hdr->source;
536 new_port = flow->tuplehash[FLOW_OFFLOAD_DIR_ORIGINAL].tuple.dst_port;
537 hdr->source = new_port;
538 break;
ac2a6666
PNA		 539 }
540
f4401262 541 nf_flow_nat_port(skb, thoff, protocol, port, new_port);
ac2a6666
PNA		 542}
543EXPORT_SYMBOL_GPL(nf_flow_dnat_port);
544
a268de77
FF		 545int nf_flow_table_init(struct nf_flowtable *flowtable)
546{
547 int err;
548
740b486a 549 INIT_DELAYED_WORK(&flowtable->gc_work, nf_flow_offload_work_gc);
c29f74e0 550 flow_block_init(&flowtable->flow_block);
422c032a 551 init_rwsem(&flowtable->flow_block_lock);
a268de77
FF		 552
553 err = rhashtable_init(&flowtable->rhashtable,
554 &nf_flow_offload_rhash_params);
555 if (err < 0)
556 return err;
557
558 queue_delayed_work(system_power_efficient_wq,
559 &flowtable->gc_work, HZ);
560
84453a90
FF		 561 mutex_lock(&flowtable_lock);
562 list_add(&flowtable->list, &flowtables);
563 mutex_unlock(&flowtable_lock);
564
a268de77
FF		 565 return 0;
566}
567EXPORT_SYMBOL_GPL(nf_flow_table_init);
568
217cff36
PNA		 569static void nf_flow_table_do_cleanup(struct nf_flowtable *flow_table,
570 struct flow_offload *flow, void *data)
c0ea1bcb
PNA		 571{
572 struct net_device *dev = data;
573
59c466dd
FF		 574 if (!dev) {
575 flow_offload_teardown(flow);
c0ea1bcb 576 return;
59c466dd 577 }
b32d2f34
PNA		 578
579 if (net_eq(nf_ct_net(flow->ct), dev_net(dev)) &&
a3fb3698
TY		 580 (flow->tuplehash[0].tuple.iifidx == dev->ifindex ||
581 flow->tuplehash[1].tuple.iifidx == dev->ifindex))
445db8d0 582 flow_offload_teardown(flow);
c0ea1bcb
PNA		 583}
584
a8284c68
PNA		 585void nf_flow_table_gc_cleanup(struct nf_flowtable *flowtable,
586 struct net_device *dev)
c0ea1bcb 587{
84453a90 588 nf_flow_table_iterate(flowtable, nf_flow_table_do_cleanup, dev);
c0ea1bcb 589 flush_delayed_work(&flowtable->gc_work);
91bfaa15 590 nf_flow_table_offload_flush(flowtable);
c0ea1bcb
PNA		 591}
592
5f1be84a 593void nf_flow_table_cleanup(struct net_device *dev)
c0ea1bcb 594{
84453a90
FF		 595 struct nf_flowtable *flowtable;
596
597 mutex_lock(&flowtable_lock);
598 list_for_each_entry(flowtable, &flowtables, list)
a8284c68 599 nf_flow_table_gc_cleanup(flowtable, dev);
84453a90 600 mutex_unlock(&flowtable_lock);
c0ea1bcb
PNA		 601}
602EXPORT_SYMBOL_GPL(nf_flow_table_cleanup);
603
b408c5b0
PNA		 604void nf_flow_table_free(struct nf_flowtable *flow_table)
605{
84453a90
FF		 606 mutex_lock(&flowtable_lock);
607 list_del(&flow_table->list);
608 mutex_unlock(&flowtable_lock);
978703f4 609
a268de77 610 cancel_delayed_work_sync(&flow_table->gc_work);
9afb4b27
PNA		 611 nf_flow_table_offload_flush(flow_table);
612 /* ... no more pending work after this stage ... */
b408c5b0 613 nf_flow_table_iterate(flow_table, nf_flow_table_do_cleanup, NULL);
759eebbc 614 nf_flow_table_gc_run(flow_table);
9afb4b27 615 nf_flow_table_offload_flush_cleanup(flow_table);
a268de77 616 rhashtable_destroy(&flow_table->rhashtable);
b408c5b0
PNA		 617}
618EXPORT_SYMBOL_GPL(nf_flow_table_free);
619
b0381776
VB		 620static int nf_flow_table_init_net(struct net *net)
621{
622 net->ft.stat = alloc_percpu(struct nf_flow_table_stat);
623 return net->ft.stat ? 0 : -ENOMEM;
624}
625
626static void nf_flow_table_fini_net(struct net *net)
627{
628 free_percpu(net->ft.stat);
629}
630
631static int nf_flow_table_pernet_init(struct net *net)
632{
633 int ret;
634
635 ret = nf_flow_table_init_net(net);
636 if (ret < 0)
637 return ret;
638
639 ret = nf_flow_table_init_proc(net);
640 if (ret < 0)
641 goto out_proc;
642
643 return 0;
644
645out_proc:
646 nf_flow_table_fini_net(net);
647 return ret;
648}
649
650static void nf_flow_table_pernet_exit(struct list_head *net_exit_list)
651{
652 struct net *net;
653
654 list_for_each_entry(net, net_exit_list, exit_list) {
655 nf_flow_table_fini_proc(net);
656 nf_flow_table_fini_net(net);
657 }
658}
659
660static struct pernet_operations nf_flow_table_net_ops = {
661 .init = nf_flow_table_pernet_init,
662 .exit_batch = nf_flow_table_pernet_exit,
663};
664
c29f74e0
PNA		 665static int __init nf_flow_table_module_init(void)
666{
b0381776
VB		 667 int ret;
668
669 ret = register_pernet_subsys(&nf_flow_table_net_ops);
670 if (ret < 0)
671 return ret;
672
673 ret = nf_flow_table_offload_init();
674 if (ret)
675 goto out_offload;
676
677 return 0;
678
679out_offload:
680 unregister_pernet_subsys(&nf_flow_table_net_ops);
681 return ret;
c29f74e0
PNA		 682}
683
684static void __exit nf_flow_table_module_exit(void)
685{
686 nf_flow_table_offload_exit();
b0381776 687 unregister_pernet_subsys(&nf_flow_table_net_ops);
c29f74e0
PNA		 688}
689
690module_init(nf_flow_table_module_init);
691module_exit(nf_flow_table_module_exit);
692
ac2a6666
PNA		 693MODULE_LICENSE("GPL");
694MODULE_AUTHOR("Pablo Neira Ayuso <pablo@netfilter.org>");
4cacc395 695MODULE_DESCRIPTION("Netfilter flow table module");
Linux 6.x block layer and io_uring tree(s)