]>
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; | |
133 | fl4.saddr = (rt_mode & IP_VS_RT_MODE_CONNECT) ? *saddr : 0; | |
ad4d3ef8 JA |
134 | fl4.flowi4_flags = (rt_mode & IP_VS_RT_MODE_KNOWN_NH) ? |
135 | FLOWI_FLAG_KNOWN_NH : 0; | |
f2edb9f7 JA |
136 | |
137 | retry: | |
138 | rt = ip_route_output_key(net, &fl4); | |
139 | if (IS_ERR(rt)) { | |
140 | /* Invalid saddr ? */ | |
141 | if (PTR_ERR(rt) == -EINVAL && *saddr && | |
142 | rt_mode & IP_VS_RT_MODE_CONNECT && !loop) { | |
143 | *saddr = 0; | |
c90558da | 144 | flowi4_update_output(&fl4, 0, 0, daddr, 0); |
f2edb9f7 JA |
145 | goto retry; |
146 | } | |
147 | IP_VS_DBG_RL("ip_route_output error, dest: %pI4\n", &daddr); | |
148 | return NULL; | |
149 | } else if (!*saddr && rt_mode & IP_VS_RT_MODE_CONNECT && fl4.saddr) { | |
150 | ip_rt_put(rt); | |
151 | *saddr = fl4.saddr; | |
c90558da | 152 | flowi4_update_output(&fl4, 0, 0, daddr, fl4.saddr); |
f2edb9f7 JA |
153 | loop++; |
154 | goto retry; | |
155 | } | |
156 | *saddr = fl4.saddr; | |
157 | return rt; | |
158 | } | |
159 | ||
4a4739d5 AG |
160 | #ifdef CONFIG_IP_VS_IPV6 |
161 | static inline int __ip_vs_is_local_route6(struct rt6_info *rt) | |
162 | { | |
163 | return rt->dst.dev && rt->dst.dev->flags & IFF_LOOPBACK; | |
164 | } | |
165 | #endif | |
166 | ||
167 | static inline bool crosses_local_route_boundary(int skb_af, struct sk_buff *skb, | |
168 | int rt_mode, | |
169 | bool new_rt_is_local) | |
170 | { | |
171 | bool rt_mode_allow_local = !!(rt_mode & IP_VS_RT_MODE_LOCAL); | |
172 | bool rt_mode_allow_non_local = !!(rt_mode & IP_VS_RT_MODE_LOCAL); | |
173 | bool rt_mode_allow_redirect = !!(rt_mode & IP_VS_RT_MODE_RDR); | |
174 | bool source_is_loopback; | |
175 | bool old_rt_is_local; | |
176 | ||
177 | #ifdef CONFIG_IP_VS_IPV6 | |
178 | if (skb_af == AF_INET6) { | |
179 | int addr_type = ipv6_addr_type(&ipv6_hdr(skb)->saddr); | |
180 | ||
181 | source_is_loopback = | |
182 | (!skb->dev || skb->dev->flags & IFF_LOOPBACK) && | |
183 | (addr_type & IPV6_ADDR_LOOPBACK); | |
184 | old_rt_is_local = __ip_vs_is_local_route6( | |
185 | (struct rt6_info *)skb_dst(skb)); | |
186 | } else | |
187 | #endif | |
188 | { | |
189 | source_is_loopback = ipv4_is_loopback(ip_hdr(skb)->saddr); | |
190 | old_rt_is_local = skb_rtable(skb)->rt_flags & RTCF_LOCAL; | |
191 | } | |
192 | ||
193 | if (unlikely(new_rt_is_local)) { | |
194 | if (!rt_mode_allow_local) | |
195 | return true; | |
196 | if (!rt_mode_allow_redirect && !old_rt_is_local) | |
197 | return true; | |
198 | } else { | |
199 | if (!rt_mode_allow_non_local) | |
200 | return true; | |
201 | if (source_is_loopback) | |
202 | return true; | |
203 | } | |
204 | return false; | |
205 | } | |
206 | ||
919aa0b2 AG |
207 | static inline void maybe_update_pmtu(int skb_af, struct sk_buff *skb, int mtu) |
208 | { | |
209 | struct sock *sk = skb->sk; | |
210 | struct rtable *ort = skb_rtable(skb); | |
211 | ||
212 | if (!skb->dev && sk && sk->sk_state != TCP_TIME_WAIT) | |
213 | ort->dst.ops->update_pmtu(&ort->dst, sk, NULL, mtu); | |
214 | } | |
215 | ||
c63e4de2 AG |
216 | static inline bool ensure_mtu_is_adequate(int skb_af, int rt_mode, |
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) { | |
222 | struct net *net = dev_net(skb_dst(skb)->dev); | |
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 */ | |
228 | if (!ipvsh->fragoffs) | |
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 | { | |
237 | struct netns_ipvs *ipvs = net_ipvs(skb_net(skb)); | |
238 | ||
239 | /* If we're going to tunnel the packet and pmtu discovery | |
240 | * is disabled, we'll just fragment it anyway | |
241 | */ | |
242 | if ((rt_mode & IP_VS_RT_MODE_TUNNEL) && !sysctl_pmtu_disc(ipvs)) | |
243 | return true; | |
244 | ||
245 | if (unlikely(ip_hdr(skb)->frag_off & htons(IP_DF) && | |
246 | skb->len > mtu && !skb_is_gso(skb))) { | |
247 | icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED, | |
248 | htonl(mtu)); | |
249 | IP_VS_DBG(1, "frag needed for %pI4\n", | |
250 | &ip_hdr(skb)->saddr); | |
251 | return false; | |
252 | } | |
253 | } | |
254 | ||
255 | return true; | |
256 | } | |
257 | ||
17a8f8e3 | 258 | /* Get route to destination or remote server */ |
4115ded1 | 259 | static int |
4a4739d5 | 260 | __ip_vs_get_out_rt(int skb_af, struct sk_buff *skb, struct ip_vs_dest *dest, |
c63e4de2 AG |
261 | __be32 daddr, int rt_mode, __be32 *ret_saddr, |
262 | struct ip_vs_iphdr *ipvsh) | |
1da177e4 | 263 | { |
fc604767 | 264 | struct net *net = dev_net(skb_dst(skb)->dev); |
026ace06 | 265 | struct ip_vs_dest_dst *dest_dst; |
1da177e4 | 266 | struct rtable *rt; /* Route to the other host */ |
4115ded1 | 267 | int mtu; |
026ace06 | 268 | int local, noref = 1; |
1da177e4 LT |
269 | |
270 | if (dest) { | |
026ace06 JA |
271 | dest_dst = __ip_vs_dst_check(dest); |
272 | if (likely(dest_dst)) | |
273 | rt = (struct rtable *) dest_dst->dst_cache; | |
274 | else { | |
275 | dest_dst = ip_vs_dest_dst_alloc(); | |
ac69269a | 276 | spin_lock_bh(&dest->dst_lock); |
026ace06 JA |
277 | if (!dest_dst) { |
278 | __ip_vs_dst_set(dest, NULL, NULL, 0); | |
ac69269a | 279 | spin_unlock_bh(&dest->dst_lock); |
026ace06 JA |
280 | goto err_unreach; |
281 | } | |
c90558da | 282 | rt = do_output_route4(net, dest->addr.ip, rt_mode, |
026ace06 | 283 | &dest_dst->dst_saddr.ip); |
f2edb9f7 | 284 | if (!rt) { |
026ace06 | 285 | __ip_vs_dst_set(dest, NULL, NULL, 0); |
ac69269a | 286 | spin_unlock_bh(&dest->dst_lock); |
026ace06 | 287 | ip_vs_dest_dst_free(dest_dst); |
4115ded1 | 288 | goto err_unreach; |
1da177e4 | 289 | } |
026ace06 | 290 | __ip_vs_dst_set(dest, dest_dst, &rt->dst, 0); |
ac69269a | 291 | spin_unlock_bh(&dest->dst_lock); |
c90558da | 292 | IP_VS_DBG(10, "new dst %pI4, src %pI4, refcnt=%d\n", |
026ace06 | 293 | &dest->addr.ip, &dest_dst->dst_saddr.ip, |
c90558da | 294 | atomic_read(&rt->dst.__refcnt)); |
