1 // SPDX-License-Identifier: GPL-2.0-only
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/ip.h>
8 #include <linux/ipv6.h>
9 #include <linux/netdevice.h>
10 #include <linux/if_ether.h>
11 #include <net/ip.h>
12 #include <net/ipv6.h>
13 #include <net/ip6_route.h>
14 #include <net/neighbour.h>
15 #include <net/netfilter/nf_flow_table.h>
16 #include <net/netfilter/nf_conntrack_acct.h>
17 /* For layer 4 checksum field offset. */
18 #include <linux/tcp.h>
19 #include <linux/udp.h>
20 
nf_flow_state_check(struct flow_offload * flow,int proto,struct sk_buff * skb,unsigned int thoff)21 static int nf_flow_state_check(struct flow_offload *flow, int proto,
22 			       struct sk_buff *skb, unsigned int thoff)
23 {
24 	struct tcphdr *tcph;
25 
26 	if (proto != IPPROTO_TCP)
27 		return 0;
28 
29 	tcph = (void *)(skb_network_header(skb) + thoff);
30 	if (unlikely(tcph->fin || tcph->rst)) {
31 		flow_offload_teardown(flow);
32 		return -1;
33 	}
34 
35 	return 0;
36 }
37 
nf_flow_nat_ip_tcp(struct sk_buff * skb,unsigned int thoff,__be32 addr,__be32 new_addr)38 static void nf_flow_nat_ip_tcp(struct sk_buff *skb, unsigned int thoff,
39 			       __be32 addr, __be32 new_addr)
40 {
41 	struct tcphdr *tcph;
42 
43 	tcph = (void *)(skb_network_header(skb) + thoff);
44 	inet_proto_csum_replace4(&tcph->check, skb, addr, new_addr, true);
45 }
46 
nf_flow_nat_ip_udp(struct sk_buff * skb,unsigned int thoff,__be32 addr,__be32 new_addr)47 static void nf_flow_nat_ip_udp(struct sk_buff *skb, unsigned int thoff,
48 			       __be32 addr, __be32 new_addr)
49 {
50 	struct udphdr *udph;
51 
52 	udph = (void *)(skb_network_header(skb) + thoff);
53 	if (udph->check || skb->ip_summed == CHECKSUM_PARTIAL) {
54 		inet_proto_csum_replace4(&udph->check, skb, addr,
55 					 new_addr, true);
56 		if (!udph->check)
57 			udph->check = CSUM_MANGLED_0;
58 	}
59 }
60 
nf_flow_nat_ip_l4proto(struct sk_buff * skb,struct iphdr * iph,unsigned int thoff,__be32 addr,__be32 new_addr)61 static void nf_flow_nat_ip_l4proto(struct sk_buff *skb, struct iphdr *iph,
62 				   unsigned int thoff, __be32 addr,
63 				   __be32 new_addr)
64 {
65 	switch (iph->protocol) {
66 	case IPPROTO_TCP:
67 		nf_flow_nat_ip_tcp(skb, thoff, addr, new_addr);
68 		break;
69 	case IPPROTO_UDP:
70 		nf_flow_nat_ip_udp(skb, thoff, addr, new_addr);
71 		break;
72 	}
73 }
74 
nf_flow_snat_ip(const struct flow_offload * flow,struct sk_buff * skb,struct iphdr * iph,unsigned int thoff,enum flow_offload_tuple_dir dir)75 static void nf_flow_snat_ip(const struct flow_offload *flow,
76 			    struct sk_buff *skb, struct iphdr *iph,
77 			    unsigned int thoff, enum flow_offload_tuple_dir dir)
78 {
79 	__be32 addr, new_addr;
80 
81 	switch (dir) {
82 	case FLOW_OFFLOAD_DIR_ORIGINAL:
83 		addr = iph->saddr;
84 		new_addr = flow->tuplehash[FLOW_OFFLOAD_DIR_REPLY].tuple.dst_v4.s_addr;
85 		iph->saddr = new_addr;
