]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * ip_vs_xmit.c: various packet transmitters for IPVS | |
3 | * | |
1da177e4 LT |
4 | * Authors: Wensong Zhang <wensong@linuxvirtualserver.org> |
5 | * Julian Anastasov <ja@ssi.bg> | |
6 | * | |
7 | * This program is free software; you can redistribute it and/or | |
8 | * modify it under the terms of the GNU General Public License | |
9 | * as published by the Free Software Foundation; either version | |
10 | * 2 of the License, or (at your option) any later version. | |
11 | * | |
12 | * Changes: | |
13 | * | |
cb59155f JA |
14 | * Description of forwarding methods: |
15 | * - all transmitters are called from LOCAL_IN (remote clients) and | |
16 | * LOCAL_OUT (local clients) but for ICMP can be called from FORWARD | |
17 | * - not all connections have destination server, for example, | |
18 | * connections in backup server when fwmark is used | |
19 | * - bypass connections use daddr from packet | |
026ace06 JA |
20 | * - we can use dst without ref while sending in RCU section, we use |
21 | * ref when returning NF_ACCEPT for NAT-ed packet via loopback | |
cb59155f JA |
22 | * LOCAL_OUT rules: |
23 | * - skb->dev is NULL, skb->protocol is not set (both are set in POST_ROUTING) | |
24 | * - skb->pkt_type is not set yet | |
25 | * - the only place where we can see skb->sk != NULL | |
1da177e4 LT |
26 | */ |
27 | ||
9aada7ac HE |
28 | #define KMSG_COMPONENT "IPVS" |
29 | #define pr_fmt(fmt) KMSG_COMPONENT ": " fmt | |
30 | ||
1da177e4 | 31 | #include <linux/kernel.h> |
5a0e3ad6 | 32 | #include <linux/slab.h> |
1da177e4 | 33 | #include <linux/tcp.h> /* for tcphdr */ |
c439cb2e | 34 | #include <net/ip.h> |
1da177e4 LT |
35 | #include <net/tcp.h> /* for csum_tcpudp_magic */ |
36 | #include <net/udp.h> | |
37 | #include <net/icmp.h> /* for icmp_send */ | |
38 | #include <net/route.h> /* for ip_route_output */ | |
38cdcc9a JV |
39 | #include <net/ipv6.h> |
40 | #include <net/ip6_route.h> | |
ea1d5d77 | 41 | #include <net/ip_tunnels.h> |
714f095f | 42 | #include <net/addrconf.h> |
38cdcc9a | 43 | #include <linux/icmpv6.h> |
1da177e4 LT |
44 | #include <linux/netfilter.h> |
45 | #include <linux/netfilter_ipv4.h> | |
46 | ||
47 | #include <net/ip_vs.h> | |
48 | ||
17a8f8e3 CG |
49 | enum { |
50 | IP_VS_RT_MODE_LOCAL = 1, /* Allow local dest */ | |
51 | IP_VS_RT_MODE_NON_LOCAL = 2, /* Allow non-local dest */ | |
52 | IP_VS_RT_MODE_RDR = 4, /* Allow redirect from remote daddr to | |
53 | * local | |
54 | */ | |
f2edb9f7 | 55 | IP_VS_RT_MODE_CONNECT = 8, /* Always bind route to saddr */ |
ad4d3ef8 | 56 | IP_VS_RT_MODE_KNOWN_NH = 16,/* Route via remote addr */ |
4115ded1 | 57 | IP_VS_RT_MODE_TUNNEL = 32,/* Tunnel mode */ |
17a8f8e3 | 58 | }; |
1da177e4 | 59 | |
026ace06 JA |
60 | static inline struct ip_vs_dest_dst *ip_vs_dest_dst_alloc(void) |
61 | { | |
62 | return kmalloc(sizeof(struct ip_vs_dest_dst), GFP_ATOMIC); | |
63 | } | |
64 | ||
65 | static inline void ip_vs_dest_dst_free(struct ip_vs_dest_dst *dest_dst) | |
66 | { | |
67 | kfree(dest_dst); | |
68 | } | |
69 | ||
1da177e4 LT |
70 | /* |
71 | * Destination cache to speed up outgoing route lookup | |
72 | */ | |
73 | static inline void | |
026ace06 JA |
74 | __ip_vs_dst_set(struct ip_vs_dest *dest, struct ip_vs_dest_dst *dest_dst, |
75 | struct dst_entry *dst, u32 dst_cookie) | |
1da177e4 | 76 | { |
026ace06 JA |
77 | struct ip_vs_dest_dst *old; |
78 | ||
79 | old = rcu_dereference_protected(dest->dest_dst, | |
80 | lockdep_is_held(&dest->dst_lock)); | |
81 | ||
82 | if (dest_dst) { | |
83 | dest_dst->dst_cache = dst; | |
84 | dest_dst->dst_cookie = dst_cookie; | |
85 | } | |
86 | rcu_assign_pointer(dest->dest_dst, dest_dst); | |
1da177e4 | 87 | |
026ace06 JA |
88 | if (old) |
89 | call_rcu(&old->rcu_head, ip_vs_dest_dst_rcu_free); | |
1da177e4 LT |
90 | } |
91 | ||
026ace06 | 92 | static inline struct ip_vs_dest_dst * |
c90558da | 93 | __ip_vs_dst_check(struct ip_vs_dest *dest) |
1da177e4 | 94 | { |
026ace06 JA |
95 | struct ip_vs_dest_dst *dest_dst = rcu_dereference(dest->dest_dst); |
96 | struct dst_entry *dst; | |
1da177e4 | 97 | |
026ace06 | 98 | if (!dest_dst) |
1da177e4 | 99 | return NULL; |
026ace06 JA |
100 | dst = dest_dst->dst_cache; |
101 | if (dst->obsolete && | |
102 | dst->ops->check(dst, dest_dst->dst_cookie) == NULL) | |
1da177e4 | 103 | return NULL; |
026ace06 | 104 | return dest_dst; |
1da177e4 LT |
105 | } |
106 | ||
590e3f79 JDB |
107 | static inline bool |
108 | __mtu_check_toobig_v6(const struct sk_buff *skb, u32 mtu) | |
109 | { | |
4cdd3408 PM |
110 | if (IP6CB(skb)->frag_max_size) { |
111 | /* frag_max_size tell us that, this packet have been | |
112 | * defragmented by netfilter IPv6 conntrack module. | |
113 | */ | |
114 | if (IP6CB(skb)->frag_max_size > mtu) | |
115 | return true; /* largest fragment violate MTU */ | |
116 | } | |
117 | else if (skb->len > mtu && !skb_is_gso(skb)) { | |
590e3f79 JDB |
118 | return true; /* Packet size violate MTU size */ |
119 | } | |
120 | return false; | |
121 | } | |
122 | ||
f2edb9f7 JA |
123 | /* Get route to daddr, update *saddr, optionally bind route to saddr */ |
124 | static struct rtable *do_output_route4(struct net *net, __be32 daddr, | |
c90558da | 125 | int rt_mode, __be32 *saddr) |
f2edb9f7 JA |
126 | { |
127 | struct flowi4 fl4; | |
128 | struct rtable *rt; | |
129 | int loop = 0; | |
130 | ||
131 | memset(&fl4, 0, sizeof(fl4)); | |
132 | fl4.daddr = daddr; | |
ad4d3ef8 JA |
133 | fl4.flowi4_flags = (rt_mode & IP_VS_RT_MODE_KNOWN_NH) ? |
134 | FLOWI_FLAG_KNOWN_NH : 0; | |
f2edb9f7 JA |
135 | |
136 | retry: | |
137 | rt = ip_route_output_key(net, &fl4); | |
138 | if (IS_ERR(rt)) { | |
139 | /* Invalid saddr ? */ | |
140 | if (PTR_ERR(rt) == -EINVAL && *saddr && | |
141 | rt_mode & IP_VS_RT_MODE_CONNECT && !loop) { | |
142 | *saddr = 0; | |
c90558da | 143 | flowi4_update_output(&fl4, 0, 0, daddr, 0); |
f2edb9f7 JA |
144 | goto retry; |
145 | } | |
146 | IP_VS_DBG_RL("ip_route_output error, dest: %pI4\n", &daddr); | |
147 | return NULL; | |
148 | } else if (!*saddr && rt_mode & IP_VS_RT_MODE_CONNECT && fl4.saddr) { | |
149 | ip_rt_put(rt); | |
150 | *saddr = fl4.saddr; | |
c90558da | 151 | flowi4_update_output(&fl4, 0, 0, daddr, fl4.saddr); |
f2edb9f7 JA |
152 | loop++; |
153 | goto retry; | |
154 | } | |
155 | *saddr = fl4.saddr; | |
156 | return rt; | |
157 | } | |
158 | ||
4a4739d5 AG |
159 | #ifdef CONFIG_IP_VS_IPV6 |
160 | static inline int __ip_vs_is_local_route6(struct rt6_info *rt) | |
161 | { | |
162 | return rt->dst.dev && rt->dst.dev->flags & IFF_LOOPBACK; | |
163 | } | |
164 | #endif | |
165 | ||
166 | static inline bool crosses_local_route_boundary(int skb_af, struct sk_buff *skb, | |
167 | int rt_mode, | |
168 | bool new_rt_is_local) | |
169 | { | |
170 | bool rt_mode_allow_local = !!(rt_mode & IP_VS_RT_MODE_LOCAL); | |
171 | bool rt_mode_allow_non_local = !!(rt_mode & IP_VS_RT_MODE_LOCAL); | |
172 | bool rt_mode_allow_redirect = !!(rt_mode & IP_VS_RT_MODE_RDR); | |
173 | bool source_is_loopback; | |
174 | bool old_rt_is_local; | |