1da177e4 | 295 | } |
44e3125c | 296 | daddr = dest->addr.ip; |
c92f5ca2 | 297 | if (ret_saddr) |
026ace06 | 298 | *ret_saddr = dest_dst->dst_saddr.ip; |
1da177e4 | 299 | } else { |
f2edb9f7 | 300 | __be32 saddr = htonl(INADDR_ANY); |
c92f5ca2 | 301 | |
026ace06 JA |
302 | noref = 0; |
303 | ||
f2edb9f7 JA |
304 | /* For such unconfigured boxes avoid many route lookups |
305 | * for performance reasons because we do not remember saddr | |
306 | */ | |
307 | rt_mode &= ~IP_VS_RT_MODE_CONNECT; | |
c90558da | 308 | rt = do_output_route4(net, daddr, rt_mode, &saddr); |
f2edb9f7 | 309 | if (!rt) |
4115ded1 | 310 | goto err_unreach; |
c92f5ca2 | 311 | if (ret_saddr) |
f2edb9f7 | 312 | *ret_saddr = saddr; |
1da177e4 LT |
313 | } |
314 | ||
4115ded1 | 315 | local = (rt->rt_flags & RTCF_LOCAL) ? 1 : 0; |
4a4739d5 AG |
316 | if (unlikely(crosses_local_route_boundary(skb_af, skb, rt_mode, |
317 | local))) { | |
318 | IP_VS_DBG_RL("We are crossing local and non-local addresses" | |
3d53666b | 319 | " daddr=%pI4\n", &daddr); |
4115ded1 | 320 | goto err_put; |
fc604767 | 321 | } |
4a4739d5 AG |
322 | |
323 | if (unlikely(local)) { | |
4115ded1 | 324 | /* skb to local stack, preserve old route */ |
026ace06 JA |
325 | if (!noref) |
326 | ip_rt_put(rt); | |
4115ded1 | 327 | return local; |
fc604767 | 328 | } |
4115ded1 JA |
329 | |
330 | if (likely(!(rt_mode & IP_VS_RT_MODE_TUNNEL))) { | |
331 | mtu = dst_mtu(&rt->dst); | |
4115ded1 | 332 | } else { |
4115ded1 JA |
333 | mtu = dst_mtu(&rt->dst) - sizeof(struct iphdr); |
334 | if (mtu < 68) { | |
335 | IP_VS_DBG_RL("%s(): mtu less than 68\n", __func__); | |
336 | goto err_put; | |
337 | } | |
919aa0b2 | 338 | maybe_update_pmtu(skb_af, skb, mtu); |
fc604767 JA |
339 | } |
340 | ||
c63e4de2 | 341 | if (!ensure_mtu_is_adequate(skb_af, rt_mode, ipvsh, skb, mtu)) |
4115ded1 | 342 | goto err_put; |
4115ded1 JA |
343 | |
344 | skb_dst_drop(skb); | |
026ace06 JA |
345 | if (noref) { |
346 | if (!local) | |
347 | skb_dst_set_noref_force(skb, &rt->dst); | |
348 | else | |
349 | skb_dst_set(skb, dst_clone(&rt->dst)); | |
350 | } else | |
351 | skb_dst_set(skb, &rt->dst); | |
4115ded1 JA |
352 | |
353 | return local; | |
354 | ||
355 | err_put: | |
026ace06 JA |
356 | if (!noref) |
357 | ip_rt_put(rt); | |
4115ded1 JA |
358 | return -1; |
359 | ||
360 | err_unreach: | |
361 | dst_link_failure(skb); | |
362 | return -1; | |
1da177e4 LT |
363 | } |
364 | ||
38cdcc9a | 365 | #ifdef CONFIG_IP_VS_IPV6 |
714f095f HS |
366 | static struct dst_entry * |
367 | __ip_vs_route_output_v6(struct net *net, struct in6_addr *daddr, | |
368 | struct in6_addr *ret_saddr, int do_xfrm) | |
369 | { | |
370 | struct dst_entry *dst; | |
4c9483b2 DM |
371 | struct flowi6 fl6 = { |
372 | .daddr = *daddr, | |
714f095f HS |
373 | }; |
374 | ||
4c9483b2 | 375 | dst = ip6_route_output(net, NULL, &fl6); |
714f095f HS |
376 | if (dst->error) |
377 | goto out_err; | |
378 | if (!ret_saddr) | |
379 | return dst; | |
4c9483b2 | 380 | if (ipv6_addr_any(&fl6.saddr) && |
714f095f | 381 | ipv6_dev_get_saddr(net, ip6_dst_idev(dst)->dev, |
4c9483b2 | 382 | &fl6.daddr, 0, &fl6.saddr) < 0) |
714f095f | 383 | goto out_err; |
452edd59 | 384 | if (do_xfrm) { |
4c9483b2 | 385 | dst = xfrm_lookup(net, dst, flowi6_to_flowi(&fl6), NULL, 0); |
452edd59 DM |
386 | if (IS_ERR(dst)) { |
387 | dst = NULL; | |
388 | goto out_err; | |
389 | } | |
390 | } | |
4e3fd7a0 | 391 | *ret_saddr = fl6.saddr; |
714f095f HS |
392 | return dst; |
393 | ||
394 | out_err: | |
395 | dst_release(dst); | |
396 | IP_VS_DBG_RL("ip6_route_output error, dest: %pI6\n", daddr); | |
397 | return NULL; | |
398 | } | |
399 | ||
fc604767 JA |
400 | /* |
401 | * Get route to destination or remote server | |
fc604767 | 402 | */ |
4115ded1 | 403 | static int |
4a4739d5 | 404 | __ip_vs_get_out_rt_v6(int skb_af, struct sk_buff *skb, struct ip_vs_dest *dest, |
fc604767 | 405 | struct in6_addr *daddr, struct in6_addr *ret_saddr, |
4115ded1 | 406 | struct ip_vs_iphdr *ipvsh, int do_xfrm, int rt_mode) |
38cdcc9a | 407 | { |
fc604767 | 408 | struct net *net = dev_net(skb_dst(skb)->dev); |
026ace06 | 409 | struct ip_vs_dest_dst *dest_dst; |
38cdcc9a | 410 | struct rt6_info *rt; /* Route to the other host */ |
714f095f | 411 | struct dst_entry *dst; |
4115ded1 | 412 | int mtu; |
026ace06 | 413 | int local, noref = 1; |
38cdcc9a JV |
414 | |
415 | if (dest) { | |
026ace06 JA |
416 | dest_dst = __ip_vs_dst_check(dest); |
417 | if (likely(dest_dst)) | |
418 | rt = (struct rt6_info *) dest_dst->dst_cache; | |
419 | else { | |
714f095f | 420 | u32 cookie; |
38cdcc9a | 421 | |
026ace06 | 422 | dest_dst = ip_vs_dest_dst_alloc(); |
ac69269a | 423 | spin_lock_bh(&dest->dst_lock); |
026ace06 JA |
424 | if (!dest_dst) { |
425 | __ip_vs_dst_set(dest, NULL, NULL, 0); | |
ac69269a | 426 | spin_unlock_bh(&dest->dst_lock); |
026ace06 JA |
427 | goto err_unreach; |
428 | } | |
714f095f | 429 | dst = __ip_vs_route_output_v6(net, &dest->addr.in6, |
026ace06 | 430 | &dest_dst->dst_saddr.in6, |
714f095f HS |
431 | do_xfrm); |
432 | if (!dst) { | |
026ace06 | 433 | __ip_vs_dst_set(dest, NULL, NULL, 0); |
ac69269a | 434 | spin_unlock_bh(&dest->dst_lock); |
026ace06 | 435 | ip_vs_dest_dst_free(dest_dst); |
4115ded1 | 436 | goto err_unreach; |
38cdcc9a | 437 | } |
714f095f HS |
438 | rt = (struct rt6_info *) dst; |
439 | cookie = rt->rt6i_node ? rt->rt6i_node->fn_sernum : 0; | |
026ace06 | 440 | __ip_vs_dst_set(dest, dest_dst, &rt->dst, cookie); |
ac69269a | 441 | spin_unlock_bh(&dest->dst_lock); |
714f095f | 442 | IP_VS_DBG(10, "new dst %pI6, src %pI6, refcnt=%d\n", |
026ace06 | 443 | &dest->addr.in6, &dest_dst->dst_saddr.in6, |
d8d1f30b | 444 | atomic_read(&rt->dst.__refcnt)); |
38cdcc9a | 445 | } |
714f095f | 446 | if (ret_saddr) |
026ace06 | 447 | *ret_saddr = dest_dst->dst_saddr.in6; |
