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>
8 #include <linux/ipv6.h>
9 #include <linux/netdevice.h>
12 #include <net/ip6_route.h>
13 #include <net/neighbour.h>
14 #include <net/netfilter/nf_flow_table.h>
15 /* For layer 4 checksum field offset. */
16 #include <linux/tcp.h>
17 #include <linux/udp.h>
19 static int nf_flow_state_check(struct flow_offload *flow, int proto,
20 struct sk_buff *skb, unsigned int thoff)
24 if (proto != IPPROTO_TCP)
27 if (!pskb_may_pull(skb, thoff + sizeof(*tcph)))
30 tcph = (void *)(skb_network_header(skb) + thoff);
31 if (unlikely(tcph->fin || tcph->rst)) {
32 flow_offload_teardown(flow);
39 static int nf_flow_nat_ip_tcp(struct sk_buff *skb, unsigned int thoff,
40 __be32 addr, __be32 new_addr)
44 if (!pskb_may_pull(skb, thoff + sizeof(*tcph)) ||
45 skb_try_make_writable(skb, thoff + sizeof(*tcph)))
48 tcph = (void *)(skb_network_header(skb) + thoff);
49 inet_proto_csum_replace4(&tcph->check, skb, addr, new_addr, true);
54 static int nf_flow_nat_ip_udp(struct sk_buff *skb, unsigned int thoff,
55 __be32 addr, __be32 new_addr)
59 if (!pskb_may_pull(skb, thoff + sizeof(*udph)) ||
60 skb_try_make_writable(skb, thoff + sizeof(*udph)))
63 udph = (void *)(skb_network_header(skb) + thoff);
64 if (udph->check || skb->ip_summed == CHECKSUM_PARTIAL) {
65 inet_proto_csum_replace4(&udph->check, skb, addr,
68 udph->check = CSUM_MANGLED_0;
74 static int nf_flow_nat_ip_l4proto(struct sk_buff *skb, struct iphdr *iph,
75 unsigned int thoff, __be32 addr,
78 switch (iph->protocol) {
80 if (nf_flow_nat_ip_tcp(skb, thoff, addr, new_addr) < 0)
84 if (nf_flow_nat_ip_udp(skb, thoff, addr, new_addr) < 0)
92 static int nf_flow_snat_ip(const struct flow_offload *flow, struct sk_buff *skb,
93 struct iphdr *iph, unsigned int thoff,
94 enum flow_offload_tuple_dir dir)
96 __be32 addr, new_addr;
99 case FLOW_OFFLOAD_DIR_ORIGINAL:
101 new_addr = flow->tuplehash[FLOW_OFFLOAD_DIR_REPLY].tuple.dst_v4.s_addr;
102 iph->saddr = new_addr;
104 case FLOW_OFFLOAD_DIR_REPLY:
106 new_addr = flow->tuplehash[FLOW_OFFLOAD_DIR_ORIGINAL].tuple.src_v4.s_addr;
107 iph->daddr = new_addr;
112 csum_replace4(&iph->check, addr, new_addr);
114 return nf_flow_nat_ip_l4proto(skb, iph, thoff, addr, new_addr);
117 static int nf_flow_dnat_ip(const struct flow_offload *flow, struct sk_buff *skb,
118 struct iphdr *iph, unsigned int thoff,
119 enum flow_offload_tuple_dir dir)
121 __be32 addr, new_addr;
124 case FLOW_OFFLOAD_DIR_ORIGINAL:
126 new_addr = flow->tuplehash[FLOW_OFFLOAD_DIR_REPLY].tuple.src_v4.s_addr;
127 iph->daddr = new_addr;
129 case FLOW_OFFLOAD_DIR_REPLY:
131 new_addr = flow->tuplehash[FLOW_OFFLOAD_DIR_ORIGINAL].tuple.dst_v4.s_addr;
132 iph->saddr = new_addr;
137 csum_replace4(&iph->check, addr, new_addr);
139 return nf_flow_nat_ip_l4proto(skb, iph, thoff, addr, new_addr);
142 static int nf_flow_nat_ip(const struct flow_offload *flow, struct sk_buff *skb,
143 unsigned int thoff, enum flow_offload_tuple_dir dir)
145 struct iphdr *iph = ip_hdr(skb);
147 if (flow->flags & FLOW_OFFLOAD_SNAT &&
148 (nf_flow_snat_port(flow, skb, thoff, iph->protocol, dir) < 0 ||
149 nf_flow_snat_ip(flow, skb, iph, thoff, dir) < 0))
151 if (flow->flags & FLOW_OFFLOAD_DNAT &&
152 (nf_flow_dnat_port(flow, skb, thoff, iph->protocol, dir) < 0 ||
153 nf_flow_dnat_ip(flow, skb, iph, thoff, dir) < 0))
159 static bool ip_has_options(unsigned int thoff)
161 return thoff != sizeof(struct iphdr);
164 static int nf_flow_tuple_ip(struct sk_buff *skb, const struct net_device *dev,