175 | ||
176 | #ifdef CONFIG_IP_VS_IPV6 | |
177 | if (skb_af == AF_INET6) { | |
178 | int addr_type = ipv6_addr_type(&ipv6_hdr(skb)->saddr); | |
179 | ||
180 | source_is_loopback = | |
181 | (!skb->dev || skb->dev->flags & IFF_LOOPBACK) && | |
182 | (addr_type & IPV6_ADDR_LOOPBACK); | |
183 | old_rt_is_local = __ip_vs_is_local_route6( | |
184 | (struct rt6_info *)skb_dst(skb)); | |
185 | } else | |
186 | #endif | |
187 | { | |
188 | source_is_loopback = ipv4_is_loopback(ip_hdr(skb)->saddr); | |
189 | old_rt_is_local = skb_rtable(skb)->rt_flags & RTCF_LOCAL; | |
190 | } | |
191 | ||
192 | if (unlikely(new_rt_is_local)) { | |
193 | if (!rt_mode_allow_local) | |
194 | return true; | |
195 | if (!rt_mode_allow_redirect && !old_rt_is_local) | |
196 | return true; | |
197 | } else { | |
198 | if (!rt_mode_allow_non_local) | |
199 | return true; | |
200 | if (source_is_loopback) | |
201 | return true; | |
202 | } | |
203 | return false; | |
204 | } | |
205 | ||
919aa0b2 AG |
206 | static inline void maybe_update_pmtu(int skb_af, struct sk_buff *skb, int mtu) |
207 | { | |
208 | struct sock *sk = skb->sk; | |
209 | struct rtable *ort = skb_rtable(skb); | |
210 | ||
a8399231 | 211 | if (!skb->dev && sk && sk_fullsock(sk)) |
919aa0b2 AG |
212 | ort->dst.ops->update_pmtu(&ort->dst, sk, NULL, mtu); |
213 | } | |
214 | ||
20868a40 EB |
215 | static inline bool ensure_mtu_is_adequate(struct netns_ipvs *ipvs, int skb_af, |
216 | int rt_mode, | |
c63e4de2 AG |
217 | struct ip_vs_iphdr *ipvsh, |
218 | struct sk_buff *skb, int mtu) | |
219 | { | |
220 | #ifdef CONFIG_IP_VS_IPV6 | |
221 | if (skb_af == AF_INET6) { | |
20868a40 | 222 | struct net *net = ipvs->net; |
c63e4de2 AG |
223 | |
224 | if (unlikely(__mtu_check_toobig_v6(skb, mtu))) { | |
225 | if (!skb->dev) | |
226 | skb->dev = net->loopback_dev; | |
227 | /* only send ICMP too big on first fragment */ | |
89621f31 | 228 | if (!ipvsh->fragoffs && !ip_vs_iph_icmp(ipvsh)) |
c63e4de2 AG |
229 | icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu); |
230 | IP_VS_DBG(1, "frag needed for %pI6c\n", | |
231 | &ipv6_hdr(skb)->saddr); | |
232 | return false; | |
233 | } | |
234 | } else | |
235 | #endif | |
236 | { | |
c63e4de2 AG |
237 | /* If we're going to tunnel the packet and pmtu discovery |
238 | * is disabled, we'll just fragment it anyway | |
239 | */ | |
240 | if ((rt_mode & IP_VS_RT_MODE_TUNNEL) && !sysctl_pmtu_disc(ipvs)) | |
241 | return true; | |
242 | ||
243 | if (unlikely(ip_hdr(skb)->frag_off & htons(IP_DF) && | |
89621f31 AG |
244 | skb->len > mtu && !skb_is_gso(skb) && |
245 | !ip_vs_iph_icmp(ipvsh))) { | |
c63e4de2 AG |
246 | icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED, |
247 | htonl(mtu)); | |
248 | IP_VS_DBG(1, "frag needed for %pI4\n", | |
249 | &ip_hdr(skb)->saddr); | |
250 | return false; | |
251 | } | |
252 | } | |
253 | ||
254 | return true; | |
255 | } | |
256 | ||
8d8e20e2 DB |
257 | static inline bool decrement_ttl(struct netns_ipvs *ipvs, |
258 | int skb_af, | |
259 | struct sk_buff *skb) | |
260 | { | |
261 | struct net *net = ipvs->net; | |
262 | ||
263 | #ifdef CONFIG_IP_VS_IPV6 | |
264 | if (skb_af == AF_INET6) { | |
265 | struct dst_entry *dst = skb_dst(skb); | |
266 | ||
267 | /* check and decrement ttl */ | |
268 | if (ipv6_hdr(skb)->hop_limit <= 1) { | |
269 | /* Force OUTPUT device used as source address */ | |
270 | skb->dev = dst->dev; | |
271 | icmpv6_send(skb, ICMPV6_TIME_EXCEED, | |
272 | ICMPV6_EXC_HOPLIMIT, 0); | |
273 | __IP6_INC_STATS(net, ip6_dst_idev(dst), | |
274 | IPSTATS_MIB_INHDRERRORS); | |
275 | ||
276 | return false; | |
277 | } | |
278 | ||
279 | /* don't propagate ttl change to cloned packets */ | |
280 | if (!skb_make_writable(skb, sizeof(struct ipv6hdr))) | |
281 | return false; | |
282 | ||
283 | ipv6_hdr(skb)->hop_limit--; | |
284 | } else | |
285 | #endif | |
286 | { | |
287 | if (ip_hdr(skb)->ttl <= 1) { | |
288 | /* Tell the sender its packet died... */ | |
289 | __IP_INC_STATS(net, IPSTATS_MIB_INHDRERRORS); | |
290 | icmp_send(skb, ICMP_TIME_EXCEEDED, ICMP_EXC_TTL, 0); | |
291 | return false; | |
292 | } | |
293 | ||
294 | /* don't propagate ttl change to cloned packets */ | |
295 | if (!skb_make_writable(skb, sizeof(struct iphdr))) | |
296 | return false; | |
297 | ||
298 | /* Decrease ttl */ | |
299 | ip_decrease_ttl(ip_hdr(skb)); | |
300 | } | |
301 | ||
302 | return true; | |
303 | } | |
304 | ||
17a8f8e3 | 305 | /* Get route to destination or remote server */ |
4115ded1 | 306 | static int |
ecfe87b8 EB |
307 | __ip_vs_get_out_rt(struct netns_ipvs *ipvs, int skb_af, struct sk_buff *skb, |
308 | struct ip_vs_dest *dest, | |
c63e4de2 AG |
309 | __be32 daddr, int rt_mode, __be32 *ret_saddr, |
310 | struct ip_vs_iphdr *ipvsh) | |
1da177e4 | 311 | { |
ecfe87b8 | 312 | struct net *net = ipvs->net; |
026ace06 | 313 | struct ip_vs_dest_dst *dest_dst; |
1da177e4 | 314 | struct rtable *rt; /* Route to the other host */ |
4115ded1 | 315 | int mtu; |
026ace06 | 316 | int local, noref = 1; |
1da177e4 LT |
317 | |
318 | if (dest) { | |
026ace06 JA |
319 | dest_dst = __ip_vs_dst_check(dest); |
320 | if (likely(dest_dst)) | |
321 | rt = (struct rtable *) dest_dst->dst_cache; | |
322 | else { | |
323 | dest_dst = ip_vs_dest_dst_alloc(); | |
ac69269a | 324 | spin_lock_bh(&dest->dst_lock); |
026ace06 JA |
325 | if (!dest_dst) { |
326 | __ip_vs_dst_set(dest, NULL, NULL, 0); | |
ac69269a | 327 | spin_unlock_bh(&dest->dst_lock); |
026ace06 JA |
328 | goto err_unreach; |
329 | } | |
c90558da | 330 | rt = do_output_route4(net, dest->addr.ip, rt_mode, |
026ace06 | 331 | &dest_dst->dst_saddr.ip); |
f2edb9f7 | 332 | if (!rt) { |
026ace06 | 333 | __ip_vs_dst_set(dest, NULL, NULL, 0); |
ac69269a | 334 | spin_unlock_bh(&dest->dst_lock); |
026ace06 | 335 | ip_vs_dest_dst_free(dest_dst); |
4115ded1 | 336 | goto err_unreach; |
1da177e4 | 337 | } |
026ace06 | 338 | __ip_vs_dst_set(dest, dest_dst, &rt->dst, 0); |
ac69269a | 339 | spin_unlock_bh(&dest->dst_lock); |
c90558da | 340 | IP_VS_DBG(10, "new dst %pI4, src %pI4, refcnt=%d\n", |
026ace06 | 341 | &dest->addr.ip, &dest_dst->dst_saddr.ip, |
c90558da | 342 | atomic_read(&rt->dst.__refcnt)); |
1da177e4 | 343 | } |
c92f5ca2 | 344 | if (ret_saddr) |
026ace06 | 345 | *ret_saddr = dest_dst->dst_saddr.ip; |
1da177e4 | 346 | } else { |
f2edb9f7 | 347 | __be32 saddr = htonl(INADDR_ANY); |
c92f5ca2 | 348 | |
026ace06 JA |
349 | noref = 0; |
350 | ||
f2edb9f7 JA |
351 | /* For such unconfigured boxes avoid many route lookups |
352 | * for performance reasons because we do not remember saddr | |
353 | */ | |
354 | rt_mode &= ~IP_VS_RT_MODE_CONNECT; | |
c90558da | 355 | rt = do_output_route4(net, daddr, rt_mode, &saddr); |
f2edb9f7 | 356 | if (!rt) |
4115ded1 | 357 | goto err_unreach; |
c92f5ca2 | 358 | if (ret_saddr) |
f2edb9f7 | 359 | *ret_saddr = saddr; |
1da177e4 LT |
360 | } |
361 | ||
4115ded1 | 362 | local = (rt->rt_flags & RTCF_LOCAL) ? 1 : 0; |
4a4739d5 AG |
363 | if (unlikely(crosses_local_route_boundary(skb_af, skb, rt_mode, |
364 | local))) { | |
365 | IP_VS_DBG_RL("We are crossing local and non-local addresses" | |