38cdcc9a | 448 | } else { |
026ace06 | 449 | noref = 0; |
fc604767 | 450 | dst = __ip_vs_route_output_v6(net, daddr, ret_saddr, do_xfrm); |
714f095f | 451 | if (!dst) |
4115ded1 | 452 | goto err_unreach; |
714f095f | 453 | rt = (struct rt6_info *) dst; |
38cdcc9a JV |
454 | } |
455 | ||
fc604767 | 456 | local = __ip_vs_is_local_route6(rt); |
4a4739d5 AG |
457 | |
458 | if (unlikely(crosses_local_route_boundary(skb_af, skb, rt_mode, | |
459 | local))) { | |
460 | IP_VS_DBG_RL("We are crossing local and non-local addresses" | |
3d53666b | 461 | " daddr=%pI6\n", daddr); |
4115ded1 | 462 | goto err_put; |
fc604767 | 463 | } |
4a4739d5 AG |
464 | |
465 | if (unlikely(local)) { | |
4115ded1 | 466 | /* skb to local stack, preserve old route */ |
026ace06 JA |
467 | if (!noref) |
468 | dst_release(&rt->dst); | |
4115ded1 | 469 | return local; |
fc604767 | 470 | } |
4115ded1 JA |
471 | |
472 | /* MTU checking */ | |
473 | if (likely(!(rt_mode & IP_VS_RT_MODE_TUNNEL))) | |
474 | mtu = dst_mtu(&rt->dst); | |
475 | else { | |
4115ded1 JA |
476 | mtu = dst_mtu(&rt->dst) - sizeof(struct ipv6hdr); |
477 | if (mtu < IPV6_MIN_MTU) { | |
478 | IP_VS_DBG_RL("%s(): mtu less than %d\n", __func__, | |
479 | IPV6_MIN_MTU); | |
480 | goto err_put; | |
481 | } | |
919aa0b2 | 482 | maybe_update_pmtu(skb_af, skb, mtu); |
fc604767 JA |
483 | } |
484 | ||
c63e4de2 | 485 | if (!ensure_mtu_is_adequate(skb_af, rt_mode, ipvsh, skb, mtu)) |
4115ded1 | 486 | goto err_put; |
4115ded1 JA |
487 | |
488 | skb_dst_drop(skb); | |
026ace06 JA |
489 | if (noref) { |
490 | if (!local) | |
491 | skb_dst_set_noref_force(skb, &rt->dst); | |
492 | else | |
493 | skb_dst_set(skb, dst_clone(&rt->dst)); | |
494 | } else | |
495 | skb_dst_set(skb, &rt->dst); | |
4115ded1 JA |
496 | |
497 | return local; | |
498 | ||
499 | err_put: | |
026ace06 JA |
500 | if (!noref) |
501 | dst_release(&rt->dst); | |
4115ded1 JA |
502 | return -1; |
503 | ||
504 | err_unreach: | |
505 | dst_link_failure(skb); | |
506 | return -1; | |
38cdcc9a JV |
507 | } |
508 | #endif | |
509 | ||
1da177e4 | 510 | |
b8abdf09 JA |
511 | /* return NF_ACCEPT to allow forwarding or other NF_xxx on error */ |
512 | static inline int ip_vs_tunnel_xmit_prepare(struct sk_buff *skb, | |
513 | struct ip_vs_conn *cp) | |
514 | { | |
515 | int ret = NF_ACCEPT; | |
516 | ||
517 | skb->ipvs_property = 1; | |
518 | if (unlikely(cp->flags & IP_VS_CONN_F_NFCT)) | |
519 | ret = ip_vs_confirm_conntrack(skb); | |
520 | if (ret == NF_ACCEPT) { | |
521 | nf_reset(skb); | |
522 | skb_forward_csum(skb); | |
523 | } | |
524 | return ret; | |
525 | } | |
526 | ||
527 | /* return NF_STOLEN (sent) or NF_ACCEPT if local=1 (not sent) */ | |
528 | static inline int ip_vs_nat_send_or_cont(int pf, struct sk_buff *skb, | |
529 | struct ip_vs_conn *cp, int local) | |
530 | { | |
531 | int ret = NF_STOLEN; | |
532 | ||
533 | skb->ipvs_property = 1; | |
534 | if (likely(!(cp->flags & IP_VS_CONN_F_NFCT))) | |
535 | ip_vs_notrack(skb); | |
536 | else | |
537 | ip_vs_update_conntrack(skb, cp, 1); | |
538 | if (!local) { | |
539 | skb_forward_csum(skb); | |
540 | NF_HOOK(pf, NF_INET_LOCAL_OUT, skb, NULL, skb_dst(skb)->dev, | |
541 | dst_output); | |
542 | } else | |
543 | ret = NF_ACCEPT; | |
544 | return ret; | |
545 | } | |
546 | ||
547 | /* return NF_STOLEN (sent) or NF_ACCEPT if local=1 (not sent) */ | |
548 | static inline int ip_vs_send_or_cont(int pf, struct sk_buff *skb, | |
549 | struct ip_vs_conn *cp, int local) | |
550 | { | |
551 | int ret = NF_STOLEN; | |
552 | ||
553 | skb->ipvs_property = 1; | |
554 | if (likely(!(cp->flags & IP_VS_CONN_F_NFCT))) | |
555 | ip_vs_notrack(skb); | |
556 | if (!local) { | |
557 | skb_forward_csum(skb); | |
558 | NF_HOOK(pf, NF_INET_LOCAL_OUT, skb, NULL, skb_dst(skb)->dev, | |
559 | dst_output); | |
560 | } else | |
561 | ret = NF_ACCEPT; | |
562 | return ret; | |
563 | } | |
1da177e4 LT |
564 | |
565 | ||
566 | /* | |
567 | * NULL transmitter (do nothing except return NF_ACCEPT) | |
568 | */ | |
569 | int | |
570 | ip_vs_null_xmit(struct sk_buff *skb, struct ip_vs_conn *cp, | |
d4383f04 | 571 | struct ip_vs_protocol *pp, struct ip_vs_iphdr *ipvsh) |
1da177e4 LT |
572 | { |
573 | /* we do not touch skb and do not need pskb ptr */ | |
b8abdf09 | 574 | return ip_vs_send_or_cont(NFPROTO_IPV4, skb, cp, 1); |
1da177e4 LT |
575 | } |
576 | ||
577 | ||
578 | /* | |
579 | * Bypass transmitter | |
580 | * Let packets bypass the destination when the destination is not | |
581 | * available, it may be only used in transparent cache cluster. | |
582 | */ | |
583 | int | |
584 | ip_vs_bypass_xmit(struct sk_buff *skb, struct ip_vs_conn *cp, | |
d4383f04 | 585 | struct ip_vs_protocol *pp, struct ip_vs_iphdr *ipvsh) |
1da177e4 | 586 | { |
eddc9ec5 | 587 | struct iphdr *iph = ip_hdr(skb); |
1da177e4 LT |
588 | |
589 | EnterFunction(10); | |
590 | ||
026ace06 | 591 | rcu_read_lock(); |
4a4739d5 | 592 | if (__ip_vs_get_out_rt(cp->af, skb, NULL, iph->daddr, |
c63e4de2 | 593 | IP_VS_RT_MODE_NON_LOCAL, NULL, ipvsh) < 0) |
1da177e4 | 594 | goto tx_error; |
1da177e4 | 595 | |
4115ded1 | 596 | ip_send_check(iph); |
1da177e4 LT |
597 | |
598 | /* Another hack: avoid icmp_send in ip_fragment */ | |
60ff7467 | 599 | skb->ignore_df = 1; |
1da177e4 | 600 | |
b8abdf09 | 601 | ip_vs_send_or_cont(NFPROTO_IPV4, skb, cp, 0); |
026ace06 | 602 | rcu_read_unlock(); |
1da177e4 LT |
603 | |
604 | LeaveFunction(10); | |
605 | return NF_STOLEN; | |
606 | ||
1da177e4 LT |
607 | tx_error: |
608 | kfree_skb(skb); | |
026ace06 | 609 | rcu_read_unlock(); |
1da177e4 LT |
610 | LeaveFunction(10); |
611 | return NF_STOLEN; | |
612 | } | |
613 | ||
b3cdd2a7 JV |
614 | #ifdef CONFIG_IP_VS_IPV6 |
615 | int | |
616 | ip_vs_bypass_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp, | |
4115ded1 | 617 | struct ip_vs_protocol *pp, struct ip_vs_iphdr *ipvsh) |