165 struct flow_offload_tuple *tuple)
167 struct flow_ports *ports;
171 if (!pskb_may_pull(skb, sizeof(*iph)))
175 thoff = iph->ihl * 4;
177 if (ip_is_fragment(iph) ||
178 unlikely(ip_has_options(thoff)))
181 if (iph->protocol != IPPROTO_TCP &&
182 iph->protocol != IPPROTO_UDP)
188 thoff = iph->ihl * 4;
189 if (!pskb_may_pull(skb, thoff + sizeof(*ports)))
192 ports = (struct flow_ports *)(skb_network_header(skb) + thoff);
194 tuple->src_v4.s_addr = iph->saddr;
195 tuple->dst_v4.s_addr = iph->daddr;
196 tuple->src_port = ports->source;
197 tuple->dst_port = ports->dest;
198 tuple->l3proto = AF_INET;
199 tuple->l4proto = iph->protocol;
200 tuple->iifidx = dev->ifindex;
205 /* Based on ip_exceeds_mtu(). */
206 static bool nf_flow_exceeds_mtu(const struct sk_buff *skb, unsigned int mtu)
211 if (skb_is_gso(skb) && skb_gso_validate_network_len(skb, mtu))
218 nf_flow_offload_ip_hook(void *priv, struct sk_buff *skb,
219 const struct nf_hook_state *state)
221 struct flow_offload_tuple_rhash *tuplehash;
222 struct nf_flowtable *flow_table = priv;
223 struct flow_offload_tuple tuple = {};
224 enum flow_offload_tuple_dir dir;
225 struct flow_offload *flow;
226 struct net_device *outdev;
232 if (skb->protocol != htons(ETH_P_IP))
235 if (nf_flow_tuple_ip(skb, state->in, &tuple) < 0)
238 tuplehash = flow_offload_lookup(flow_table, &tuple);
239 if (tuplehash == NULL)
242 dir = tuplehash->tuple.dir;
243 flow = container_of(tuplehash, struct flow_offload, tuplehash[dir]);
244 rt = (struct rtable *)flow->tuplehash[dir].tuple.dst_cache;
245 outdev = rt->dst.dev;
247 if (unlikely(nf_flow_exceeds_mtu(skb, flow->tuplehash[dir].tuple.mtu)) &&
248 (ip_hdr(skb)->frag_off & htons(IP_DF)) != 0)
251 if (skb_try_make_writable(skb, sizeof(*iph)))
254 thoff = ip_hdr(skb)->ihl * 4;
255 if (nf_flow_state_check(flow, ip_hdr(skb)->protocol, skb, thoff))
258 if (nf_flow_nat_ip(flow, skb, thoff, dir) < 0)
261 flow->timeout = (u32)jiffies + NF_FLOW_TIMEOUT;
263 ip_decrease_ttl(iph);
266 nexthop = rt_nexthop(rt, flow->tuplehash[!dir].tuple.src_v4.s_addr);
267 skb_dst_set_noref(skb, &rt->dst);
268 neigh_xmit(NEIGH_ARP_TABLE, outdev, &nexthop, skb);
272 EXPORT_SYMBOL_GPL(nf_flow_offload_ip_hook);
274 static int nf_flow_nat_ipv6_tcp(struct sk_buff *skb, unsigned int thoff,
275 struct in6_addr *addr,
276 struct in6_addr *new_addr)
280 if (!pskb_may_pull(skb, thoff + sizeof(*tcph)) ||
281 skb_try_make_writable(skb, thoff + sizeof(*tcph)))
284 tcph = (void *)(skb_network_header(skb) + thoff);
285 inet_proto_csum_replace16(&tcph->check, skb, addr->s6_addr32,
286 new_addr->s6_addr32, true);
291 static int nf_flow_nat_ipv6_udp(struct sk_buff *skb, unsigned int thoff,
292 struct in6_addr *addr,
293 struct in6_addr *new_addr)
297 if (!pskb_may_pull(skb, thoff + sizeof(*udph)) ||
298 skb_try_make_writable(skb, thoff + sizeof(*udph)))
301 udph = (void *)(skb_network_header(skb) + thoff);
302 if (udph->check || skb->ip_summed == CHECKSUM_PARTIAL) {
303 inet_proto_csum_replace16(&udph->check, skb, addr->s6_addr32,
304 new_addr->s6_addr32, true);
306 udph->check = CSUM_MANGLED_0;
312 static int nf_flow_nat_ipv6_l4proto(struct sk_buff *skb, struct ipv6hdr *ip6h,
313 unsigned int thoff, struct in6_addr *addr,
314 struct in6_addr *new_addr)
316 switch (ip6h->nexthdr) {
318 if (nf_flow_nat_ipv6_tcp(skb, thoff, addr, new_addr) < 0)
322 if (nf_flow_nat_ipv6_udp(skb, thoff, addr, new_addr) < 0)
330 static int nf_flow_snat_ipv6(const struct flow_offload *flow,
331 struct sk_buff *skb, struct ipv6hdr *ip6h,