3d53666b | 366 | " daddr=%pI4\n", &daddr); |
4115ded1 | 367 | goto err_put; |
fc604767 | 368 | } |
4a4739d5 AG |
369 | |
370 | if (unlikely(local)) { | |
4115ded1 | 371 | /* skb to local stack, preserve old route */ |
026ace06 JA |
372 | if (!noref) |
373 | ip_rt_put(rt); | |
4115ded1 | 374 | return local; |
fc604767 | 375 | } |
4115ded1 | 376 | |
8d8e20e2 DB |
377 | if (!decrement_ttl(ipvs, skb_af, skb)) |
378 | goto err_put; | |
379 | ||
4115ded1 JA |
380 | if (likely(!(rt_mode & IP_VS_RT_MODE_TUNNEL))) { |
381 | mtu = dst_mtu(&rt->dst); | |
4115ded1 | 382 | } else { |
4115ded1 JA |
383 | mtu = dst_mtu(&rt->dst) - sizeof(struct iphdr); |
384 | if (mtu < 68) { | |
385 | IP_VS_DBG_RL("%s(): mtu less than 68\n", __func__); | |
386 | goto err_put; | |
387 | } | |
919aa0b2 | 388 | maybe_update_pmtu(skb_af, skb, mtu); |
fc604767 JA |
389 | } |
390 | ||
20868a40 | 391 | if (!ensure_mtu_is_adequate(ipvs, skb_af, rt_mode, ipvsh, skb, mtu)) |
4115ded1 | 392 | goto err_put; |
4115ded1 JA |
393 | |
394 | skb_dst_drop(skb); | |
026ace06 JA |
395 | if (noref) { |
396 | if (!local) | |
dbfc4fb7 | 397 | skb_dst_set_noref(skb, &rt->dst); |
026ace06 JA |
398 | else |
399 | skb_dst_set(skb, dst_clone(&rt->dst)); | |
400 | } else | |
401 | skb_dst_set(skb, &rt->dst); | |
4115ded1 JA |
402 | |
403 | return local; | |
404 | ||
405 | err_put: | |
026ace06 JA |
406 | if (!noref) |
407 | ip_rt_put(rt); | |
4115ded1 JA |
408 | return -1; |
409 | ||
410 | err_unreach: | |
411 | dst_link_failure(skb); | |
412 | return -1; | |
1da177e4 LT |
413 | } |
414 | ||
38cdcc9a | 415 | #ifdef CONFIG_IP_VS_IPV6 |
714f095f HS |
416 | static struct dst_entry * |
417 | __ip_vs_route_output_v6(struct net *net, struct in6_addr *daddr, | |
48e8aa6e | 418 | struct in6_addr *ret_saddr, int do_xfrm, int rt_mode) |
714f095f HS |
419 | { |
420 | struct dst_entry *dst; | |
4c9483b2 DM |
421 | struct flowi6 fl6 = { |
422 | .daddr = *daddr, | |
714f095f HS |
423 | }; |
424 | ||
48e8aa6e MKL |
425 | if (rt_mode & IP_VS_RT_MODE_KNOWN_NH) |
426 | fl6.flowi6_flags = FLOWI_FLAG_KNOWN_NH; | |
427 | ||
4c9483b2 | 428 | dst = ip6_route_output(net, NULL, &fl6); |
714f095f HS |
429 | if (dst->error) |
430 | goto out_err; | |
431 | if (!ret_saddr) | |
432 | return dst; | |
4c9483b2 | 433 | if (ipv6_addr_any(&fl6.saddr) && |
714f095f | 434 | ipv6_dev_get_saddr(net, ip6_dst_idev(dst)->dev, |
4c9483b2 | 435 | &fl6.daddr, 0, &fl6.saddr) < 0) |
714f095f | 436 | goto out_err; |
452edd59 | 437 | if (do_xfrm) { |
4c9483b2 | 438 | dst = xfrm_lookup(net, dst, flowi6_to_flowi(&fl6), NULL, 0); |
452edd59 DM |
439 | if (IS_ERR(dst)) { |
440 | dst = NULL; | |
441 | goto out_err; | |
442 | } | |
443 | } | |
4e3fd7a0 | 444 | *ret_saddr = fl6.saddr; |
714f095f HS |
445 | return dst; |
446 | ||
447 | out_err: | |
448 | dst_release(dst); | |
449 | IP_VS_DBG_RL("ip6_route_output error, dest: %pI6\n", daddr); | |
450 | return NULL; | |
451 | } | |
452 | ||
fc604767 JA |
453 | /* |
454 | * Get route to destination or remote server | |
fc604767 | 455 | */ |
4115ded1 | 456 | static int |
f5745f8a EB |
457 | __ip_vs_get_out_rt_v6(struct netns_ipvs *ipvs, int skb_af, struct sk_buff *skb, |
458 | struct ip_vs_dest *dest, | |
fc604767 | 459 | struct in6_addr *daddr, struct in6_addr *ret_saddr, |
4115ded1 | 460 | struct ip_vs_iphdr *ipvsh, int do_xfrm, int rt_mode) |
38cdcc9a | 461 | { |
f5745f8a | 462 | struct net *net = ipvs->net; |
026ace06 | 463 | struct ip_vs_dest_dst *dest_dst; |
38cdcc9a | 464 | struct rt6_info *rt; /* Route to the other host */ |
714f095f | 465 | struct dst_entry *dst; |
4115ded1 | 466 | int mtu; |
026ace06 | 467 | int local, noref = 1; |
38cdcc9a JV |
468 | |
469 | if (dest) { | |
026ace06 JA |
470 | dest_dst = __ip_vs_dst_check(dest); |
471 | if (likely(dest_dst)) | |
472 | rt = (struct rt6_info *) dest_dst->dst_cache; | |
473 | else { | |
714f095f | 474 | u32 cookie; |
38cdcc9a | 475 | |
026ace06 | 476 | dest_dst = ip_vs_dest_dst_alloc(); |
ac69269a | 477 | spin_lock_bh(&dest->dst_lock); |
026ace06 JA |
478 | if (!dest_dst) { |
479 | __ip_vs_dst_set(dest, NULL, NULL, 0); | |
ac69269a | 480 | spin_unlock_bh(&dest->dst_lock); |
026ace06 JA |
481 | goto err_unreach; |
482 | } | |
714f095f | 483 | dst = __ip_vs_route_output_v6(net, &dest->addr.in6, |
026ace06 | 484 | &dest_dst->dst_saddr.in6, |
48e8aa6e | 485 | do_xfrm, rt_mode); |
714f095f | 486 | if (!dst) { |
026ace06 | 487 | __ip_vs_dst_set(dest, NULL, NULL, 0); |
ac69269a | 488 | spin_unlock_bh(&dest->dst_lock); |
026ace06 | 489 | ip_vs_dest_dst_free(dest_dst); |
4115ded1 | 490 | goto err_unreach; |
38cdcc9a | 491 | } |
714f095f | 492 | rt = (struct rt6_info *) dst; |
b197df4f | 493 | cookie = rt6_get_cookie(rt); |
026ace06 | 494 | __ip_vs_dst_set(dest, dest_dst, &rt->dst, cookie); |
ac69269a | 495 | spin_unlock_bh(&dest->dst_lock); |
714f095f | 496 | IP_VS_DBG(10, "new dst %pI6, src %pI6, refcnt=%d\n", |
026ace06 | 497 | &dest->addr.in6, &dest_dst->dst_saddr.in6, |
d8d1f30b | 498 | atomic_read(&rt->dst.__refcnt)); |
38cdcc9a | 499 | } |
714f095f | 500 | if (ret_saddr) |
026ace06 | 501 | *ret_saddr = dest_dst->dst_saddr.in6; |
38cdcc9a | 502 | } else { |
026ace06 | 503 | noref = 0; |
48e8aa6e MKL |
504 | dst = __ip_vs_route_output_v6(net, daddr, ret_saddr, do_xfrm, |
505 | rt_mode); | |
714f095f | 506 | if (!dst) |
4115ded1 | 507 | goto err_unreach; |
714f095f | 508 | rt = (struct rt6_info *) dst; |
38cdcc9a JV |
509 | } |
510 | ||
fc604767 | 511 | local = __ip_vs_is_local_route6(rt); |
4a4739d5 AG |
512 | |
513 | if (unlikely(crosses_local_route_boundary(skb_af, skb, rt_mode, | |
514 | local))) { | |
515 | IP_VS_DBG_RL("We are crossing local and non-local addresses" | |
3d53666b | 516 | " daddr=%pI6\n", daddr); |
4115ded1 | 517 | goto err_put; |
fc604767 | 518 | } |
4a4739d5 AG |
519 | |
520 | if (unlikely(local)) { | |
4115ded1 | 521 | /* skb to local stack, preserve old route */ |
026ace06 JA |
522 | if (!noref) |
523 | dst_release(&rt->dst); | |
4115ded1 | 524 | return local; |
fc604767 | 525 | } |
4115ded1 | 526 | |
8d8e20e2 DB |
527 | if (!decrement_ttl(ipvs, skb_af, skb)) |
528 | goto err_put; | |
529 | ||
4115ded1 JA |
530 | /* MTU checking */ |
531 | if (likely(!(rt_mode & IP_VS_RT_MODE_TUNNEL))) | |
532 | mtu = dst_mtu(&rt->dst); | |
533 | else { | |
4115ded1 JA |
534 | mtu = dst_mtu(&rt->dst) - sizeof(struct ipv6hdr); |
535 | if (mtu < IPV6_MIN_MTU) { | |
536 | IP_VS_DBG_RL("%s(): mtu less than %d\n", __func__, | |
537 | IPV6_MIN_MTU); | |
538 | goto err_put; | |
539 | } | |
919aa0b2 | 540 | maybe_update_pmtu(skb_af, skb, mtu); |
fc604767 JA |
541 | } |
542 | ||
20868a40 | 543 | if (!ensure_mtu_is_adequate(ipvs, skb_af, rt_mode, ipvsh, skb, mtu)) |
4115ded1 | 544 | goto err_put; |
4115ded1 JA |
545 | |
546 | skb_dst_drop(skb); | |
026ace06 JA |
547 | if (noref) { |
548 | if (!local) | |
dbfc4fb7 | 549 | skb_dst_set_noref(skb, &rt->dst); |
026ace06 JA |
550 | else |
551 | skb_dst_set(skb, dst_clone(&rt->dst)); | |
552 | } else | |
553 | skb_dst_set(skb, &rt->dst); | |
4115ded1 JA |