b3cdd2a7 | 618 | { |
b3cdd2a7 JV |
619 | EnterFunction(10); |
620 | ||
026ace06 | 621 | rcu_read_lock(); |
4a4739d5 | 622 | if (__ip_vs_get_out_rt_v6(cp->af, skb, NULL, &ipvsh->daddr.in6, NULL, |
4115ded1 | 623 | ipvsh, 0, IP_VS_RT_MODE_NON_LOCAL) < 0) |
b3cdd2a7 | 624 | goto tx_error; |
b3cdd2a7 JV |
625 | |
626 | /* Another hack: avoid icmp_send in ip_fragment */ | |
60ff7467 | 627 | skb->ignore_df = 1; |
b3cdd2a7 | 628 | |
b8abdf09 | 629 | ip_vs_send_or_cont(NFPROTO_IPV6, skb, cp, 0); |
026ace06 | 630 | rcu_read_unlock(); |
b3cdd2a7 JV |
631 | |
632 | LeaveFunction(10); | |
633 | return NF_STOLEN; | |
634 | ||
b3cdd2a7 JV |
635 | tx_error: |
636 | kfree_skb(skb); | |
026ace06 | 637 | rcu_read_unlock(); |
b3cdd2a7 JV |
638 | LeaveFunction(10); |
639 | return NF_STOLEN; | |
640 | } | |
641 | #endif | |
1da177e4 LT |
642 | |
643 | /* | |
644 | * NAT transmitter (only for outside-to-inside nat forwarding) | |
645 | * Not used for related ICMP | |
646 | */ | |
647 | int | |
648 | ip_vs_nat_xmit(struct sk_buff *skb, struct ip_vs_conn *cp, | |
d4383f04 | 649 | struct ip_vs_protocol *pp, struct ip_vs_iphdr *ipvsh) |
1da177e4 LT |
650 | { |
651 | struct rtable *rt; /* Route to the other host */ | |
4115ded1 | 652 | int local, rc, was_input; |
1da177e4 LT |
653 | |
654 | EnterFunction(10); | |
655 | ||
026ace06 | 656 | rcu_read_lock(); |
1da177e4 LT |
657 | /* check if it is a connection of no-client-port */ |
658 | if (unlikely(cp->flags & IP_VS_CONN_F_NO_CPORT)) { | |
014d730d | 659 | __be16 _pt, *p; |
4115ded1 JA |
660 | |
661 | p = skb_header_pointer(skb, ipvsh->len, sizeof(_pt), &_pt); | |
1da177e4 LT |
662 | if (p == NULL) |
663 | goto tx_error; | |
664 | ip_vs_conn_fill_cport(cp, *p); | |
665 | IP_VS_DBG(10, "filled cport=%d\n", ntohs(*p)); | |
666 | } | |
667 | ||
4115ded1 | 668 | was_input = rt_is_input_route(skb_rtable(skb)); |
4a4739d5 | 669 | local = __ip_vs_get_out_rt(cp->af, skb, cp->dest, cp->daddr.ip, |
4115ded1 JA |
670 | IP_VS_RT_MODE_LOCAL | |
671 | IP_VS_RT_MODE_NON_LOCAL | | |
c63e4de2 | 672 | IP_VS_RT_MODE_RDR, NULL, ipvsh); |
4115ded1 JA |
673 | if (local < 0) |
674 | goto tx_error; | |
675 | rt = skb_rtable(skb); | |
fc604767 JA |
676 | /* |
677 | * Avoid duplicate tuple in reply direction for NAT traffic | |
678 | * to local address when connection is sync-ed | |
679 | */ | |
c0cd1156 | 680 | #if IS_ENABLED(CONFIG_NF_CONNTRACK) |
fc604767 JA |
681 | if (cp->flags & IP_VS_CONN_F_SYNC && local) { |
682 | enum ip_conntrack_info ctinfo; | |
05b4b065 | 683 | struct nf_conn *ct = nf_ct_get(skb, &ctinfo); |
fc604767 JA |
684 | |
685 | if (ct && !nf_ct_is_untracked(ct)) { | |
0d79641a JA |
686 | IP_VS_DBG_RL_PKT(10, AF_INET, pp, skb, 0, |
687 | "ip_vs_nat_xmit(): " | |
fc604767 | 688 | "stopping DNAT to local address"); |
4115ded1 | 689 | goto tx_error; |
fc604767 JA |
690 | } |
691 | } | |
692 | #endif | |
693 | ||
694 | /* From world but DNAT to loopback address? */ | |
4115ded1 | 695 | if (local && ipv4_is_loopback(cp->daddr.ip) && was_input) { |
0d79641a | 696 | IP_VS_DBG_RL_PKT(1, AF_INET, pp, skb, 0, "ip_vs_nat_xmit(): " |
fc604767 | 697 | "stopping DNAT to loopback address"); |
4115ded1 | 698 | goto tx_error; |
1da177e4 LT |
699 | } |
700 | ||
701 | /* copy-on-write the packet before mangling it */ | |
af1e1cf0 | 702 | if (!skb_make_writable(skb, sizeof(struct iphdr))) |
4115ded1 | 703 | goto tx_error; |
1da177e4 | 704 | |
d8d1f30b | 705 | if (skb_cow(skb, rt->dst.dev->hard_header_len)) |
4115ded1 | 706 | goto tx_error; |
1da177e4 | 707 | |
1da177e4 | 708 | /* mangle the packet */ |
d4383f04 | 709 | if (pp->dnat_handler && !pp->dnat_handler(skb, pp, cp, ipvsh)) |
4115ded1 | 710 | goto tx_error; |
e7ade46a | 711 | ip_hdr(skb)->daddr = cp->daddr.ip; |
eddc9ec5 | 712 | ip_send_check(ip_hdr(skb)); |
1da177e4 | 713 | |
0d79641a | 714 | IP_VS_DBG_PKT(10, AF_INET, pp, skb, 0, "After DNAT"); |
1da177e4 LT |
715 | |
716 | /* FIXME: when application helper enlarges the packet and the length | |
717 | is larger than the MTU of outgoing device, there will be still | |
718 | MTU problem. */ | |
719 | ||
720 | /* Another hack: avoid icmp_send in ip_fragment */ | |
60ff7467 | 721 | skb->ignore_df = 1; |
1da177e4 | 722 | |
b8abdf09 | 723 | rc = ip_vs_nat_send_or_cont(NFPROTO_IPV4, skb, cp, local); |
026ace06 | 724 | rcu_read_unlock(); |
1da177e4 LT |
725 | |
726 | LeaveFunction(10); | |
b8abdf09 | 727 | return rc; |
1da177e4 | 728 | |
1da177e4 | 729 | tx_error: |
1da177e4 | 730 | kfree_skb(skb); |
026ace06 | 731 | rcu_read_unlock(); |
f4bc17cd | 732 | LeaveFunction(10); |
1da177e4 | 733 | return NF_STOLEN; |
1da177e4 LT |
734 | } |
735 | ||
b3cdd2a7 JV |
736 | #ifdef CONFIG_IP_VS_IPV6 |
737 | int | |
738 | ip_vs_nat_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp, | |
4115ded1 | 739 | struct ip_vs_protocol *pp, struct ip_vs_iphdr *ipvsh) |
b3cdd2a7 JV |
740 | { |
741 | struct rt6_info *rt; /* Route to the other host */ | |
b8abdf09 | 742 | int local, rc; |
b3cdd2a7 JV |
743 | |
744 | EnterFunction(10); | |
745 | ||
026ace06 | 746 | rcu_read_lock(); |
b3cdd2a7 | 747 | /* check if it is a connection of no-client-port */ |
4115ded1 | 748 | if (unlikely(cp->flags & IP_VS_CONN_F_NO_CPORT && !ipvsh->fragoffs)) { |
b3cdd2a7 | 749 | __be16 _pt, *p; |
4115ded1 | 750 | p = skb_header_pointer(skb, ipvsh->len, sizeof(_pt), &_pt); |
b3cdd2a7 JV |
751 | if (p == NULL) |
752 | goto tx_error; | |
753 | ip_vs_conn_fill_cport(cp, *p); | |
754 | IP_VS_DBG(10, "filled cport=%d\n", ntohs(*p)); | |
755 | } | |
756 | ||
4a4739d5 AG |
757 | local = __ip_vs_get_out_rt_v6(cp->af, skb, cp->dest, &cp->daddr.in6, |
758 | NULL, ipvsh, 0, | |
4115ded1 JA |
759 | IP_VS_RT_MODE_LOCAL | |
760 | IP_VS_RT_MODE_NON_LOCAL | | |
761 | IP_VS_RT_MODE_RDR); | |
762 | if (local < 0) | |
763 | goto tx_error; | |
764 | rt = (struct rt6_info *) skb_dst(skb); | |