333 enum flow_offload_tuple_dir dir)
335 struct in6_addr addr, new_addr;
338 case FLOW_OFFLOAD_DIR_ORIGINAL:
340 new_addr = flow->tuplehash[FLOW_OFFLOAD_DIR_REPLY].tuple.dst_v6;
341 ip6h->saddr = new_addr;
343 case FLOW_OFFLOAD_DIR_REPLY:
345 new_addr = flow->tuplehash[FLOW_OFFLOAD_DIR_ORIGINAL].tuple.src_v6;
346 ip6h->daddr = new_addr;
352 return nf_flow_nat_ipv6_l4proto(skb, ip6h, thoff, &addr, &new_addr);
355 static int nf_flow_dnat_ipv6(const struct flow_offload *flow,
356 struct sk_buff *skb, struct ipv6hdr *ip6h,
358 enum flow_offload_tuple_dir dir)
360 struct in6_addr addr, new_addr;
363 case FLOW_OFFLOAD_DIR_ORIGINAL:
365 new_addr = flow->tuplehash[FLOW_OFFLOAD_DIR_REPLY].tuple.src_v6;
366 ip6h->daddr = new_addr;
368 case FLOW_OFFLOAD_DIR_REPLY:
370 new_addr = flow->tuplehash[FLOW_OFFLOAD_DIR_ORIGINAL].tuple.dst_v6;
371 ip6h->saddr = new_addr;
377 return nf_flow_nat_ipv6_l4proto(skb, ip6h, thoff, &addr, &new_addr);
380 static int nf_flow_nat_ipv6(const struct flow_offload *flow,
382 enum flow_offload_tuple_dir dir)
384 struct ipv6hdr *ip6h = ipv6_hdr(skb);
385 unsigned int thoff = sizeof(*ip6h);
387 if (flow->flags & FLOW_OFFLOAD_SNAT &&
388 (nf_flow_snat_port(flow, skb, thoff, ip6h->nexthdr, dir) < 0 ||
389 nf_flow_snat_ipv6(flow, skb, ip6h, thoff, dir) < 0))
391 if (flow->flags & FLOW_OFFLOAD_DNAT &&
392 (nf_flow_dnat_port(flow, skb, thoff, ip6h->nexthdr, dir) < 0 ||
393 nf_flow_dnat_ipv6(flow, skb, ip6h, thoff, dir) < 0))
399 static int nf_flow_tuple_ipv6(struct sk_buff *skb, const struct net_device *dev,
400 struct flow_offload_tuple *tuple)
402 struct flow_ports *ports;
403 struct ipv6hdr *ip6h;
406 if (!pskb_may_pull(skb, sizeof(*ip6h)))
409 ip6h = ipv6_hdr(skb);
411 if (ip6h->nexthdr != IPPROTO_TCP &&
412 ip6h->nexthdr != IPPROTO_UDP)
415 if (ip6h->hop_limit <= 1)
418 thoff = sizeof(*ip6h);
419 if (!pskb_may_pull(skb, thoff + sizeof(*ports)))
422 ports = (struct flow_ports *)(skb_network_header(skb) + thoff);
424 tuple->src_v6 = ip6h->saddr;
425 tuple->dst_v6 = ip6h->daddr;
426 tuple->src_port = ports->source;
427 tuple->dst_port = ports->dest;
428 tuple->l3proto = AF_INET6;
429 tuple->l4proto = ip6h->nexthdr;
430 tuple->iifidx = dev->ifindex;
436 nf_flow_offload_ipv6_hook(void *priv, struct sk_buff *skb,
437 const struct nf_hook_state *state)
439 struct flow_offload_tuple_rhash *tuplehash;
440 struct nf_flowtable *flow_table = priv;
441 struct flow_offload_tuple tuple = {};
442 enum flow_offload_tuple_dir dir;
443 struct flow_offload *flow;
444 struct net_device *outdev;
445 struct in6_addr *nexthop;
446 struct ipv6hdr *ip6h;
449 if (skb->protocol != htons(ETH_P_IPV6))
452 if (nf_flow_tuple_ipv6(skb, state->in, &tuple) < 0)
455 tuplehash = flow_offload_lookup(flow_table, &tuple);
456 if (tuplehash == NULL)
459 dir = tuplehash->tuple.dir;
460 flow = container_of(tuplehash, struct flow_offload, tuplehash[dir]);
461 rt = (struct rt6_info *)flow->tuplehash[dir].tuple.dst_cache;
462 outdev = rt->dst.dev;
464 if (unlikely(nf_flow_exceeds_mtu(skb, flow->tuplehash[dir].tuple.mtu)))
467 if (nf_flow_state_check(flow, ipv6_hdr(skb)->nexthdr, skb,
471 if (skb_try_make_writable(skb, sizeof(*ip6h)))
474 if (nf_flow_nat_ipv6(flow, skb, dir) < 0)
477 flow->timeout = (u32)jiffies + NF_FLOW_TIMEOUT;
478 ip6h = ipv6_hdr(skb);
482 nexthop = rt6_nexthop(rt, &flow->tuplehash[!dir].tuple.src_v6);
483 skb_dst_set_noref(skb, &rt->dst);
484 neigh_xmit(NEIGH_ND_TABLE, outdev, nexthop, skb);
488 EXPORT_SYMBOL_GPL(nf_flow_offload_ipv6_hook);