86 		break;
87 	case FLOW_OFFLOAD_DIR_REPLY:
88 		addr = iph->daddr;
89 		new_addr = flow->tuplehash[FLOW_OFFLOAD_DIR_ORIGINAL].tuple.src_v4.s_addr;
90 		iph->daddr = new_addr;
91 		break;
92 	}
93 	csum_replace4(&iph->check, addr, new_addr);
94 
95 	nf_flow_nat_ip_l4proto(skb, iph, thoff, addr, new_addr);
96 }
97 
nf_flow_dnat_ip(const struct flow_offload * flow,struct sk_buff * skb,struct iphdr * iph,unsigned int thoff,enum flow_offload_tuple_dir dir)98 static void nf_flow_dnat_ip(const struct flow_offload *flow,
99 			    struct sk_buff *skb, struct iphdr *iph,
100 			    unsigned int thoff, enum flow_offload_tuple_dir dir)
101 {
102 	__be32 addr, new_addr;
103 
104 	switch (dir) {
105 	case FLOW_OFFLOAD_DIR_ORIGINAL:
106 		addr = iph->daddr;
107 		new_addr = flow->tuplehash[FLOW_OFFLOAD_DIR_REPLY].tuple.src_v4.s_addr;
108 		iph->daddr = new_addr;
109 		break;
110 	case FLOW_OFFLOAD_DIR_REPLY:
111 		addr = iph->saddr;
112 		new_addr = flow->tuplehash[FLOW_OFFLOAD_DIR_ORIGINAL].tuple.dst_v4.s_addr;
113 		iph->saddr = new_addr;
114 		break;
115 	}
116 	csum_replace4(&iph->check, addr, new_addr);
117 
118 	nf_flow_nat_ip_l4proto(skb, iph, thoff, addr, new_addr);
119 }
120 
nf_flow_nat_ip(const struct flow_offload * flow,struct sk_buff * skb,unsigned int thoff,enum flow_offload_tuple_dir dir,struct iphdr * iph)121 static void nf_flow_nat_ip(const struct flow_offload *flow, struct sk_buff *skb,
122 			  unsigned int thoff, enum flow_offload_tuple_dir dir,
123 			  struct iphdr *iph)
124 {
125 	if (test_bit(NF_FLOW_SNAT, &flow->flags)) {
126 		nf_flow_snat_port(flow, skb, thoff, iph->protocol, dir);
127 		nf_flow_snat_ip(flow, skb, iph, thoff, dir);
128 	}
129 	if (test_bit(NF_FLOW_DNAT, &flow->flags)) {
130 		nf_flow_dnat_port(flow, skb, thoff, iph->protocol, dir);
131 		nf_flow_dnat_ip(flow, skb, iph, thoff, dir);
132 	}
133 }
134 
ip_has_options(unsigned int thoff)135 static bool ip_has_options(unsigned int thoff)
136 {
137 	return thoff != sizeof(struct iphdr);
138 }
139 
nf_flow_tuple_encap(struct sk_buff * skb,struct flow_offload_tuple * tuple)140 static void nf_flow_tuple_encap(struct sk_buff *skb,
141 				struct flow_offload_tuple *tuple)
142 {
143 	struct vlan_ethhdr *veth;
144 	struct pppoe_hdr *phdr;
145 	int i = 0;
146 
147 	if (skb_vlan_tag_present(skb)) {
148 		tuple->encap[i].id = skb_vlan_tag_get(skb);
149 		tuple->encap[i].proto = skb->vlan_proto;
150 		i++;
151 	}
152 	switch (skb->protocol) {
153 	case htons(ETH_P_8021Q):
154 		veth = (struct vlan_ethhdr *)skb_mac_header(skb);
155 		tuple->encap[i].id = ntohs(veth->h_vlan_TCI);
156 		tuple->encap[i].proto = skb->protocol;
157 		break;
158 	case htons(ETH_P_PPP_SES):
159 		phdr = (struct pppoe_hdr *)skb_mac_header(skb);
160 		tuple->encap[i].id = ntohs(phdr->sid);
161 		tuple->encap[i].proto = skb->protocol;
162 		break;
163 	}
164 }
165 