554 | |
555 | return local; | |
556 | ||
557 | err_put: | |
026ace06 JA |
558 | if (!noref) |
559 | dst_release(&rt->dst); | |
4115ded1 JA |
560 | return -1; |
561 | ||
562 | err_unreach: | |
326bf17e AG |
563 | /* The ip6_link_failure function requires the dev field to be set |
564 | * in order to get the net (further for the sake of fwmark | |
565 | * reflection). | |
566 | */ | |
567 | if (!skb->dev) | |
568 | skb->dev = skb_dst(skb)->dev; | |
569 | ||
4115ded1 JA |
570 | dst_link_failure(skb); |
571 | return -1; | |
38cdcc9a JV |
572 | } |
573 | #endif | |
574 | ||
1da177e4 | 575 | |
b8abdf09 JA |
576 | /* return NF_ACCEPT to allow forwarding or other NF_xxx on error */ |
577 | static inline int ip_vs_tunnel_xmit_prepare(struct sk_buff *skb, | |
578 | struct ip_vs_conn *cp) | |
579 | { | |
580 | int ret = NF_ACCEPT; | |
581 | ||
582 | skb->ipvs_property = 1; | |
583 | if (unlikely(cp->flags & IP_VS_CONN_F_NFCT)) | |
584 | ret = ip_vs_confirm_conntrack(skb); | |
585 | if (ret == NF_ACCEPT) { | |
586 | nf_reset(skb); | |
587 | skb_forward_csum(skb); | |
588 | } | |
589 | return ret; | |
590 | } | |
591 | ||
71563f34 AG |
592 | /* In the event of a remote destination, it's possible that we would have |
593 | * matches against an old socket (particularly a TIME-WAIT socket). This | |
594 | * causes havoc down the line (ip_local_out et. al. expect regular sockets | |
595 | * and invalid memory accesses will happen) so simply drop the association | |
596 | * in this case. | |
597 | */ | |
598 | static inline void ip_vs_drop_early_demux_sk(struct sk_buff *skb) | |
599 | { | |
600 | /* If dev is set, the packet came from the LOCAL_IN callback and | |
601 | * not from a local TCP socket. | |
602 | */ | |
603 | if (skb->dev) | |
604 | skb_orphan(skb); | |
605 | } | |
606 | ||
b8abdf09 JA |
607 | /* return NF_STOLEN (sent) or NF_ACCEPT if local=1 (not sent) */ |
608 | static inline int ip_vs_nat_send_or_cont(int pf, struct sk_buff *skb, | |
609 | struct ip_vs_conn *cp, int local) | |
610 | { | |
611 | int ret = NF_STOLEN; | |
612 | ||
613 | skb->ipvs_property = 1; | |
614 | if (likely(!(cp->flags & IP_VS_CONN_F_NFCT))) | |
615 | ip_vs_notrack(skb); | |
616 | else | |
617 | ip_vs_update_conntrack(skb, cp, 1); | |
71563f34 AG |
618 | |
619 | /* Remove the early_demux association unless it's bound for the | |
620 | * exact same port and address on this host after translation. | |
621 | */ | |
622 | if (!local || cp->vport != cp->dport || | |
623 | !ip_vs_addr_equal(cp->af, &cp->vaddr, &cp->daddr)) | |
624 | ip_vs_drop_early_demux_sk(skb); | |
625 | ||
b8abdf09 JA |
626 | if (!local) { |
627 | skb_forward_csum(skb); | |
58dbc6f2 | 628 | NF_HOOK(pf, NF_INET_LOCAL_OUT, cp->ipvs->net, NULL, skb, |
13206b6b | 629 | NULL, skb_dst(skb)->dev, dst_output); |
b8abdf09 JA |
630 | } else |
631 | ret = NF_ACCEPT; | |
71563f34 | 632 | |
b8abdf09 JA |
633 | return ret; |
634 | } | |
635 | ||
636 | /* return NF_STOLEN (sent) or NF_ACCEPT if local=1 (not sent) */ | |
637 | static inline int ip_vs_send_or_cont(int pf, struct sk_buff *skb, | |
638 | struct ip_vs_conn *cp, int local) | |
639 | { | |
640 | int ret = NF_STOLEN; | |
641 | ||
642 | skb->ipvs_property = 1; | |
643 | if (likely(!(cp->flags & IP_VS_CONN_F_NFCT))) | |
644 | ip_vs_notrack(skb); | |
645 | if (!local) { | |
71563f34 | 646 | ip_vs_drop_early_demux_sk(skb); |
b8abdf09 | 647 | skb_forward_csum(skb); |
58dbc6f2 | 648 | NF_HOOK(pf, NF_INET_LOCAL_OUT, cp->ipvs->net, NULL, skb, |
13206b6b | 649 | NULL, skb_dst(skb)->dev, dst_output); |
b8abdf09 JA |
650 | } else |
651 | ret = NF_ACCEPT; | |
652 | return ret; | |
653 | } | |
1da177e4 LT |
654 | |
655 | ||
656 | /* | |
657 | * NULL transmitter (do nothing except return NF_ACCEPT) | |
658 | */ | |
659 | int | |
660 | ip_vs_null_xmit(struct sk_buff *skb, struct ip_vs_conn *cp, | |
d4383f04 | 661 | struct ip_vs_protocol *pp, struct ip_vs_iphdr *ipvsh) |
1da177e4 LT |
662 | { |
663 | /* we do not touch skb and do not need pskb ptr */ | |
b8abdf09 | 664 | return ip_vs_send_or_cont(NFPROTO_IPV4, skb, cp, 1); |
1da177e4 LT |
665 | } |
666 | ||
667 | ||
668 | /* | |
669 | * Bypass transmitter | |
670 | * Let packets bypass the destination when the destination is not | |
671 | * available, it may be only used in transparent cache cluster. | |
672 | */ | |
673 | int | |
674 | ip_vs_bypass_xmit(struct sk_buff *skb, struct ip_vs_conn *cp, | |
d4383f04 | 675 | struct ip_vs_protocol *pp, struct ip_vs_iphdr *ipvsh) |
1da177e4 | 676 | { |
eddc9ec5 | 677 | struct iphdr *iph = ip_hdr(skb); |
1da177e4 LT |
678 | |
679 | EnterFunction(10); | |
680 | ||
026ace06 | 681 | rcu_read_lock(); |
ecfe87b8 | 682 | if (__ip_vs_get_out_rt(cp->ipvs, cp->af, skb, NULL, iph->daddr, |
c63e4de2 | 683 | IP_VS_RT_MODE_NON_LOCAL, NULL, ipvsh) < 0) |
1da177e4 | 684 | goto tx_error; |
1da177e4 | 685 | |
4115ded1 | 686 | ip_send_check(iph); |
1da177e4 LT |
687 | |
688 | /* Another hack: avoid icmp_send in ip_fragment */ | |
60ff7467 | 689 | skb->ignore_df = 1; |
1da177e4 | 690 | |
b8abdf09 | 691 | ip_vs_send_or_cont(NFPROTO_IPV4, skb, cp, 0); |
026ace06 | 692 | rcu_read_unlock(); |
1da177e4 LT |
693 | |
694 | LeaveFunction(10); | |
695 | return NF_STOLEN; | |
696 | ||
1da177e4 LT |
697 | tx_error: |
698 | kfree_skb(skb); | |
026ace06 | 699 | rcu_read_unlock(); |
1da177e4 LT |
700 | LeaveFunction(10); |
701 | return NF_STOLEN; | |
702 | } | |
703 | ||
b3cdd2a7 JV |
704 | #ifdef CONFIG_IP_VS_IPV6 |
705 | int | |
706 | ip_vs_bypass_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp, | |
4115ded1 | 707 | struct ip_vs_protocol *pp, struct ip_vs_iphdr *ipvsh) |
b3cdd2a7 | 708 | { |
3481894f AG |
709 | struct ipv6hdr *iph = ipv6_hdr(skb); |
710 | ||
b3cdd2a7 JV |
711 | EnterFunction(10); |
712 | ||
026ace06 | 713 | rcu_read_lock(); |
f5745f8a EB |
714 | if (__ip_vs_get_out_rt_v6(cp->ipvs, cp->af, skb, NULL, |
715 | &iph->daddr, NULL, | |
4115ded1 | 716 | ipvsh, 0, IP_VS_RT_MODE_NON_LOCAL) < 0) |
b3cdd2a7 | 717 | goto tx_error; |
b3cdd2a7 JV |
718 | |
719 | /* Another hack: avoid icmp_send in ip_fragment */ | |
60ff7467 | 720 | skb->ignore_df = 1; |
b3cdd2a7 | 721 | |
b8abdf09 | 722 | ip_vs_send_or_cont(NFPROTO_IPV6, skb, cp, 0); |
026ace06 | 723 | rcu_read_unlock(); |
b3cdd2a7 JV |
724 | |
725 | LeaveFunction(10); | |
726 | return NF_STOLEN; | |
727 | ||
b3cdd2a7 JV |
728 | tx_error: |
729 | kfree_skb(skb); | |
026ace06 | 730 | rcu_read_unlock(); |
b3cdd2a7 JV |
731 | LeaveFunction(10); |
732 | return NF_STOLEN; | |
733 | } | |
734 | #endif | |
1da177e4 LT |
735 | |
736 | /* | |
737 | * NAT transmitter (only for outside-to-inside nat forwarding) | |
738 | * Not used for related ICMP | |
739 | */ | |
740 | int | |
741 | ip_vs_nat_xmit(struct sk_buff *skb, struct ip_vs_conn *cp, | |
d4383f04 | 742 | struct ip_vs_protocol *pp, struct ip_vs_iphdr *ipvsh) |
1da177e4 LT |
743 | { |
744 | struct rtable *rt; /* Route to the other host */ | |
4115ded1 | 745 | int local, rc, was_input; |
1da177e4 LT |
746 | |
747 | EnterFunction(10); | |
748 | ||
026ace06 | 749 | rcu_read_lock(); |
1da177e4 LT |