fc604767 JA |
765 | /* |
766 | * Avoid duplicate tuple in reply direction for NAT traffic | |
767 | * to local address when connection is sync-ed | |
768 | */ | |
c0cd1156 | 769 | #if IS_ENABLED(CONFIG_NF_CONNTRACK) |
fc604767 JA |
770 | if (cp->flags & IP_VS_CONN_F_SYNC && local) { |
771 | enum ip_conntrack_info ctinfo; | |
05b4b065 | 772 | struct nf_conn *ct = nf_ct_get(skb, &ctinfo); |
fc604767 JA |
773 | |
774 | if (ct && !nf_ct_is_untracked(ct)) { | |
0d79641a | 775 | IP_VS_DBG_RL_PKT(10, AF_INET6, pp, skb, 0, |
fc604767 JA |
776 | "ip_vs_nat_xmit_v6(): " |
777 | "stopping DNAT to local address"); | |
4115ded1 | 778 | goto tx_error; |
fc604767 JA |
779 | } |
780 | } | |
781 | #endif | |
782 | ||
783 | /* From world but DNAT to loopback address? */ | |
784 | if (local && skb->dev && !(skb->dev->flags & IFF_LOOPBACK) && | |
785 | ipv6_addr_type(&rt->rt6i_dst.addr) & IPV6_ADDR_LOOPBACK) { | |
0d79641a | 786 | IP_VS_DBG_RL_PKT(1, AF_INET6, pp, skb, 0, |
fc604767 JA |
787 | "ip_vs_nat_xmit_v6(): " |
788 | "stopping DNAT to loopback address"); | |
4115ded1 | 789 | goto tx_error; |
b3cdd2a7 JV |
790 | } |
791 | ||
792 | /* copy-on-write the packet before mangling it */ | |
793 | if (!skb_make_writable(skb, sizeof(struct ipv6hdr))) | |
4115ded1 | 794 | goto tx_error; |
b3cdd2a7 | 795 | |
d8d1f30b | 796 | if (skb_cow(skb, rt->dst.dev->hard_header_len)) |
4115ded1 | 797 | goto tx_error; |
b3cdd2a7 | 798 | |
b3cdd2a7 | 799 | /* mangle the packet */ |
4115ded1 | 800 | if (pp->dnat_handler && !pp->dnat_handler(skb, pp, cp, ipvsh)) |
b3cdd2a7 | 801 | goto tx_error; |
4e3fd7a0 | 802 | ipv6_hdr(skb)->daddr = cp->daddr.in6; |
fc604767 | 803 | |
0d79641a | 804 | IP_VS_DBG_PKT(10, AF_INET6, pp, skb, 0, "After DNAT"); |
b3cdd2a7 JV |
805 | |
806 | /* FIXME: when application helper enlarges the packet and the length | |
807 | is larger than the MTU of outgoing device, there will be still | |
808 | MTU problem. */ | |
809 | ||
810 | /* Another hack: avoid icmp_send in ip_fragment */ | |
60ff7467 | 811 | skb->ignore_df = 1; |
b3cdd2a7 | 812 | |
b8abdf09 | 813 | rc = ip_vs_nat_send_or_cont(NFPROTO_IPV6, skb, cp, local); |
026ace06 | 814 | rcu_read_unlock(); |
b3cdd2a7 JV |
815 | |
816 | LeaveFunction(10); | |
b8abdf09 | 817 | return rc; |
b3cdd2a7 | 818 | |
b3cdd2a7 JV |
819 | tx_error: |
820 | LeaveFunction(10); | |
821 | kfree_skb(skb); | |
026ace06 | 822 | rcu_read_unlock(); |
b3cdd2a7 | 823 | return NF_STOLEN; |
b3cdd2a7 JV |
824 | } |
825 | #endif | |
826 | ||
8052ba29 AG |
827 | /* When forwarding a packet, we must ensure that we've got enough headroom |
828 | * for the encapsulation packet in the skb. This also gives us an | |
829 | * opportunity to figure out what the payload_len, dsfield, ttl, and df | |
830 | * values should be, so that we won't need to look at the old ip header | |
831 | * again | |
832 | */ | |
833 | static struct sk_buff * | |
834 | ip_vs_prepare_tunneled_skb(struct sk_buff *skb, int skb_af, | |
835 | unsigned int max_headroom, __u8 *next_protocol, | |
836 | __u32 *payload_len, __u8 *dsfield, __u8 *ttl, | |
837 | __be16 *df) | |
838 | { | |
839 | struct sk_buff *new_skb = NULL; | |
840 | struct iphdr *old_iph = NULL; | |
841 | #ifdef CONFIG_IP_VS_IPV6 | |
842 | struct ipv6hdr *old_ipv6h = NULL; | |
843 | #endif | |
844 | ||
845 | if (skb_headroom(skb) < max_headroom || skb_cloned(skb)) { | |
846 | new_skb = skb_realloc_headroom(skb, max_headroom); | |
847 | if (!new_skb) | |
848 | goto error; | |
849 | consume_skb(skb); | |
850 | skb = new_skb; | |
851 | } | |
852 | ||
853 | #ifdef CONFIG_IP_VS_IPV6 | |
854 | if (skb_af == AF_INET6) { | |
855 | old_ipv6h = ipv6_hdr(skb); | |
856 | *next_protocol = IPPROTO_IPV6; | |
857 | if (payload_len) | |
858 | *payload_len = | |
859 | ntohs(old_ipv6h->payload_len) + | |
860 | sizeof(*old_ipv6h); | |
861 | *dsfield = ipv6_get_dsfield(old_ipv6h); | |
862 | *ttl = old_ipv6h->hop_limit; | |
863 | if (df) | |
864 | *df = 0; | |
865 | } else | |
866 | #endif | |
867 | { | |
868 | old_iph = ip_hdr(skb); | |
869 | /* Copy DF, reset fragment offset and MF */ | |
870 | if (df) | |
871 | *df = (old_iph->frag_off & htons(IP_DF)); | |
872 | *next_protocol = IPPROTO_IPIP; | |
873 | ||
874 | /* fix old IP header checksum */ | |
875 | ip_send_check(old_iph); | |
876 | *dsfield = ipv4_get_dsfield(old_iph); | |
877 | *ttl = old_iph->ttl; | |
878 | if (payload_len) | |
879 | *payload_len = ntohs(old_iph->tot_len); | |
880 | } | |
881 | ||
882 | return skb; | |
883 | error: | |
884 | kfree_skb(skb); | |
885 | return ERR_PTR(-ENOMEM); | |
886 | } | |
887 | ||
888 | static inline int __tun_gso_type_mask(int encaps_af, int orig_af) | |
889 | { | |
890 | if (encaps_af == AF_INET) { | |
891 | if (orig_af == AF_INET) | |
892 | return SKB_GSO_IPIP; | |
893 | ||
894 | return SKB_GSO_SIT; | |
895 | } | |
896 | ||
897 | /* GSO: we need to provide proper SKB_GSO_ value for IPv6: | |
898 | * SKB_GSO_SIT/IPV6 | |
899 | */ | |
900 | return 0; | |
901 | } | |
1da177e4 LT |
902 | |
903 | /* | |
904 | * IP Tunneling transmitter | |
905 | * | |
906 | * This function encapsulates the packet in a new IP packet, its | |
907 | * destination will be set to cp->daddr. Most code of this function | |
908 | * is taken from ipip.c. | |
909 | * | |
910 | * It is used in VS/TUN cluster. The load balancer selects a real | |
911 | * server from a cluster based on a scheduling algorithm, | |
912 | * encapsulates the request packet and forwards it to the selected | |
913 | * server. For example, all real servers are configured with | |
914 | * "ifconfig tunl0 <Virtual IP Address> up". When the server receives | |
915 | * the encapsulated packet, it will decapsulate the packet, processe | |
916 | * the request and return the response packets directly to the client | |
917 | * without passing the load balancer. This can greatly increase the | |