nf_flow_tuple_ip(struct sk_buff * skb,const struct net_device * dev,struct flow_offload_tuple * tuple,u32 * hdrsize,u32 offset)166 static int nf_flow_tuple_ip(struct sk_buff *skb, const struct net_device *dev,
167 			    struct flow_offload_tuple *tuple, u32 *hdrsize,
168 			    u32 offset)
169 {
170 	struct flow_ports *ports;
171 	unsigned int thoff;
172 	struct iphdr *iph;
173 	u8 ipproto;
174 
175 	if (!pskb_may_pull(skb, sizeof(*iph) + offset))
176 		return -1;
177 
178 	iph = (struct iphdr *)(skb_network_header(skb) + offset);
179 	thoff = (iph->ihl * 4);
180 
181 	if (ip_is_fragment(iph) ||
182 	    unlikely(ip_has_options(thoff)))
183 		return -1;
184 
185 	thoff += offset;
186 
187 	ipproto = iph->protocol;
188 	switch (ipproto) {
189 	case IPPROTO_TCP:
190 		*hdrsize = sizeof(struct tcphdr);
191 		break;
192 	case IPPROTO_UDP:
193 		*hdrsize = sizeof(struct udphdr);
194 		break;
195 #ifdef CONFIG_NF_CT_PROTO_GRE
196 	case IPPROTO_GRE:
197 		*hdrsize = sizeof(struct gre_base_hdr);
198 		break;
199 #endif
200 	default:
201 		return -1;
202 	}
203 
204 	if (iph->ttl <= 1)
205 		return -1;
206 
207 	if (!pskb_may_pull(skb, thoff + *hdrsize))
208 		return -1;
209 
210 	switch (ipproto) {
211 	case IPPROTO_TCP:
212 	case IPPROTO_UDP:
213 		ports = (struct flow_ports *)(skb_network_header(skb) + thoff);
214 		tuple->src_port		= ports->source;
215 		tuple->dst_port		= ports->dest;
216 		break;
217 	case IPPROTO_GRE: {
218 		struct gre_base_hdr *greh;
219 
220 		greh = (struct gre_base_hdr *)(skb_network_header(skb) + thoff);
221 		if ((greh->flags & GRE_VERSION) != GRE_VERSION_0)
222 			return -1;
223 		break;
224 	}
225 	}
226 
227 	iph = (struct iphdr *)(skb_network_header(skb) + offset);
228 
229 	tuple->src_v4.s_addr	= iph->saddr;
230 	tuple->dst_v4.s_addr	= iph->daddr;
231 	tuple->l3proto		= AF_INET;
232 	tuple->l4proto		= ipproto;
233 	tuple->iifidx		= dev->ifindex;
234 	nf_flow_tuple_encap(skb, tuple);
235 
236 	return 0;
237 }
238 
239 /* Based on ip_exceeds_mtu(). */
nf_flow_exceeds_mtu(const struct sk_buff * skb,unsigned int mtu)240 static bool nf_flow_exceeds_mtu(const struct sk_buff *skb, unsigned int mtu)
241 {
242 	if (skb->len <= mtu)
243 		return false;
244 
245 	if (skb_is_gso(skb) && skb_gso_validate_network_len(skb, mtu))
246 		return false;
247 
248 	return true;
249 }
250 
nf_flow_dst_check(struct flow_offload_tuple * tuple)251 static inline bool nf_flow_dst_check(struct flow_offload_tuple *tuple)
252 {
253 	if (tuple->xmit_type != FLOW_OFFLOAD_XMIT_NEIGH &&
254 	    tuple->xmit_type != FLOW_OFFLOAD_XMIT_XFRM)
255 		return true;
256 
257 	return dst_check(tuple->dst_cache, tuple->dst_cookie);
258 }
259 
nf_flow_xmit_xfrm(struct sk_buff * skb,const struct nf_hook_state * state,struct dst_entry * dst)260 static unsigned int nf_flow_xmit_xfrm(struct sk_buff *skb,
261 				      const struct nf_hook_state *state,