750 | /* check if it is a connection of no-client-port */ |
751 | if (unlikely(cp->flags & IP_VS_CONN_F_NO_CPORT)) { | |
014d730d | 752 | __be16 _pt, *p; |
4115ded1 JA |
753 | |
754 | p = skb_header_pointer(skb, ipvsh->len, sizeof(_pt), &_pt); | |
1da177e4 LT |
755 | if (p == NULL) |
756 | goto tx_error; | |
757 | ip_vs_conn_fill_cport(cp, *p); | |
758 | IP_VS_DBG(10, "filled cport=%d\n", ntohs(*p)); | |
759 | } | |
760 | ||
4115ded1 | 761 | was_input = rt_is_input_route(skb_rtable(skb)); |
ecfe87b8 | 762 | local = __ip_vs_get_out_rt(cp->ipvs, cp->af, skb, cp->dest, cp->daddr.ip, |
4115ded1 JA |
763 | IP_VS_RT_MODE_LOCAL | |
764 | IP_VS_RT_MODE_NON_LOCAL | | |
c63e4de2 | 765 | IP_VS_RT_MODE_RDR, NULL, ipvsh); |
4115ded1 JA |
766 | if (local < 0) |
767 | goto tx_error; | |
768 | rt = skb_rtable(skb); | |
fc604767 JA |
769 | /* |
770 | * Avoid duplicate tuple in reply direction for NAT traffic | |
771 | * to local address when connection is sync-ed | |
772 | */ | |
c0cd1156 | 773 | #if IS_ENABLED(CONFIG_NF_CONNTRACK) |
fc604767 JA |
774 | if (cp->flags & IP_VS_CONN_F_SYNC && local) { |
775 | enum ip_conntrack_info ctinfo; | |
05b4b065 | 776 | struct nf_conn *ct = nf_ct_get(skb, &ctinfo); |
fc604767 | 777 | |
ab8bc7ed | 778 | if (ct) { |
b0e010c5 | 779 | IP_VS_DBG_RL_PKT(10, AF_INET, pp, skb, ipvsh->off, |
0d79641a | 780 | "ip_vs_nat_xmit(): " |
fc604767 | 781 | "stopping DNAT to local address"); |
4115ded1 | 782 | goto tx_error; |
fc604767 JA |
783 | } |
784 | } | |
785 | #endif | |
786 | ||
787 | /* From world but DNAT to loopback address? */ | |
4115ded1 | 788 | if (local && ipv4_is_loopback(cp->daddr.ip) && was_input) { |
b0e010c5 AG |
789 | IP_VS_DBG_RL_PKT(1, AF_INET, pp, skb, ipvsh->off, |
790 | "ip_vs_nat_xmit(): stopping DNAT to loopback " | |
791 | "address"); | |
4115ded1 | 792 | goto tx_error; |
1da177e4 LT |
793 | } |
794 | ||
795 | /* copy-on-write the packet before mangling it */ | |
af1e1cf0 | 796 | if (!skb_make_writable(skb, sizeof(struct iphdr))) |
4115ded1 | 797 | goto tx_error; |
1da177e4 | 798 | |
d8d1f30b | 799 | if (skb_cow(skb, rt->dst.dev->hard_header_len)) |
4115ded1 | 800 | goto tx_error; |
1da177e4 | 801 | |
1da177e4 | 802 | /* mangle the packet */ |
d4383f04 | 803 | if (pp->dnat_handler && !pp->dnat_handler(skb, pp, cp, ipvsh)) |
4115ded1 | 804 | goto tx_error; |
e7ade46a | 805 | ip_hdr(skb)->daddr = cp->daddr.ip; |
eddc9ec5 | 806 | ip_send_check(ip_hdr(skb)); |
1da177e4 | 807 | |
b0e010c5 | 808 | IP_VS_DBG_PKT(10, AF_INET, pp, skb, ipvsh->off, "After DNAT"); |
1da177e4 LT |
809 | |
810 | /* FIXME: when application helper enlarges the packet and the length | |
811 | is larger than the MTU of outgoing device, there will be still | |
812 | MTU problem. */ | |
813 | ||
814 | /* Another hack: avoid icmp_send in ip_fragment */ | |
60ff7467 | 815 | skb->ignore_df = 1; |
1da177e4 | 816 | |
b8abdf09 | 817 | rc = ip_vs_nat_send_or_cont(NFPROTO_IPV4, skb, cp, local); |
026ace06 | 818 | rcu_read_unlock(); |
1da177e4 LT |
819 | |
820 | LeaveFunction(10); | |
b8abdf09 | 821 | return rc; |
1da177e4 | 822 | |
1da177e4 | 823 | tx_error: |
1da177e4 | 824 | kfree_skb(skb); |
026ace06 | 825 | rcu_read_unlock(); |
f4bc17cd | 826 | LeaveFunction(10); |
1da177e4 | 827 | return NF_STOLEN; |
1da177e4 LT |
828 | } |
829 | ||
b3cdd2a7 JV |
830 | #ifdef CONFIG_IP_VS_IPV6 |
831 | int | |
832 | ip_vs_nat_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp, | |
4115ded1 | 833 | struct ip_vs_protocol *pp, struct ip_vs_iphdr *ipvsh) |
b3cdd2a7 JV |
834 | { |
835 | struct rt6_info *rt; /* Route to the other host */ | |
b8abdf09 | 836 | int local, rc; |
b3cdd2a7 JV |
837 | |
838 | EnterFunction(10); | |
839 | ||
026ace06 | 840 | rcu_read_lock(); |
b3cdd2a7 | 841 | /* check if it is a connection of no-client-port */ |
4115ded1 | 842 | if (unlikely(cp->flags & IP_VS_CONN_F_NO_CPORT && !ipvsh->fragoffs)) { |
b3cdd2a7 | 843 | __be16 _pt, *p; |
4115ded1 | 844 | p = skb_header_pointer(skb, ipvsh->len, sizeof(_pt), &_pt); |
b3cdd2a7 JV |
845 | if (p == NULL) |
846 | goto tx_error; | |
847 | ip_vs_conn_fill_cport(cp, *p); | |
848 | IP_VS_DBG(10, "filled cport=%d\n", ntohs(*p)); | |
849 | } | |
850 | ||
f5745f8a EB |
851 | local = __ip_vs_get_out_rt_v6(cp->ipvs, cp->af, skb, cp->dest, |
852 | &cp->daddr.in6, | |
4a4739d5 | 853 | NULL, ipvsh, 0, |
4115ded1 JA |
854 | IP_VS_RT_MODE_LOCAL | |
855 | IP_VS_RT_MODE_NON_LOCAL | | |
856 | IP_VS_RT_MODE_RDR); | |
857 | if (local < 0) | |
858 | goto tx_error; | |
859 | rt = (struct rt6_info *) skb_dst(skb); | |
fc604767 JA |
860 | /* |
861 | * Avoid duplicate tuple in reply direction for NAT traffic | |
862 | * to local address when connection is sync-ed | |
863 | */ | |
c0cd1156 | 864 | #if IS_ENABLED(CONFIG_NF_CONNTRACK) |
fc604767 JA |
865 | if (cp->flags & IP_VS_CONN_F_SYNC && local) { |
866 | enum ip_conntrack_info ctinfo; | |
05b4b065 | 867 | struct nf_conn *ct = nf_ct_get(skb, &ctinfo); |
fc604767 | 868 | |
ab8bc7ed | 869 | if (ct) { |
b0e010c5 | 870 | IP_VS_DBG_RL_PKT(10, AF_INET6, pp, skb, ipvsh->off, |
fc604767 JA |
871 | "ip_vs_nat_xmit_v6(): " |
872 | "stopping DNAT to local address"); | |
4115ded1 | 873 | goto tx_error; |
fc604767 JA |
874 | } |
875 | } | |
876 | #endif | |
877 | ||
878 | /* From world but DNAT to loopback address? */ | |
879 | if (local && skb->dev && !(skb->dev->flags & IFF_LOOPBACK) && | |
fd0273d7 | 880 | ipv6_addr_type(&cp->daddr.in6) & IPV6_ADDR_LOOPBACK) { |
b0e010c5 | 881 | IP_VS_DBG_RL_PKT(1, AF_INET6, pp, skb, ipvsh->off, |
fc604767 JA |
882 | "ip_vs_nat_xmit_v6(): " |
883 | "stopping DNAT to loopback address"); | |
4115ded1 | 884 | goto tx_error; |
b3cdd2a7 JV |
885 | } |
886 | ||
887 | /* copy-on-write the packet before mangling it */ | |
888 | if (!skb_make_writable(skb, sizeof(struct ipv6hdr))) | |
4115ded1 | 889 | goto tx_error; |
b3cdd2a7 | 890 | |
d8d1f30b | 891 | if (skb_cow(skb, rt->dst.dev->hard_header_len)) |
4115ded1 | 892 | goto tx_error; |
b3cdd2a7 | 893 | |
b3cdd2a7 | 894 | /* mangle the packet */ |
4115ded1 | 895 | if (pp->dnat_handler && !pp->dnat_handler(skb, pp, cp, ipvsh)) |
b3cdd2a7 | 896 | goto tx_error; |
4e3fd7a0 | 897 | ipv6_hdr(skb)->daddr = cp->daddr.in6; |
fc604767 | 898 | |
b0e010c5 | 899 | IP_VS_DBG_PKT(10, AF_INET6, pp, skb, ipvsh->off, "After DNAT"); |
b3cdd2a7 JV |
900 | |
901 | /* FIXME: when application helper enlarges the packet and the length | |
902 | is larger than the MTU of outgoing device, there will be still | |
903 | MTU problem. */ | |
904 | ||
905 | /* Another hack: avoid icmp_send in ip_fragment */ | |
60ff7467 | 906 | skb->ignore_df = 1; |
b3cdd2a7 | 907 | |
b8abdf09 | 908 | rc = ip_vs_nat_send_or_cont(NFPROTO_IPV6, skb, cp, local); |
026ace06 | 909 | rcu_read_unlock(); |
b3cdd2a7 JV |
910 | |
911 | LeaveFunction(10); | |
b8abdf09 | 912 | return rc; |
b3cdd2a7 | 913 | |
b3cdd2a7 JV |
914 | tx_error: |
915 | LeaveFunction(10); | |
916 | kfree_skb(skb); | |
026ace06 | 917 | rcu_read_unlock(); |
b3cdd2a7 | 918 | return NF_STOLEN; |
b3cdd2a7 JV |
919 | } |
920 | #endif | |
921 | ||
8052ba29 AG |