918 | * scalability of virtual server. | |
919 | * | |
920 | * Used for ANY protocol | |
921 | */ | |
922 | int | |
923 | ip_vs_tunnel_xmit(struct sk_buff *skb, struct ip_vs_conn *cp, | |
d4383f04 | 924 | struct ip_vs_protocol *pp, struct ip_vs_iphdr *ipvsh) |
1da177e4 | 925 | { |
3654e611 | 926 | struct netns_ipvs *ipvs = net_ipvs(skb_net(skb)); |
1da177e4 | 927 | struct rtable *rt; /* Route to the other host */ |
c92f5ca2 | 928 | __be32 saddr; /* Source for tunnel */ |
1da177e4 | 929 | struct net_device *tdev; /* Device to other host */ |
8052ba29 AG |
930 | __u8 next_protocol = 0; |
931 | __u8 dsfield = 0; | |
932 | __u8 ttl = 0; | |
933 | __be16 df = 0; | |
934 | __be16 *dfp = NULL; | |
1da177e4 | 935 | struct iphdr *iph; /* Our new IP header */ |
c2636b4d | 936 | unsigned int max_headroom; /* The extra header space needed */ |
4115ded1 | 937 | int ret, local; |
1da177e4 LT |
938 | |
939 | EnterFunction(10); | |
940 | ||
026ace06 | 941 | rcu_read_lock(); |
4a4739d5 | 942 | local = __ip_vs_get_out_rt(cp->af, skb, cp->dest, cp->daddr.ip, |
4115ded1 JA |
943 | IP_VS_RT_MODE_LOCAL | |
944 | IP_VS_RT_MODE_NON_LOCAL | | |
945 | IP_VS_RT_MODE_CONNECT | | |
c63e4de2 | 946 | IP_VS_RT_MODE_TUNNEL, &saddr, ipvsh); |
4115ded1 JA |
947 | if (local < 0) |
948 | goto tx_error; | |
026ace06 JA |
949 | if (local) { |
950 | rcu_read_unlock(); | |
b8abdf09 | 951 | return ip_vs_send_or_cont(NFPROTO_IPV4, skb, cp, 1); |
026ace06 | 952 | } |
1da177e4 | 953 | |
4115ded1 | 954 | rt = skb_rtable(skb); |
d8d1f30b | 955 | tdev = rt->dst.dev; |
1da177e4 | 956 | |
1da177e4 LT |
957 | /* |
958 | * Okay, now see if we can stuff it in the buffer as-is. | |
959 | */ | |
960 | max_headroom = LL_RESERVED_SPACE(tdev) + sizeof(struct iphdr); | |
961 | ||
8052ba29 AG |
962 | /* We only care about the df field if sysctl_pmtu_disc(ipvs) is set */ |
963 | dfp = sysctl_pmtu_disc(ipvs) ? &df : NULL; | |
964 | skb = ip_vs_prepare_tunneled_skb(skb, cp->af, max_headroom, | |
965 | &next_protocol, NULL, &dsfield, | |
966 | &ttl, dfp); | |
967 | if (IS_ERR(skb)) | |
968 | goto tx_error; | |
1da177e4 | 969 | |
8052ba29 AG |
970 | skb = iptunnel_handle_offloads( |
971 | skb, false, __tun_gso_type_mask(AF_INET, cp->af)); | |
ea1d5d77 JA |
972 | if (IS_ERR(skb)) |
973 | goto tx_error; | |
974 | ||
975 | skb->transport_header = skb->network_header; | |
976 | ||
e2d1bca7 ACM |
977 | skb_push(skb, sizeof(struct iphdr)); |
978 | skb_reset_network_header(skb); | |
1da177e4 LT |
979 | memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt)); |
980 | ||
1da177e4 LT |
981 | /* |
982 | * Push down and install the IPIP header. | |
983 | */ | |
eddc9ec5 | 984 | iph = ip_hdr(skb); |
1da177e4 LT |
985 | iph->version = 4; |
986 | iph->ihl = sizeof(struct iphdr)>>2; | |
987 | iph->frag_off = df; | |
8052ba29 AG |
988 | iph->protocol = next_protocol; |
989 | iph->tos = dsfield; | |
c92f5ca2 JA |
990 | iph->daddr = cp->daddr.ip; |
991 | iph->saddr = saddr; | |
8052ba29 | 992 | iph->ttl = ttl; |
73f156a6 | 993 | ip_select_ident(skb, NULL); |
1da177e4 LT |
994 | |
995 | /* Another hack: avoid icmp_send in ip_fragment */ | |
60ff7467 | 996 | skb->ignore_df = 1; |
1da177e4 | 997 | |
b8abdf09 | 998 | ret = ip_vs_tunnel_xmit_prepare(skb, cp); |
f4bc17cd JA |
999 | if (ret == NF_ACCEPT) |
1000 | ip_local_out(skb); | |
1001 | else if (ret == NF_DROP) | |
1002 | kfree_skb(skb); | |
026ace06 | 1003 | rcu_read_unlock(); |
1da177e4 LT |
1004 | |
1005 | LeaveFunction(10); | |
1006 | ||
1007 | return NF_STOLEN; | |
1008 | ||
1da177e4 | 1009 | tx_error: |
ea1d5d77 JA |
1010 | if (!IS_ERR(skb)) |
1011 | kfree_skb(skb); | |
026ace06 | 1012 | rcu_read_unlock(); |
1da177e4 LT |
1013 | LeaveFunction(10); |
1014 | return NF_STOLEN; | |
1015 | } | |
1016 | ||
b3cdd2a7 JV |
1017 | #ifdef CONFIG_IP_VS_IPV6 |
1018 | int | |
1019 | ip_vs_tunnel_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp, | |
d4383f04 | 1020 | struct ip_vs_protocol *pp, struct ip_vs_iphdr *ipvsh) |
b3cdd2a7 JV |
1021 | { |
1022 | struct rt6_info *rt; /* Route to the other host */ | |
714f095f | 1023 | struct in6_addr saddr; /* Source for tunnel */ |
b3cdd2a7 | 1024 | struct net_device *tdev; /* Device to other host */ |
8052ba29 AG |
1025 | __u8 next_protocol = 0; |
1026 | __u32 payload_len = 0; | |
1027 | __u8 dsfield = 0; | |
1028 | __u8 ttl = 0; | |
b3cdd2a7 JV |
1029 | struct ipv6hdr *iph; /* Our new IP header */ |
1030 | unsigned int max_headroom; /* The extra header space needed */ | |
4115ded1 | 1031 | int ret, local; |
b3cdd2a7 JV |
1032 | |
1033 | EnterFunction(10); | |
1034 | ||
026ace06 | 1035 | rcu_read_lock(); |
4a4739d5 | 1036 | local = __ip_vs_get_out_rt_v6(cp->af, skb, cp->dest, &cp->daddr.in6, |
4115ded1 JA |
1037 | &saddr, ipvsh, 1, |
1038 | IP_VS_RT_MODE_LOCAL | | |
1039 | IP_VS_RT_MODE_NON_LOCAL | | |
1040 | IP_VS_RT_MODE_TUNNEL); | |
1041 | if (local < 0) | |
1042 | goto tx_error; | |
026ace06 JA |
1043 | if (local) { |
1044 | rcu_read_unlock(); | |
b8abdf09 | 1045 | return ip_vs_send_or_cont(NFPROTO_IPV6, skb, cp, 1); |
026ace06 | 1046 | } |
b3cdd2a7 | 1047 | |
4115ded1 | 1048 | rt = (struct rt6_info *) skb_dst(skb); |
d8d1f30b | 1049 | tdev = rt->dst.dev; |
b3cdd2a7 | 1050 | |
b3cdd2a7 JV |
1051 | /* |
1052 | * Okay, now see if we can stuff it in the buffer as-is. | |
1053 | */ | |
1054 | max_headroom = LL_RESERVED_SPACE(tdev) + sizeof(struct ipv6hdr); | |
1055 | ||
8052ba29 AG |
1056 | skb = ip_vs_prepare_tunneled_skb(skb, cp->af, max_headroom, |
1057 | &next_protocol, &payload_len, | |
1058 | &dsfield, &ttl, NULL); | |
1059 | if (IS_ERR(skb)) | |
1060 | goto tx_error; | |
b3cdd2a7 | 1061 | |
8052ba29 AG |
1062 | skb = iptunnel_handle_offloads( |
1063 | skb, false, __tun_gso_type_mask(AF_INET6, cp->af)); | |
ea1d5d77 JA |
1064 | if (IS_ERR(skb)) |
1065 | goto tx_error; | |
1066 | ||