262 				      struct dst_entry *dst)
263 {
264 	skb_orphan(skb);
265 	skb_dst_set_noref(skb, dst);
266 	dst_output(state->net, state->sk, skb);
267 	return NF_STOLEN;
268 }
269 
nf_flow_skb_encap_protocol(const struct sk_buff * skb,__be16 proto,u32 * offset)270 static bool nf_flow_skb_encap_protocol(const struct sk_buff *skb, __be16 proto,
271 				       u32 *offset)
272 {
273 	struct vlan_ethhdr *veth;
274 
275 	switch (skb->protocol) {
276 	case htons(ETH_P_8021Q):
277 		veth = (struct vlan_ethhdr *)skb_mac_header(skb);
278 		if (veth->h_vlan_encapsulated_proto == proto) {
279 			*offset += VLAN_HLEN;
280 			return true;
281 		}
282 		break;
283 	case htons(ETH_P_PPP_SES):
284 		if (nf_flow_pppoe_proto(skb) == proto) {
285 			*offset += PPPOE_SES_HLEN;
286 			return true;
287 		}
288 		break;
289 	}
290 
291 	return false;
292 }
293 
nf_flow_encap_pop(struct sk_buff * skb,struct flow_offload_tuple_rhash * tuplehash)294 static void nf_flow_encap_pop(struct sk_buff *skb,
295 			      struct flow_offload_tuple_rhash *tuplehash)
296 {
297 	struct vlan_hdr *vlan_hdr;
298 	int i;
299 
300 	for (i = 0; i < tuplehash->tuple.encap_num; i++) {
301 		if (skb_vlan_tag_present(skb)) {
302 			__vlan_hwaccel_clear_tag(skb);
303 			continue;
304 		}
305 		switch (skb->protocol) {
306 		case htons(ETH_P_8021Q):
307 			vlan_hdr = (struct vlan_hdr *)skb->data;
308 			__skb_pull(skb, VLAN_HLEN);
309 			vlan_set_encap_proto(skb, vlan_hdr);
310 			skb_reset_network_header(skb);
311 			break;
312 		case htons(ETH_P_PPP_SES):
313 			skb->protocol = nf_flow_pppoe_proto(skb);
314 			skb_pull(skb, PPPOE_SES_HLEN);
315 			skb_reset_network_header(skb);
316 			break;
317 		}
318 	}
319 }
320 
nf_flow_queue_xmit(struct net * net,struct sk_buff * skb,const struct flow_offload_tuple_rhash * tuplehash,unsigned short type)321 static unsigned int nf_flow_queue_xmit(struct net *net, struct sk_buff *skb,
322 				       const struct flow_offload_tuple_rhash *tuplehash,
323 				       unsigned short type)
324 {
325 	struct net_device *outdev;
326 
327 	outdev = dev_get_by_index_rcu(net, tuplehash->tuple.out.ifidx);
328 	if (!outdev)
329 		return NF_DROP;
330 
331 	skb->dev = outdev;
332 	dev_hard_header(skb, skb->dev, type, tuplehash->tuple.out.h_dest,
333 			tuplehash->tuple.out.h_source, skb->len);
334 	dev_queue_xmit(skb);
335 
336 	return NF_STOLEN;
337 }
338 
339 unsigned int
nf_flow_offload_ip_hook(void * priv,struct sk_buff * skb,const struct nf_hook_state * state)340 nf_flow_offload_ip_hook(void *priv, struct sk_buff *skb,
341 			const struct nf_hook_state *state)
342 {
343 	struct flow_offload_tuple_rhash *tuplehash;
344 	struct nf_flowtable *flow_table = priv;
345 	struct flow_offload_tuple tuple = {};
346 	enum flow_offload_tuple_dir dir;
347 	struct flow_offload *flow;
348 	struct net_device *outdev;
349 	u32 hdrsize, offset = 0;