922 | /* When forwarding a packet, we must ensure that we've got enough headroom |
923 | * for the encapsulation packet in the skb. This also gives us an | |
924 | * opportunity to figure out what the payload_len, dsfield, ttl, and df | |
925 | * values should be, so that we won't need to look at the old ip header | |
926 | * again | |
927 | */ | |
928 | static struct sk_buff * | |
929 | ip_vs_prepare_tunneled_skb(struct sk_buff *skb, int skb_af, | |
930 | unsigned int max_headroom, __u8 *next_protocol, | |
931 | __u32 *payload_len, __u8 *dsfield, __u8 *ttl, | |
932 | __be16 *df) | |
933 | { | |
934 | struct sk_buff *new_skb = NULL; | |
935 | struct iphdr *old_iph = NULL; | |
936 | #ifdef CONFIG_IP_VS_IPV6 | |
937 | struct ipv6hdr *old_ipv6h = NULL; | |
938 | #endif | |
939 | ||
71563f34 AG |
940 | ip_vs_drop_early_demux_sk(skb); |
941 | ||
8052ba29 AG |
942 | if (skb_headroom(skb) < max_headroom || skb_cloned(skb)) { |
943 | new_skb = skb_realloc_headroom(skb, max_headroom); | |
944 | if (!new_skb) | |
945 | goto error; | |
50656d9d CO |
946 | if (skb->sk) |
947 | skb_set_owner_w(new_skb, skb->sk); | |
8052ba29 AG |
948 | consume_skb(skb); |
949 | skb = new_skb; | |
950 | } | |
951 | ||
952 | #ifdef CONFIG_IP_VS_IPV6 | |
953 | if (skb_af == AF_INET6) { | |
954 | old_ipv6h = ipv6_hdr(skb); | |
955 | *next_protocol = IPPROTO_IPV6; | |
956 | if (payload_len) | |
957 | *payload_len = | |
958 | ntohs(old_ipv6h->payload_len) + | |
959 | sizeof(*old_ipv6h); | |
960 | *dsfield = ipv6_get_dsfield(old_ipv6h); | |
961 | *ttl = old_ipv6h->hop_limit; | |
962 | if (df) | |
963 | *df = 0; | |
964 | } else | |
965 | #endif | |
966 | { | |
967 | old_iph = ip_hdr(skb); | |
968 | /* Copy DF, reset fragment offset and MF */ | |
969 | if (df) | |
970 | *df = (old_iph->frag_off & htons(IP_DF)); | |
971 | *next_protocol = IPPROTO_IPIP; | |
972 | ||
973 | /* fix old IP header checksum */ | |
974 | ip_send_check(old_iph); | |
975 | *dsfield = ipv4_get_dsfield(old_iph); | |
976 | *ttl = old_iph->ttl; | |
977 | if (payload_len) | |
978 | *payload_len = ntohs(old_iph->tot_len); | |
979 | } | |
980 | ||
981 | return skb; | |
982 | error: | |
983 | kfree_skb(skb); | |
984 | return ERR_PTR(-ENOMEM); | |
985 | } | |
986 | ||
987 | static inline int __tun_gso_type_mask(int encaps_af, int orig_af) | |
988 | { | |
7e13318d TH |
989 | switch (encaps_af) { |
990 | case AF_INET: | |
991 | return SKB_GSO_IPXIP4; | |
992 | case AF_INET6: | |
993 | return SKB_GSO_IPXIP6; | |
994 | default: | |
995 | return 0; | |
8052ba29 | 996 | } |
8052ba29 | 997 | } |
1da177e4 LT |
998 | |
999 | /* | |
1000 | * IP Tunneling transmitter | |
1001 | * | |
1002 | * This function encapsulates the packet in a new IP packet, its | |
1003 | * destination will be set to cp->daddr. Most code of this function | |
1004 | * is taken from ipip.c. | |
1005 | * | |
1006 | * It is used in VS/TUN cluster. The load balancer selects a real | |
1007 | * server from a cluster based on a scheduling algorithm, | |
1008 | * encapsulates the request packet and forwards it to the selected | |
1009 | * server. For example, all real servers are configured with | |
1010 | * "ifconfig tunl0 <Virtual IP Address> up". When the server receives | |
1011 | * the encapsulated packet, it will decapsulate the packet, processe | |
1012 | * the request and return the response packets directly to the client | |
1013 | * without passing the load balancer. This can greatly increase the | |
1014 | * scalability of virtual server. | |
1015 | * | |
1016 | * Used for ANY protocol | |
1017 | */ | |
1018 | int | |
1019 | ip_vs_tunnel_xmit(struct sk_buff *skb, struct ip_vs_conn *cp, | |
d4383f04 | 1020 | struct ip_vs_protocol *pp, struct ip_vs_iphdr *ipvsh) |
1da177e4 | 1021 | { |
361c3f52 EB |
1022 | struct netns_ipvs *ipvs = cp->ipvs; |
1023 | struct net *net = ipvs->net; | |
1da177e4 | 1024 | struct rtable *rt; /* Route to the other host */ |
c92f5ca2 | 1025 | __be32 saddr; /* Source for tunnel */ |
1da177e4 | 1026 | struct net_device *tdev; /* Device to other host */ |
8052ba29 AG |
1027 | __u8 next_protocol = 0; |
1028 | __u8 dsfield = 0; | |
1029 | __u8 ttl = 0; | |
1030 | __be16 df = 0; | |
1031 | __be16 *dfp = NULL; | |
1da177e4 | 1032 | struct iphdr *iph; /* Our new IP header */ |
c2636b4d | 1033 | unsigned int max_headroom; /* The extra header space needed */ |
4115ded1 | 1034 | int ret, local; |
1da177e4 LT |
1035 | |
1036 | EnterFunction(10); | |
1037 | ||
026ace06 | 1038 | rcu_read_lock(); |
ecfe87b8 | 1039 | local = __ip_vs_get_out_rt(ipvs, cp->af, skb, cp->dest, cp->daddr.ip, |
4115ded1 JA |
1040 | IP_VS_RT_MODE_LOCAL | |
1041 | IP_VS_RT_MODE_NON_LOCAL | | |
1042 | IP_VS_RT_MODE_CONNECT | | |
c63e4de2 | 1043 | IP_VS_RT_MODE_TUNNEL, &saddr, ipvsh); |
4115ded1 JA |
1044 | if (local < 0) |
1045 | goto tx_error; | |
026ace06 JA |
1046 | if (local) { |
1047 | rcu_read_unlock(); | |
b8abdf09 | 1048 | return ip_vs_send_or_cont(NFPROTO_IPV4, skb, cp, 1); |
026ace06 | 1049 | } |
1da177e4 | 1050 | |
4115ded1 | 1051 | rt = skb_rtable(skb); |
d8d1f30b | 1052 | tdev = rt->dst.dev; |
1da177e4 | 1053 | |
1da177e4 LT |
1054 | /* |
1055 | * Okay, now see if we can stuff it in the buffer as-is. | |
1056 | */ | |
1057 | max_headroom = LL_RESERVED_SPACE(tdev) + sizeof(struct iphdr); | |
1058 | ||
8052ba29 AG |
1059 | /* We only care about the df field if sysctl_pmtu_disc(ipvs) is set */ |
1060 | dfp = sysctl_pmtu_disc(ipvs) ? &df : NULL; | |
1061 | skb = ip_vs_prepare_tunneled_skb(skb, cp->af, max_headroom, | |
1062 | &next_protocol, NULL, &dsfield, | |
1063 | &ttl, dfp); | |
1064 | if (IS_ERR(skb)) | |
1065 | goto tx_error; | |
1da177e4 | 1066 | |
aed069df | 1067 | if (iptunnel_handle_offloads(skb, __tun_gso_type_mask(AF_INET, cp->af))) |
ea1d5d77 JA |
1068 | goto tx_error; |
1069 | ||
1070 | skb->transport_header = skb->network_header; | |
1071 | ||
e2d1bca7 ACM |
1072 | skb_push(skb, sizeof(struct iphdr)); |
1073 | skb_reset_network_header(skb); | |
1da177e4 LT |
1074 | memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt)); |
1075 | ||
1da177e4 LT |
1076 | /* |
1077 | * Push down and install the IPIP header. | |
1078 | */ | |
eddc9ec5 | 1079 | iph = ip_hdr(skb); |
1da177e4 LT |
1080 | iph->version = 4; |
1081 | iph->ihl = sizeof(struct iphdr)>>2; | |
1082 | iph->frag_off = df; | |
8052ba29 AG |
1083 | iph->protocol = next_protocol; |
1084 | iph->tos = dsfield; | |
c92f5ca2 JA |
1085 | iph->daddr = cp->daddr.ip; |
1086 | iph->saddr = saddr; | |
8052ba29 | 1087 | iph->ttl = ttl; |
b6a7719a | 1088 | ip_select_ident(net, skb, NULL); |
1da177e4 LT |
1089 | |
1090 | /* Another hack: avoid icmp_send in ip_fragment */ | |
60ff7467 | 1091 | skb->ignore_df = 1; |
1da177e4 | 1092 | |
b8abdf09 | 1093 | ret = ip_vs_tunnel_xmit_prepare(skb, cp); |
f4bc17cd | 1094 | if (ret == NF_ACCEPT) |
33224b16 | 1095 | ip_local_out(net, skb->sk, skb); |
f4bc17cd JA |
1096 | else if (ret == NF_DROP) |
1097 | kfree_skb(skb); | |
026ace06 | 1098 | rcu_read_unlock(); |
1da177e4 LT |
1099 | |
1100 | LeaveFunction(10); | |
1101 | ||
1102 | return NF_STOLEN; | |