714f095f | 1067 | skb->transport_header = skb->network_header; |
b3cdd2a7 JV |
1068 | |
1069 | skb_push(skb, sizeof(struct ipv6hdr)); | |
1070 | skb_reset_network_header(skb); | |
1071 | memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt)); | |
1072 | ||
b3cdd2a7 JV |
1073 | /* |
1074 | * Push down and install the IPIP header. | |
1075 | */ | |
1076 | iph = ipv6_hdr(skb); | |
1077 | iph->version = 6; | |
8052ba29 AG |
1078 | iph->nexthdr = next_protocol; |
1079 | iph->payload_len = htons(payload_len); | |
b3cdd2a7 | 1080 | memset(&iph->flow_lbl, 0, sizeof(iph->flow_lbl)); |
8052ba29 | 1081 | ipv6_change_dsfield(iph, 0, dsfield); |
4e3fd7a0 AD |
1082 | iph->daddr = cp->daddr.in6; |
1083 | iph->saddr = saddr; | |
8052ba29 | 1084 | iph->hop_limit = ttl; |
b3cdd2a7 JV |
1085 | |
1086 | /* Another hack: avoid icmp_send in ip_fragment */ | |
60ff7467 | 1087 | skb->ignore_df = 1; |
b3cdd2a7 | 1088 | |
b8abdf09 | 1089 | ret = ip_vs_tunnel_xmit_prepare(skb, cp); |
f4bc17cd JA |
1090 | if (ret == NF_ACCEPT) |
1091 | ip6_local_out(skb); | |
1092 | else if (ret == NF_DROP) | |
1093 | kfree_skb(skb); | |
026ace06 | 1094 | rcu_read_unlock(); |
b3cdd2a7 JV |
1095 | |
1096 | LeaveFunction(10); | |
1097 | ||
1098 | return NF_STOLEN; | |
1099 | ||
b3cdd2a7 | 1100 | tx_error: |
ea1d5d77 JA |
1101 | if (!IS_ERR(skb)) |
1102 | kfree_skb(skb); | |
026ace06 | 1103 | rcu_read_unlock(); |
b3cdd2a7 JV |
1104 | LeaveFunction(10); |
1105 | return NF_STOLEN; | |
1106 | } | |
1107 | #endif | |
1108 | ||
1da177e4 LT |
1109 | |
1110 | /* | |
1111 | * Direct Routing transmitter | |
1112 | * Used for ANY protocol | |
1113 | */ | |
1114 | int | |
1115 | ip_vs_dr_xmit(struct sk_buff *skb, struct ip_vs_conn *cp, | |
d4383f04 | 1116 | struct ip_vs_protocol *pp, struct ip_vs_iphdr *ipvsh) |
1da177e4 | 1117 | { |
4115ded1 | 1118 | int local; |
1da177e4 LT |
1119 | |
1120 | EnterFunction(10); | |
1121 | ||
026ace06 | 1122 | rcu_read_lock(); |
4a4739d5 | 1123 | local = __ip_vs_get_out_rt(cp->af, skb, cp->dest, cp->daddr.ip, |
4115ded1 JA |
1124 | IP_VS_RT_MODE_LOCAL | |
1125 | IP_VS_RT_MODE_NON_LOCAL | | |
c63e4de2 | 1126 | IP_VS_RT_MODE_KNOWN_NH, NULL, ipvsh); |
4115ded1 | 1127 | if (local < 0) |
1da177e4 | 1128 | goto tx_error; |
026ace06 JA |
1129 | if (local) { |
1130 | rcu_read_unlock(); | |
4115ded1 | 1131 | return ip_vs_send_or_cont(NFPROTO_IPV4, skb, cp, 1); |
026ace06 | 1132 | } |
1da177e4 | 1133 | |
eddc9ec5 | 1134 | ip_send_check(ip_hdr(skb)); |
1da177e4 | 1135 | |
1da177e4 | 1136 | /* Another hack: avoid icmp_send in ip_fragment */ |
60ff7467 | 1137 | skb->ignore_df = 1; |
1da177e4 | 1138 | |
b8abdf09 | 1139 | ip_vs_send_or_cont(NFPROTO_IPV4, skb, cp, 0); |
026ace06 | 1140 | rcu_read_unlock(); |
1da177e4 LT |
1141 | |
1142 | LeaveFunction(10); | |
1143 | return NF_STOLEN; | |
1144 | ||
1da177e4 LT |
1145 | tx_error: |
1146 | kfree_skb(skb); | |
026ace06 | 1147 | rcu_read_unlock(); |
1da177e4 LT |
1148 | LeaveFunction(10); |
1149 | return NF_STOLEN; | |
1150 | } | |
1151 | ||
b3cdd2a7 JV |
1152 | #ifdef CONFIG_IP_VS_IPV6 |
1153 | int | |
1154 | ip_vs_dr_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp, | |
4115ded1 | 1155 | struct ip_vs_protocol *pp, struct ip_vs_iphdr *ipvsh) |
b3cdd2a7 | 1156 | { |
4115ded1 | 1157 | int local; |
b3cdd2a7 JV |
1158 | |
1159 | EnterFunction(10); | |
1160 | ||
026ace06 | 1161 | rcu_read_lock(); |
4a4739d5 AG |
1162 | local = __ip_vs_get_out_rt_v6(cp->af, skb, cp->dest, &cp->daddr.in6, |
1163 | NULL, ipvsh, 0, | |
4115ded1 JA |
1164 | IP_VS_RT_MODE_LOCAL | |
1165 | IP_VS_RT_MODE_NON_LOCAL); | |
1166 | if (local < 0) | |
b3cdd2a7 | 1167 | goto tx_error; |
026ace06 JA |
1168 | if (local) { |
1169 | rcu_read_unlock(); | |
4115ded1 | 1170 | return ip_vs_send_or_cont(NFPROTO_IPV6, skb, cp, 1); |
026ace06 | 1171 | } |
b3cdd2a7 JV |
1172 | |
1173 | /* Another hack: avoid icmp_send in ip_fragment */ | |
60ff7467 | 1174 | skb->ignore_df = 1; |
b3cdd2a7 | 1175 | |
b8abdf09 | 1176 | ip_vs_send_or_cont(NFPROTO_IPV6, skb, cp, 0); |
026ace06 | 1177 | rcu_read_unlock(); |
b3cdd2a7 JV |
1178 | |
1179 | LeaveFunction(10); | |
1180 | return NF_STOLEN; | |
1181 | ||
b3cdd2a7 JV |
1182 | tx_error: |
1183 | kfree_skb(skb); | |
026ace06 | 1184 | rcu_read_unlock(); |
b3cdd2a7 JV |
1185 | LeaveFunction(10); |
1186 | return NF_STOLEN; | |
1187 | } | |
1188 | #endif | |
1189 | ||
1da177e4 LT |
1190 | |
1191 | /* | |
1192 | * ICMP packet transmitter | |
1193 | * called by the ip_vs_in_icmp | |
1194 | */ | |
1195 | int | |
1196 | ip_vs_icmp_xmit(struct sk_buff *skb, struct ip_vs_conn *cp, | |
d4383f04 JDB |
1197 | struct ip_vs_protocol *pp, int offset, unsigned int hooknum, |
1198 | struct ip_vs_iphdr *iph) | |
1da177e4 LT |
1199 | { |
1200 | struct rtable *rt; /* Route to the other host */ | |
1da177e4 | 1201 | int rc; |
fc604767 | 1202 | int local; |
4115ded1 | 1203 | int rt_mode, was_input; |
1da177e4 LT |
1204 | |
1205 | EnterFunction(10); | |
1206 | ||
1207 | /* The ICMP packet for VS/TUN, VS/DR and LOCALNODE will be | |
1208 | forwarded directly here, because there is no need to | |
1209 | translate address/port back */ | |
1210 | if (IP_VS_FWD_METHOD(cp) != IP_VS_CONN_F_MASQ) { | |
1211 | if (cp->packet_xmit) | |
d4383f04 | 1212 | rc = cp->packet_xmit(skb, cp, pp, iph); |
1da177e4 LT |
1213 | else |
1214 | rc = NF_ACCEPT; | |
1215 | /* do not touch skb anymore */ | |
1216 | atomic_inc(&cp->in_pkts); | |
1da177e4 LT |
1217 | goto out; |
1218 | } | |
1219 | ||
1220 | /* | |
1221 | * mangle and send the packet here (only for VS/NAT) | |
1222 | */ | |
4115ded1 | 1223 | was_input = rt_is_input_route(skb_rtable(skb)); |
1da177e4 | 1224 | |
c92f5ca2 JA |
1225 | /* LOCALNODE from FORWARD hook is not supported */ |
1226 | rt_mode = (hooknum != NF_INET_FORWARD) ? | |
1227 | IP_VS_RT_MODE_LOCAL | IP_VS_RT_MODE_NON_LOCAL | | |