350 	unsigned int thoff, mtu;
351 	struct rtable *rt;
352 	struct iphdr *iph;
353 	__be32 nexthop;
354 	int ret;
355 
356 	if (skb->protocol != htons(ETH_P_IP) &&
357 	    !nf_flow_skb_encap_protocol(skb, htons(ETH_P_IP), &offset))
358 		return NF_ACCEPT;
359 
360 	if (nf_flow_tuple_ip(skb, state->in, &tuple, &hdrsize, offset) < 0)
361 		return NF_ACCEPT;
362 
363 	tuplehash = flow_offload_lookup(flow_table, &tuple);
364 	if (tuplehash == NULL)
365 		return NF_ACCEPT;
366 
367 	dir = tuplehash->tuple.dir;
368 	flow = container_of(tuplehash, struct flow_offload, tuplehash[dir]);
369 
370 	mtu = flow->tuplehash[dir].tuple.mtu + offset;
371 	if (unlikely(nf_flow_exceeds_mtu(skb, mtu)))
372 		return NF_ACCEPT;
373 
374 	iph = (struct iphdr *)(skb_network_header(skb) + offset);
375 	thoff = (iph->ihl * 4) + offset;
376 	if (nf_flow_state_check(flow, iph->protocol, skb, thoff))
377 		return NF_ACCEPT;
378 
379 	if (!nf_flow_dst_check(&tuplehash->tuple)) {
380 		flow_offload_teardown(flow);
381 		return NF_ACCEPT;
382 	}
383 
384 	if (skb_try_make_writable(skb, thoff + hdrsize))
385 		return NF_DROP;
386 
387 	flow_offload_refresh(flow_table, flow);
388 
389 	nf_flow_encap_pop(skb, tuplehash);
390 	thoff -= offset;
391 
392 	iph = ip_hdr(skb);
393 	nf_flow_nat_ip(flow, skb, thoff, dir, iph);
394 
395 	ip_decrease_ttl(iph);
396 	skb_clear_tstamp(skb);
397 
398 	if (flow_table->flags & NF_FLOWTABLE_COUNTER)
399 		nf_ct_acct_update(flow->ct, tuplehash->tuple.dir, skb->len);
400 
401 	if (unlikely(tuplehash->tuple.xmit_type == FLOW_OFFLOAD_XMIT_XFRM)) {
402 		rt = (struct rtable *)tuplehash->tuple.dst_cache;
403 		memset(skb->cb, 0, sizeof(struct inet_skb_parm));
404 		IPCB(skb)->iif = skb->dev->ifindex;
405 		IPCB(skb)->flags = IPSKB_FORWARDED;
406 		return nf_flow_xmit_xfrm(skb, state, &rt->dst);
407 	}
408 
409 	switch (tuplehash->tuple.xmit_type) {
410 	case FLOW_OFFLOAD_XMIT_NEIGH:
411 		rt = (struct rtable *)tuplehash->tuple.dst_cache;
412 		outdev = rt->dst.dev;
413 		skb->dev = outdev;
414 		nexthop = rt_nexthop(rt, flow->tuplehash[!dir].tuple.src_v4.s_addr);
415 		skb_dst_set_noref(skb, &rt->dst);
416 		neigh_xmit(NEIGH_ARP_TABLE, outdev, &nexthop, skb);
417 		ret = NF_STOLEN;
418 		break;
419 	case FLOW_OFFLOAD_XMIT_DIRECT:
420 		ret = nf_flow_queue_xmit(state->net, skb, tuplehash, ETH_P_IP);
421 		if (ret == NF_DROP)
422 			flow_offload_teardown(flow);
423 		break;
424 	}
425 
426 	return ret;
427 }
428 EXPORT_SYMBOL_GPL(nf_flow_offload_ip_hook);
429 
nf_flow_nat_ipv6_tcp(struct sk_buff * skb,unsigned int thoff,struct in6_addr * addr,struct in6_addr * new_addr,struct ipv6hdr * ip6h)430 static void nf_flow_nat_ipv6_tcp(struct sk_buff *skb, unsigned int thoff,
431 				 struct in6_addr *addr,
432 				 struct in6_addr *new_addr,
433 				 struct ipv6hdr *ip6h)
434 {
435 	struct tcphdr *tcph;
436 