1103 | ||
1da177e4 | 1104 | tx_error: |
ea1d5d77 JA |
1105 | if (!IS_ERR(skb)) |
1106 | kfree_skb(skb); | |
026ace06 | 1107 | rcu_read_unlock(); |
1da177e4 LT |
1108 | LeaveFunction(10); |
1109 | return NF_STOLEN; | |
1110 | } | |
1111 | ||
b3cdd2a7 JV |
1112 | #ifdef CONFIG_IP_VS_IPV6 |
1113 | int | |
1114 | ip_vs_tunnel_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp, | |
d4383f04 | 1115 | struct ip_vs_protocol *pp, struct ip_vs_iphdr *ipvsh) |
b3cdd2a7 JV |
1116 | { |
1117 | struct rt6_info *rt; /* Route to the other host */ | |
714f095f | 1118 | struct in6_addr saddr; /* Source for tunnel */ |
b3cdd2a7 | 1119 | struct net_device *tdev; /* Device to other host */ |
8052ba29 AG |
1120 | __u8 next_protocol = 0; |
1121 | __u32 payload_len = 0; | |
1122 | __u8 dsfield = 0; | |
1123 | __u8 ttl = 0; | |
b3cdd2a7 JV |
1124 | struct ipv6hdr *iph; /* Our new IP header */ |
1125 | unsigned int max_headroom; /* The extra header space needed */ | |
4115ded1 | 1126 | int ret, local; |
b3cdd2a7 JV |
1127 | |
1128 | EnterFunction(10); | |
1129 | ||
026ace06 | 1130 | rcu_read_lock(); |
f5745f8a EB |
1131 | local = __ip_vs_get_out_rt_v6(cp->ipvs, cp->af, skb, cp->dest, |
1132 | &cp->daddr.in6, | |
4115ded1 JA |
1133 | &saddr, ipvsh, 1, |
1134 | IP_VS_RT_MODE_LOCAL | | |
1135 | IP_VS_RT_MODE_NON_LOCAL | | |
1136 | IP_VS_RT_MODE_TUNNEL); | |
1137 | if (local < 0) | |
1138 | goto tx_error; | |
026ace06 JA |
1139 | if (local) { |
1140 | rcu_read_unlock(); | |
b8abdf09 | 1141 | return ip_vs_send_or_cont(NFPROTO_IPV6, skb, cp, 1); |
026ace06 | 1142 | } |
b3cdd2a7 | 1143 | |
4115ded1 | 1144 | rt = (struct rt6_info *) skb_dst(skb); |
d8d1f30b | 1145 | tdev = rt->dst.dev; |
b3cdd2a7 | 1146 | |
b3cdd2a7 JV |
1147 | /* |
1148 | * Okay, now see if we can stuff it in the buffer as-is. | |
1149 | */ | |
1150 | max_headroom = LL_RESERVED_SPACE(tdev) + sizeof(struct ipv6hdr); | |
1151 | ||
8052ba29 AG |
1152 | skb = ip_vs_prepare_tunneled_skb(skb, cp->af, max_headroom, |
1153 | &next_protocol, &payload_len, | |
1154 | &dsfield, &ttl, NULL); | |
1155 | if (IS_ERR(skb)) | |
1156 | goto tx_error; | |
b3cdd2a7 | 1157 | |
aed069df | 1158 | if (iptunnel_handle_offloads(skb, __tun_gso_type_mask(AF_INET6, cp->af))) |
ea1d5d77 JA |
1159 | goto tx_error; |
1160 | ||
714f095f | 1161 | skb->transport_header = skb->network_header; |
b3cdd2a7 JV |
1162 | |
1163 | skb_push(skb, sizeof(struct ipv6hdr)); | |
1164 | skb_reset_network_header(skb); | |
1165 | memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt)); | |
1166 | ||
b3cdd2a7 JV |
1167 | /* |
1168 | * Push down and install the IPIP header. | |
1169 | */ | |
1170 | iph = ipv6_hdr(skb); | |
1171 | iph->version = 6; | |
8052ba29 AG |
1172 | iph->nexthdr = next_protocol; |
1173 | iph->payload_len = htons(payload_len); | |
b3cdd2a7 | 1174 | memset(&iph->flow_lbl, 0, sizeof(iph->flow_lbl)); |
8052ba29 | 1175 | ipv6_change_dsfield(iph, 0, dsfield); |
4e3fd7a0 AD |
1176 | iph->daddr = cp->daddr.in6; |
1177 | iph->saddr = saddr; | |
8052ba29 | 1178 | iph->hop_limit = ttl; |
b3cdd2a7 JV |
1179 | |
1180 | /* Another hack: avoid icmp_send in ip_fragment */ | |
60ff7467 | 1181 | skb->ignore_df = 1; |
b3cdd2a7 | 1182 | |
b8abdf09 | 1183 | ret = ip_vs_tunnel_xmit_prepare(skb, cp); |
f4bc17cd | 1184 | if (ret == NF_ACCEPT) |
33224b16 | 1185 | ip6_local_out(cp->ipvs->net, skb->sk, skb); |
f4bc17cd JA |
1186 | else if (ret == NF_DROP) |
1187 | kfree_skb(skb); | |
026ace06 | 1188 | rcu_read_unlock(); |
b3cdd2a7 JV |
1189 | |
1190 | LeaveFunction(10); | |
1191 | ||
1192 | return NF_STOLEN; | |
1193 | ||
b3cdd2a7 | 1194 | tx_error: |
ea1d5d77 JA |
1195 | if (!IS_ERR(skb)) |
1196 | kfree_skb(skb); | |
026ace06 | 1197 | rcu_read_unlock(); |
b3cdd2a7 JV |
1198 | LeaveFunction(10); |
1199 | return NF_STOLEN; | |
1200 | } | |
1201 | #endif | |
1202 | ||
1da177e4 LT |
1203 | |
1204 | /* | |
1205 | * Direct Routing transmitter | |
1206 | * Used for ANY protocol | |
1207 | */ | |
1208 | int | |
1209 | ip_vs_dr_xmit(struct sk_buff *skb, struct ip_vs_conn *cp, | |
d4383f04 | 1210 | struct ip_vs_protocol *pp, struct ip_vs_iphdr *ipvsh) |
1da177e4 | 1211 | { |
4115ded1 | 1212 | int local; |
1da177e4 LT |
1213 | |
1214 | EnterFunction(10); | |
1215 | ||
026ace06 | 1216 | rcu_read_lock(); |
ecfe87b8 | 1217 | local = __ip_vs_get_out_rt(cp->ipvs, cp->af, skb, cp->dest, cp->daddr.ip, |
4115ded1 JA |
1218 | IP_VS_RT_MODE_LOCAL | |
1219 | IP_VS_RT_MODE_NON_LOCAL | | |
c63e4de2 | 1220 | IP_VS_RT_MODE_KNOWN_NH, NULL, ipvsh); |
4115ded1 | 1221 | if (local < 0) |
1da177e4 | 1222 | goto tx_error; |
026ace06 JA |
1223 | if (local) { |
1224 | rcu_read_unlock(); | |
4115ded1 | 1225 | return ip_vs_send_or_cont(NFPROTO_IPV4, skb, cp, 1); |
026ace06 | 1226 | } |
1da177e4 | 1227 | |
eddc9ec5 | 1228 | ip_send_check(ip_hdr(skb)); |
1da177e4 | 1229 | |
1da177e4 | 1230 | /* Another hack: avoid icmp_send in ip_fragment */ |
60ff7467 | 1231 | skb->ignore_df = 1; |
1da177e4 | 1232 | |
b8abdf09 | 1233 | ip_vs_send_or_cont(NFPROTO_IPV4, skb, cp, 0); |
026ace06 | 1234 | rcu_read_unlock(); |
1da177e4 LT |
1235 | |
1236 | LeaveFunction(10); | |
1237 | return NF_STOLEN; | |
1238 | ||
1da177e4 LT |
1239 | tx_error: |
1240 | kfree_skb(skb); | |
026ace06 | 1241 | rcu_read_unlock(); |
1da177e4 LT |
1242 | LeaveFunction(10); |
1243 | return NF_STOLEN; | |
1244 | } | |
1245 | ||
b3cdd2a7 JV |
1246 | #ifdef CONFIG_IP_VS_IPV6 |
1247 | int | |
1248 | ip_vs_dr_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp, | |
4115ded1 | 1249 | struct ip_vs_protocol *pp, struct ip_vs_iphdr *ipvsh) |
b3cdd2a7 | 1250 | { |
4115ded1 | 1251 | int local; |
b3cdd2a7 JV |
1252 | |
1253 | EnterFunction(10); | |
1254 | ||
026ace06 | 1255 | rcu_read_lock(); |
f5745f8a EB |
1256 | local = __ip_vs_get_out_rt_v6(cp->ipvs, cp->af, skb, cp->dest, |
1257 | &cp->daddr.in6, | |
4a4739d5 | 1258 | NULL, ipvsh, 0, |
4115ded1 | 1259 | IP_VS_RT_MODE_LOCAL | |
48e8aa6e MKL |
1260 | IP_VS_RT_MODE_NON_LOCAL | |
1261 | IP_VS_RT_MODE_KNOWN_NH); | |
4115ded1 | 1262 | if (local < 0) |
b3cdd2a7 | 1263 | goto tx_error; |
026ace06 JA |
1264 | if (local) { |
1265 | rcu_read_unlock(); | |
4115ded1 | 1266 | return ip_vs_send_or_cont(NFPROTO_IPV6, skb, cp, 1); |
026ace06 | 1267 | } |
b3cdd2a7 JV |
1268 | |
1269 | /* Another hack: avoid icmp_send in ip_fragment */ | |
60ff7467 | 1270 | skb->ignore_df = 1; |
b3cdd2a7 | 1271 | |
b8abdf09 | 1272 | ip_vs_send_or_cont(NFPROTO_IPV6, skb, cp, 0); |
026ace06 | 1273 | rcu_read_unlock(); |
b3cdd2a7 JV |
1274 | |
1275 | LeaveFunction(10); | |
1276 | return NF_STOLEN; | |
1277 | ||
b3cdd2a7 JV |
1278 | tx_error: |
1279 | kfree_skb(skb); | |
026ace06 | 1280 | rcu_read_unlock(); |
b3cdd2a7 JV |
1281 | LeaveFunction(10); |
1282 | return NF_STOLEN; | |
1283 | } | |
1284 | #endif | |
1285 | ||
1da177e4 LT |
1286 | |
1287 | /* | |
1288 | * ICMP packet transmitter | |
1289 | * called by the ip_vs_in_icmp | |
1290 | */ | |
1291 | int | |
1292 | ip_vs_icmp_xmit(struct sk_buff *skb, struct ip_vs_conn *cp, | |
d4383f04 JDB |
1293 | struct ip_vs_protocol *pp, int offset, unsigned int hooknum, |