1228 | IP_VS_RT_MODE_RDR : IP_VS_RT_MODE_NON_LOCAL; | |
026ace06 | 1229 | rcu_read_lock(); |
4a4739d5 | 1230 | local = __ip_vs_get_out_rt(cp->af, skb, cp->dest, cp->daddr.ip, rt_mode, |
c63e4de2 | 1231 | NULL, iph); |
4115ded1 JA |
1232 | if (local < 0) |
1233 | goto tx_error; | |
1234 | rt = skb_rtable(skb); | |
fc604767 JA |
1235 | |
1236 | /* | |
1237 | * Avoid duplicate tuple in reply direction for NAT traffic | |
1238 | * to local address when connection is sync-ed | |
1239 | */ | |
c0cd1156 | 1240 | #if IS_ENABLED(CONFIG_NF_CONNTRACK) |
fc604767 JA |
1241 | if (cp->flags & IP_VS_CONN_F_SYNC && local) { |
1242 | enum ip_conntrack_info ctinfo; | |
05b4b065 | 1243 | struct nf_conn *ct = nf_ct_get(skb, &ctinfo); |
fc604767 JA |
1244 | |
1245 | if (ct && !nf_ct_is_untracked(ct)) { | |
1246 | IP_VS_DBG(10, "%s(): " | |
1247 | "stopping DNAT to local address %pI4\n", | |
1248 | __func__, &cp->daddr.ip); | |
4115ded1 | 1249 | goto tx_error; |
fc604767 JA |
1250 | } |
1251 | } | |
1252 | #endif | |
1253 | ||
1254 | /* From world but DNAT to loopback address? */ | |
4115ded1 | 1255 | if (local && ipv4_is_loopback(cp->daddr.ip) && was_input) { |
fc604767 JA |
1256 | IP_VS_DBG(1, "%s(): " |
1257 | "stopping DNAT to loopback %pI4\n", | |
1258 | __func__, &cp->daddr.ip); | |
4115ded1 | 1259 | goto tx_error; |
1da177e4 LT |
1260 | } |
1261 | ||
1262 | /* copy-on-write the packet before mangling it */ | |
af1e1cf0 | 1263 | if (!skb_make_writable(skb, offset)) |
4115ded1 | 1264 | goto tx_error; |
1da177e4 | 1265 | |
d8d1f30b | 1266 | if (skb_cow(skb, rt->dst.dev->hard_header_len)) |
4115ded1 | 1267 | goto tx_error; |
1da177e4 | 1268 | |
1da177e4 LT |
1269 | ip_vs_nat_icmp(skb, pp, cp, 0); |
1270 | ||
1271 | /* Another hack: avoid icmp_send in ip_fragment */ | |
60ff7467 | 1272 | skb->ignore_df = 1; |
1da177e4 | 1273 | |
b8abdf09 | 1274 | rc = ip_vs_nat_send_or_cont(NFPROTO_IPV4, skb, cp, local); |
026ace06 | 1275 | rcu_read_unlock(); |
1da177e4 LT |
1276 | goto out; |
1277 | ||
1da177e4 | 1278 | tx_error: |
026ace06 JA |
1279 | kfree_skb(skb); |
1280 | rcu_read_unlock(); | |
1da177e4 LT |
1281 | rc = NF_STOLEN; |
1282 | out: | |
1283 | LeaveFunction(10); | |
1284 | return rc; | |
1da177e4 | 1285 | } |
b3cdd2a7 JV |
1286 | |
1287 | #ifdef CONFIG_IP_VS_IPV6 | |
1288 | int | |
1289 | ip_vs_icmp_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp, | |
d4383f04 | 1290 | struct ip_vs_protocol *pp, int offset, unsigned int hooknum, |
4115ded1 | 1291 | struct ip_vs_iphdr *ipvsh) |
b3cdd2a7 JV |
1292 | { |
1293 | struct rt6_info *rt; /* Route to the other host */ | |
b3cdd2a7 | 1294 | int rc; |
fc604767 | 1295 | int local; |
c92f5ca2 | 1296 | int rt_mode; |
b3cdd2a7 JV |
1297 | |
1298 | EnterFunction(10); | |
1299 | ||
1300 | /* The ICMP packet for VS/TUN, VS/DR and LOCALNODE will be | |
1301 | forwarded directly here, because there is no need to | |
1302 | translate address/port back */ | |
1303 | if (IP_VS_FWD_METHOD(cp) != IP_VS_CONN_F_MASQ) { | |
1304 | if (cp->packet_xmit) | |
4115ded1 | 1305 | rc = cp->packet_xmit(skb, cp, pp, ipvsh); |
b3cdd2a7 JV |
1306 | else |
1307 | rc = NF_ACCEPT; | |
1308 | /* do not touch skb anymore */ | |
1309 | atomic_inc(&cp->in_pkts); | |
1310 | goto out; | |
1311 | } | |
1312 | ||
1313 | /* | |
1314 | * mangle and send the packet here (only for VS/NAT) | |
1315 | */ | |
1316 | ||
c92f5ca2 JA |
1317 | /* LOCALNODE from FORWARD hook is not supported */ |
1318 | rt_mode = (hooknum != NF_INET_FORWARD) ? | |
1319 | IP_VS_RT_MODE_LOCAL | IP_VS_RT_MODE_NON_LOCAL | | |
1320 | IP_VS_RT_MODE_RDR : IP_VS_RT_MODE_NON_LOCAL; | |
026ace06 | 1321 | rcu_read_lock(); |
4a4739d5 AG |
1322 | local = __ip_vs_get_out_rt_v6(cp->af, skb, cp->dest, &cp->daddr.in6, |
1323 | NULL, ipvsh, 0, rt_mode); | |
4115ded1 JA |
1324 | if (local < 0) |
1325 | goto tx_error; | |
1326 | rt = (struct rt6_info *) skb_dst(skb); | |
fc604767 JA |
1327 | /* |
1328 | * Avoid duplicate tuple in reply direction for NAT traffic | |
1329 | * to local address when connection is sync-ed | |
1330 | */ | |
c0cd1156 | 1331 | #if IS_ENABLED(CONFIG_NF_CONNTRACK) |
fc604767 JA |
1332 | if (cp->flags & IP_VS_CONN_F_SYNC && local) { |
1333 | enum ip_conntrack_info ctinfo; | |
05b4b065 | 1334 | struct nf_conn *ct = nf_ct_get(skb, &ctinfo); |
fc604767 JA |
1335 | |
1336 | if (ct && !nf_ct_is_untracked(ct)) { | |
1337 | IP_VS_DBG(10, "%s(): " | |
1338 | "stopping DNAT to local address %pI6\n", | |
1339 | __func__, &cp->daddr.in6); | |
4115ded1 | 1340 | goto tx_error; |
fc604767 JA |
1341 | } |
1342 | } | |
1343 | #endif | |
1344 | ||
1345 | /* From world but DNAT to loopback address? */ | |
1346 | if (local && skb->dev && !(skb->dev->flags & IFF_LOOPBACK) && | |
1347 | ipv6_addr_type(&rt->rt6i_dst.addr) & IPV6_ADDR_LOOPBACK) { | |
1348 | IP_VS_DBG(1, "%s(): " | |
1349 | "stopping DNAT to loopback %pI6\n", | |
1350 | __func__, &cp->daddr.in6); | |
4115ded1 | 1351 | goto tx_error; |
b3cdd2a7 JV |
1352 | } |
1353 | ||
1354 | /* copy-on-write the packet before mangling it */ | |
1355 | if (!skb_make_writable(skb, offset)) | |
4115ded1 | 1356 | goto tx_error; |
b3cdd2a7 | 1357 | |
d8d1f30b | 1358 | if (skb_cow(skb, rt->dst.dev->hard_header_len)) |
4115ded1 | 1359 | goto tx_error; |
b3cdd2a7 | 1360 | |
b3cdd2a7 JV |
1361 | ip_vs_nat_icmp_v6(skb, pp, cp, 0); |
1362 | ||
1363 | /* Another hack: avoid icmp_send in ip_fragment */ | |
60ff7467 | 1364 | skb->ignore_df = 1; |
b3cdd2a7 | 1365 | |
b8abdf09 | 1366 | rc = ip_vs_nat_send_or_cont(NFPROTO_IPV6, skb, cp, local); |
026ace06 | 1367 | rcu_read_unlock(); |
b3cdd2a7 JV |
1368 | goto out; |
1369 | ||
b3cdd2a7 | 1370 | tx_error: |
026ace06 JA |
1371 | kfree_skb(skb); |
1372 | rcu_read_unlock(); | |
b3cdd2a7 JV |
1373 | rc = NF_STOLEN; |
1374 | out: | |
1375 | LeaveFunction(10); | |
1376 | return rc; | |
b3cdd2a7 JV |
1377 | } |
1378 | #endif |