437 	tcph = (void *)(skb_network_header(skb) + thoff);
438 	inet_proto_csum_replace16(&tcph->check, skb, addr->s6_addr32,
439 				  new_addr->s6_addr32, true);
440 }
441 
nf_flow_nat_ipv6_udp(struct sk_buff * skb,unsigned int thoff,struct in6_addr * addr,struct in6_addr * new_addr)442 static void nf_flow_nat_ipv6_udp(struct sk_buff *skb, unsigned int thoff,
443 				 struct in6_addr *addr,
444 				 struct in6_addr *new_addr)
445 {
446 	struct udphdr *udph;
447 
448 	udph = (void *)(skb_network_header(skb) + thoff);
449 	if (udph->check || skb->ip_summed == CHECKSUM_PARTIAL) {
450 		inet_proto_csum_replace16(&udph->check, skb, addr->s6_addr32,
451 					  new_addr->s6_addr32, true);
452 		if (!udph->check)
453 			udph->check = CSUM_MANGLED_0;
454 	}
455 }
456 
nf_flow_nat_ipv6_l4proto(struct sk_buff * skb,struct ipv6hdr * ip6h,unsigned int thoff,struct in6_addr * addr,struct in6_addr * new_addr)457 static void nf_flow_nat_ipv6_l4proto(struct sk_buff *skb, struct ipv6hdr *ip6h,
458 				     unsigned int thoff, struct in6_addr *addr,
459 				     struct in6_addr *new_addr)
460 {
461 	switch (ip6h->nexthdr) {
462 	case IPPROTO_TCP:
463 		nf_flow_nat_ipv6_tcp(skb, thoff, addr, new_addr, ip6h);
464 		break;
465 	case IPPROTO_UDP:
466 		nf_flow_nat_ipv6_udp(skb, thoff, addr, new_addr);
467 		break;
468 	}
469 }
470 
nf_flow_snat_ipv6(const struct flow_offload * flow,struct sk_buff * skb,struct ipv6hdr * ip6h,unsigned int thoff,enum flow_offload_tuple_dir dir)471 static void nf_flow_snat_ipv6(const struct flow_offload *flow,
472 			      struct sk_buff *skb, struct ipv6hdr *ip6h,
473 			      unsigned int thoff,
474 			      enum flow_offload_tuple_dir dir)
475 {
476 	struct in6_addr addr, new_addr;
477 
478 	switch (dir) {
479 	case FLOW_OFFLOAD_DIR_ORIGINAL:
480 		addr = ip6h->saddr;
481 		new_addr = flow->tuplehash[FLOW_OFFLOAD_DIR_REPLY].tuple.dst_v6;
482 		ip6h->saddr = new_addr;
483 		break;
484 	case FLOW_OFFLOAD_DIR_REPLY:
485 		addr = ip6h->daddr;
486 		new_addr = flow->tuplehash[FLOW_OFFLOAD_DIR_ORIGINAL].tuple.src_v6;
487 		ip6h->daddr = new_addr;
488 		break;
489 	}
490 
491 	nf_flow_nat_ipv6_l4proto(skb, ip6h, thoff, &addr, &new_addr);
492 }
493 
nf_flow_dnat_ipv6(const struct flow_offload * flow,struct sk_buff * skb,struct ipv6hdr * ip6h,unsigned int thoff,enum flow_offload_tuple_dir dir)494 static void nf_flow_dnat_ipv6(const struct flow_offload *flow,
495 			      struct sk_buff *skb, struct ipv6hdr *ip6h,
496 			      unsigned int thoff,
497 			      enum flow_offload_tuple_dir dir)
498 {
499 	struct in6_addr addr, new_addr;
500 
501 	switch (dir) {
502 	case FLOW_OFFLOAD_DIR_ORIGINAL:
503 		addr = ip6h->daddr;
504 		new_addr = flow->tuplehash[FLOW_OFFLOAD_DIR_REPLY].tuple.src_v6;
505 		ip6h->daddr = new_addr;
506 		break;
507 	case FLOW_OFFLOAD_DIR_REPLY:
508 		addr = ip6h->saddr;
509 		new_addr = flow->tuplehash[FLOW_OFFLOAD_DIR_ORIGINAL].tuple.dst_v6;