1294 | struct ip_vs_iphdr *iph) | |
1da177e4 LT |
1295 | { |
1296 | struct rtable *rt; /* Route to the other host */ | |
1da177e4 | 1297 | int rc; |
fc604767 | 1298 | int local; |
4115ded1 | 1299 | int rt_mode, was_input; |
1da177e4 LT |
1300 | |
1301 | EnterFunction(10); | |
1302 | ||
1303 | /* The ICMP packet for VS/TUN, VS/DR and LOCALNODE will be | |
1304 | forwarded directly here, because there is no need to | |
1305 | translate address/port back */ | |
1306 | if (IP_VS_FWD_METHOD(cp) != IP_VS_CONN_F_MASQ) { | |
1307 | if (cp->packet_xmit) | |
d4383f04 | 1308 | rc = cp->packet_xmit(skb, cp, pp, iph); |
1da177e4 LT |
1309 | else |
1310 | rc = NF_ACCEPT; | |
1311 | /* do not touch skb anymore */ | |
1312 | atomic_inc(&cp->in_pkts); | |
1da177e4 LT |
1313 | goto out; |
1314 | } | |
1315 | ||
1316 | /* | |
1317 | * mangle and send the packet here (only for VS/NAT) | |
1318 | */ | |
4115ded1 | 1319 | was_input = rt_is_input_route(skb_rtable(skb)); |
1da177e4 | 1320 | |
c92f5ca2 JA |
1321 | /* LOCALNODE from FORWARD hook is not supported */ |
1322 | rt_mode = (hooknum != NF_INET_FORWARD) ? | |
1323 | IP_VS_RT_MODE_LOCAL | IP_VS_RT_MODE_NON_LOCAL | | |
1324 | IP_VS_RT_MODE_RDR : IP_VS_RT_MODE_NON_LOCAL; | |
026ace06 | 1325 | rcu_read_lock(); |
ecfe87b8 | 1326 | local = __ip_vs_get_out_rt(cp->ipvs, cp->af, skb, cp->dest, cp->daddr.ip, rt_mode, |
c63e4de2 | 1327 | NULL, iph); |
4115ded1 JA |
1328 | if (local < 0) |
1329 | goto tx_error; | |
1330 | rt = skb_rtable(skb); | |
fc604767 JA |
1331 | |
1332 | /* | |
1333 | * Avoid duplicate tuple in reply direction for NAT traffic | |
1334 | * to local address when connection is sync-ed | |
1335 | */ | |
c0cd1156 | 1336 | #if IS_ENABLED(CONFIG_NF_CONNTRACK) |
fc604767 JA |
1337 | if (cp->flags & IP_VS_CONN_F_SYNC && local) { |
1338 | enum ip_conntrack_info ctinfo; | |
05b4b065 | 1339 | struct nf_conn *ct = nf_ct_get(skb, &ctinfo); |
fc604767 | 1340 | |
ab8bc7ed | 1341 | if (ct) { |
fc604767 JA |
1342 | IP_VS_DBG(10, "%s(): " |
1343 | "stopping DNAT to local address %pI4\n", | |
1344 | __func__, &cp->daddr.ip); | |
4115ded1 | 1345 | goto tx_error; |
fc604767 JA |
1346 | } |
1347 | } | |
1348 | #endif | |
1349 | ||
1350 | /* From world but DNAT to loopback address? */ | |
4115ded1 | 1351 | if (local && ipv4_is_loopback(cp->daddr.ip) && was_input) { |
fc604767 JA |
1352 | IP_VS_DBG(1, "%s(): " |
1353 | "stopping DNAT to loopback %pI4\n", | |
1354 | __func__, &cp->daddr.ip); | |
4115ded1 | 1355 | goto tx_error; |
1da177e4 LT |
1356 | } |
1357 | ||
1358 | /* copy-on-write the packet before mangling it */ | |
af1e1cf0 | 1359 | if (!skb_make_writable(skb, offset)) |
4115ded1 | 1360 | goto tx_error; |
1da177e4 | 1361 | |
d8d1f30b | 1362 | if (skb_cow(skb, rt->dst.dev->hard_header_len)) |
4115ded1 | 1363 | goto tx_error; |
1da177e4 | 1364 | |
1da177e4 LT |
1365 | ip_vs_nat_icmp(skb, pp, cp, 0); |
1366 | ||
1367 | /* Another hack: avoid icmp_send in ip_fragment */ | |
60ff7467 | 1368 | skb->ignore_df = 1; |
1da177e4 | 1369 | |
b8abdf09 | 1370 | rc = ip_vs_nat_send_or_cont(NFPROTO_IPV4, skb, cp, local); |
026ace06 | 1371 | rcu_read_unlock(); |
1da177e4 LT |
1372 | goto out; |
1373 | ||
1da177e4 | 1374 | tx_error: |
026ace06 JA |
1375 | kfree_skb(skb); |
1376 | rcu_read_unlock(); | |
1da177e4 LT |
1377 | rc = NF_STOLEN; |
1378 | out: | |
1379 | LeaveFunction(10); | |
1380 | return rc; | |
1da177e4 | 1381 | } |
b3cdd2a7 JV |
1382 | |
1383 | #ifdef CONFIG_IP_VS_IPV6 | |
1384 | int | |
1385 | ip_vs_icmp_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp, | |
d4383f04 | 1386 | struct ip_vs_protocol *pp, int offset, unsigned int hooknum, |
4115ded1 | 1387 | struct ip_vs_iphdr *ipvsh) |
b3cdd2a7 JV |
1388 | { |
1389 | struct rt6_info *rt; /* Route to the other host */ | |
b3cdd2a7 | 1390 | int rc; |
fc604767 | 1391 | int local; |
c92f5ca2 | 1392 | int rt_mode; |
b3cdd2a7 JV |
1393 | |
1394 | EnterFunction(10); | |
1395 | ||
1396 | /* The ICMP packet for VS/TUN, VS/DR and LOCALNODE will be | |
1397 | forwarded directly here, because there is no need to | |
1398 | translate address/port back */ | |
1399 | if (IP_VS_FWD_METHOD(cp) != IP_VS_CONN_F_MASQ) { | |
1400 | if (cp->packet_xmit) | |
4115ded1 | 1401 | rc = cp->packet_xmit(skb, cp, pp, ipvsh); |
b3cdd2a7 JV |
1402 | else |
1403 | rc = NF_ACCEPT; | |
1404 | /* do not touch skb anymore */ | |
1405 | atomic_inc(&cp->in_pkts); | |
1406 | goto out; | |
1407 | } | |
1408 | ||
1409 | /* | |
1410 | * mangle and send the packet here (only for VS/NAT) | |
1411 | */ | |
1412 | ||
c92f5ca2 JA |
1413 | /* LOCALNODE from FORWARD hook is not supported */ |
1414 | rt_mode = (hooknum != NF_INET_FORWARD) ? | |
1415 | IP_VS_RT_MODE_LOCAL | IP_VS_RT_MODE_NON_LOCAL | | |
1416 | IP_VS_RT_MODE_RDR : IP_VS_RT_MODE_NON_LOCAL; | |
026ace06 | 1417 | rcu_read_lock(); |
f5745f8a EB |
1418 | local = __ip_vs_get_out_rt_v6(cp->ipvs, cp->af, skb, cp->dest, |
1419 | &cp->daddr.in6, NULL, ipvsh, 0, rt_mode); | |
4115ded1 JA |
1420 | if (local < 0) |
1421 | goto tx_error; | |
1422 | rt = (struct rt6_info *) skb_dst(skb); | |
fc604767 JA |
1423 | /* |
1424 | * Avoid duplicate tuple in reply direction for NAT traffic | |
1425 | * to local address when connection is sync-ed | |
1426 | */ | |
c0cd1156 | 1427 | #if IS_ENABLED(CONFIG_NF_CONNTRACK) |
fc604767 JA |
1428 | if (cp->flags & IP_VS_CONN_F_SYNC && local) { |
1429 | enum ip_conntrack_info ctinfo; | |
05b4b065 | 1430 | struct nf_conn *ct = nf_ct_get(skb, &ctinfo); |
fc604767 | 1431 | |
ab8bc7ed | 1432 | if (ct) { |
fc604767 JA |
1433 | IP_VS_DBG(10, "%s(): " |
1434 | "stopping DNAT to local address %pI6\n", | |
1435 | __func__, &cp->daddr.in6); | |
4115ded1 | 1436 | goto tx_error; |
fc604767 JA |
1437 | } |
1438 | } | |
1439 | #endif | |
1440 | ||
1441 | /* From world but DNAT to loopback address? */ | |
1442 | if (local && skb->dev && !(skb->dev->flags & IFF_LOOPBACK) && | |
fd0273d7 | 1443 | ipv6_addr_type(&cp->daddr.in6) & IPV6_ADDR_LOOPBACK) { |
fc604767 JA |
1444 | IP_VS_DBG(1, "%s(): " |
1445 | "stopping DNAT to loopback %pI6\n", | |
1446 | __func__, &cp->daddr.in6); | |
4115ded1 | 1447 | goto tx_error; |
b3cdd2a7 JV |
1448 | } |
1449 | ||
1450 | /* copy-on-write the packet before mangling it */ | |
1451 | if (!skb_make_writable(skb, offset)) | |
4115ded1 | 1452 | goto tx_error; |
b3cdd2a7 | 1453 | |
d8d1f30b | 1454 | if (skb_cow(skb, rt->dst.dev->hard_header_len)) |
4115ded1 | 1455 | goto tx_error; |
b3cdd2a7 | 1456 | |
b3cdd2a7 JV |
1457 | ip_vs_nat_icmp_v6(skb, pp, cp, 0); |
1458 | ||
1459 | /* Another hack: avoid icmp_send in ip_fragment */ | |
60ff7467 | 1460 | skb->ignore_df = 1; |
b3cdd2a7 | 1461 | |
b8abdf09 | 1462 | rc = ip_vs_nat_send_or_cont(NFPROTO_IPV6, skb, cp, local); |
026ace06 | 1463 | rcu_read_unlock(); |
b3cdd2a7 JV |
1464 | goto out; |
1465 | ||
b3cdd2a7 | 1466 | tx_error: |
026ace06 JA |
1467 | kfree_skb(skb); |
1468 | rcu_read_unlock(); | |
b3cdd2a7 JV |
1469 | rc = NF_STOLEN; |
1470 | out: | |
1471 | LeaveFunction(10); | |
1472 | return rc; | |
b3cdd2a7 JV |
1473 | } |
1474 | #endif |