510 		ip6h->saddr = new_addr;
511 		break;
512 	}
513 
514 	nf_flow_nat_ipv6_l4proto(skb, ip6h, thoff, &addr, &new_addr);
515 }
516 
nf_flow_nat_ipv6(const struct flow_offload * flow,struct sk_buff * skb,enum flow_offload_tuple_dir dir,struct ipv6hdr * ip6h)517 static void nf_flow_nat_ipv6(const struct flow_offload *flow,
518 			     struct sk_buff *skb,
519 			     enum flow_offload_tuple_dir dir,
520 			     struct ipv6hdr *ip6h)
521 {
522 	unsigned int thoff = sizeof(*ip6h);
523 
524 	if (test_bit(NF_FLOW_SNAT, &flow->flags)) {
525 		nf_flow_snat_port(flow, skb, thoff, ip6h->nexthdr, dir);
526 		nf_flow_snat_ipv6(flow, skb, ip6h, thoff, dir);
527 	}
528 	if (test_bit(NF_FLOW_DNAT, &flow->flags)) {
529 		nf_flow_dnat_port(flow, skb, thoff, ip6h->nexthdr, dir);
530 		nf_flow_dnat_ipv6(flow, skb, ip6h, thoff, dir);
531 	}
532 }
533 
nf_flow_tuple_ipv6(struct sk_buff * skb,const struct net_device * dev,struct flow_offload_tuple * tuple,u32 * hdrsize,u32 offset)534 static int nf_flow_tuple_ipv6(struct sk_buff *skb, const struct net_device *dev,
535 			      struct flow_offload_tuple *tuple, u32 *hdrsize,
536 			      u32 offset)
537 {
538 	struct flow_ports *ports;
539 	struct ipv6hdr *ip6h;
540 	unsigned int thoff;
541 	u8 nexthdr;
542 
543 	thoff = sizeof(*ip6h) + offset;
544 	if (!pskb_may_pull(skb, thoff))
545 		return -1;
546 
547 	ip6h = (struct ipv6hdr *)(skb_network_header(skb) + offset);
548 
549 	nexthdr = ip6h->nexthdr;
550 	switch (nexthdr) {
551 	case IPPROTO_TCP:
552 		*hdrsize = sizeof(struct tcphdr);
553 		break;
554 	case IPPROTO_UDP:
555 		*hdrsize = sizeof(struct udphdr);
556 		break;
557 #ifdef CONFIG_NF_CT_PROTO_GRE
558 	case IPPROTO_GRE:
559 		*hdrsize = sizeof(struct gre_base_hdr);
560 		break;
561 #endif
562 	default:
563 		return -1;
564 	}
565 
566 	if (ip6h->hop_limit <= 1)
567 		return -1;
568 
569 	if (!pskb_may_pull(skb, thoff + *hdrsize))
570 		return -1;
571 
572 	switch (nexthdr) {
573 	case IPPROTO_TCP:
574 	case IPPROTO_UDP:
575 		ports = (struct flow_ports *)(skb_network_header(skb) + thoff);
576 		tuple->src_port		= ports->source;
577 		tuple->dst_port		= ports->dest;
578 		break;
579 	case IPPROTO_GRE: {
580 		struct gre_base_hdr *greh;
581 
582 		greh = (struct gre_base_hdr *)(skb_network_header(skb) + thoff);
583 		if ((greh->flags & GRE_VERSION) != GRE_VERSION_0)
584 			return -1;
585 		break;
586 	}
587 	}
588 
589 	ip6h = (struct ipv6hdr *)(skb_network_header(skb) + offset);
590 
591 	tuple->src_v6		= ip6h->saddr;
592 	tuple->dst_v6		= ip6h->daddr;
593 	tuple->l3proto		= AF_INET6;
594 	tuple->l4proto		= nexthdr;
595 	tuple->iifidx		= dev->ifindex;
596 	nf_flow_tuple_encap(skb, tuple);
597 
598 	return 0;
599 }
600 
601 unsigned int
nf_flow_offload_ipv6_hook(void * priv,struct sk_buff * skb,const struct nf_hook_state * state)602 nf_flow_offload_ipv6_hook(void *priv, struct sk_buff *skb,
603 			  const struct nf_hook_state *state)
604 {
605 	struct flow_offload_tuple_rhash *tuplehash;
606 	struct nf_flowtable *flow_table = priv;
607 	struct flow_offload_tuple tuple = {};
608 	enum flow_offload_tuple_dir dir;
609 	const struct in6_addr *nexthop;
610 	struct flow_offload *flow;
611 	struct net_device *outdev;
612 	unsigned int thoff, mtu;
613 	u32 hdrsize, offset = 0;
614 	struct ipv6hdr *ip6h;
615 	struct rt6_info *rt;
616 	int ret;
617 
618 	if (skb->protocol != htons(ETH_P_IPV6) &&
619 	    !nf_flow_skb_encap_protocol(skb, htons(ETH_P_IPV6), &offset))
620 		return NF_ACCEPT;
621 
622 	if (nf_flow_tuple_ipv6(skb, state->in, &tuple, &hdrsize, offset) < 0)
623 		return NF_ACCEPT;
624 
625 	tuplehash = flow_offload_lookup(flow_table, &tuple);
626 	if (tuplehash == NULL)
627 		return NF_ACCEPT;
628 
629 	dir = tuplehash->tuple.dir;
630 	flow = container_of(tuplehash, struct flow_offload, tuplehash[dir]);
631 
632 	mtu = flow->tuplehash[dir].tuple.mtu + offset;
633 	if (unlikely(nf_flow_exceeds_mtu(skb, mtu)))
634 		return NF_ACCEPT;
635 
636 	ip6h = (struct ipv6hdr *)(skb_network_header(skb) + offset);
637 	thoff = sizeof(*ip6h) + offset;
638 	if (nf_flow_state_check(flow, ip6h->nexthdr, skb, thoff))
639 		return NF_ACCEPT;
640 
641 	if (!nf_flow_dst_check(&tuplehash->tuple)) {
642 		flow_offload_teardown(flow);
643 		return NF_ACCEPT;
644 	}
645 
646 	if (skb_try_make_writable(skb, thoff + hdrsize))
647 		return NF_DROP;
648 
649 	flow_offload_refresh(flow_table, flow);
650 
651 	nf_flow_encap_pop(skb, tuplehash);
652 
653 	ip6h = ipv6_hdr(skb);
654 	nf_flow_nat_ipv6(flow, skb, dir, ip6h);
655 
656 	ip6h->hop_limit--;
657 	skb_clear_tstamp(skb);
658 
659 	if (flow_table->flags & NF_FLOWTABLE_COUNTER)
660 		nf_ct_acct_update(flow->ct, tuplehash->tuple.dir, skb->len);
661 
662 	if (unlikely(tuplehash->tuple.xmit_type == FLOW_OFFLOAD_XMIT_XFRM)) {
663 		rt = (struct rt6_info *)tuplehash->tuple.dst_cache;
664 		memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
665 		IP6CB(skb)->iif = skb->dev->ifindex;
666 		IP6CB(skb)->flags = IP6SKB_FORWARDED;
667 		return nf_flow_xmit_xfrm(skb, state, &rt->dst);
668 	}
669 
670 	switch (tuplehash->tuple.xmit_type) {
671 	case FLOW_OFFLOAD_XMIT_NEIGH:
672 		rt = (struct rt6_info *)tuplehash->tuple.dst_cache;
673 		outdev = rt->dst.dev;
674 		skb->dev = outdev;
675 		nexthop = rt6_nexthop(rt, &flow->tuplehash[!dir].tuple.src_v6);
676 		skb_dst_set_noref(skb, &rt->dst);
677 		neigh_xmit(NEIGH_ND_TABLE, outdev, nexthop, skb);
678 		ret = NF_STOLEN;
679 		break;
680 	case FLOW_OFFLOAD_XMIT_DIRECT:
681 		ret = nf_flow_queue_xmit(state->net, skb, tuplehash, ETH_P_IPV6);
682 		if (ret == NF_DROP)
683 			flow_offload_teardown(flow);
684 		break;
685 	}
686 
687 	return ret;
688 }
689 EXPORT_SYMBOL_GPL(nf_flow_offload_ipv6_hook);
690