]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * Handle firewalling | |
3 | * Linux ethernet bridge | |
4 | * | |
5 | * Authors: | |
8237908e BDS |
6 | * Lennert Buytenhek <buytenh@gnu.org> |
7 | * Bart De Schuymer <bdschuym@pandora.be> | |
1da177e4 LT |
8 | * |
9 | * This program is free software; you can redistribute it and/or | |
10 | * modify it under the terms of the GNU General Public License | |
11 | * as published by the Free Software Foundation; either version | |
12 | * 2 of the License, or (at your option) any later version. | |
13 | * | |
14 | * Lennert dedicates this file to Kerstin Wurdinger. | |
15 | */ | |
16 | ||
17 | #include <linux/module.h> | |
18 | #include <linux/kernel.h> | |
5a0e3ad6 | 19 | #include <linux/slab.h> |
1da177e4 LT |
20 | #include <linux/ip.h> |
21 | #include <linux/netdevice.h> | |
22 | #include <linux/skbuff.h> | |
14c85021 | 23 | #include <linux/if_arp.h> |
1da177e4 LT |
24 | #include <linux/if_ether.h> |
25 | #include <linux/if_vlan.h> | |
516299d2 MM |
26 | #include <linux/if_pppox.h> |
27 | #include <linux/ppp_defs.h> | |
1da177e4 LT |
28 | #include <linux/netfilter_bridge.h> |
29 | #include <linux/netfilter_ipv4.h> | |
30 | #include <linux/netfilter_ipv6.h> | |
31 | #include <linux/netfilter_arp.h> | |
32 | #include <linux/in_route.h> | |
f216f082 | 33 | #include <linux/inetdevice.h> |
14c85021 | 34 | |
1da177e4 LT |
35 | #include <net/ip.h> |
36 | #include <net/ipv6.h> | |
14c85021 | 37 | #include <net/route.h> |
56768644 | 38 | #include <net/netfilter/br_netfilter.h> |
14c85021 | 39 | |
c055d5b0 FW |
40 | #if IS_ENABLED(CONFIG_NF_CONNTRACK) |
41 | #include <net/netfilter/nf_conntrack.h> | |
42 | #endif | |
43 | ||
1da177e4 | 44 | #include <asm/uaccess.h> |
1da177e4 LT |
45 | #include "br_private.h" |
46 | #ifdef CONFIG_SYSCTL | |
47 | #include <linux/sysctl.h> | |
48 | #endif | |
49 | ||
1da177e4 LT |
50 | #ifdef CONFIG_SYSCTL |
51 | static struct ctl_table_header *brnf_sysctl_header; | |
9c1ea148 BH |
52 | static int brnf_call_iptables __read_mostly = 1; |
53 | static int brnf_call_ip6tables __read_mostly = 1; | |
54 | static int brnf_call_arptables __read_mostly = 1; | |
47e0e1ca HX |
55 | static int brnf_filter_vlan_tagged __read_mostly = 0; |
56 | static int brnf_filter_pppoe_tagged __read_mostly = 0; | |
4981682c | 57 | static int brnf_pass_vlan_indev __read_mostly = 0; |
1da177e4 | 58 | #else |
4df53d8b PM |
59 | #define brnf_call_iptables 1 |
60 | #define brnf_call_ip6tables 1 | |
61 | #define brnf_call_arptables 1 | |
47e0e1ca HX |
62 | #define brnf_filter_vlan_tagged 0 |
63 | #define brnf_filter_pppoe_tagged 0 | |
4981682c | 64 | #define brnf_pass_vlan_indev 0 |
1da177e4 LT |
65 | #endif |
66 | ||
739e4505 | 67 | #define IS_IP(skb) \ |
df8a39de | 68 | (!skb_vlan_tag_present(skb) && skb->protocol == htons(ETH_P_IP)) |
739e4505 FW |
69 | |
70 | #define IS_IPV6(skb) \ | |
df8a39de | 71 | (!skb_vlan_tag_present(skb) && skb->protocol == htons(ETH_P_IPV6)) |
739e4505 FW |
72 | |
73 | #define IS_ARP(skb) \ | |
df8a39de | 74 | (!skb_vlan_tag_present(skb) && skb->protocol == htons(ETH_P_ARP)) |
739e4505 | 75 | |
b6f99a21 | 76 | static inline __be16 vlan_proto(const struct sk_buff *skb) |
8b42ec39 | 77 | { |
df8a39de | 78 | if (skb_vlan_tag_present(skb)) |
13937911 JG |
79 | return skb->protocol; |
80 | else if (skb->protocol == htons(ETH_P_8021Q)) | |
81 | return vlan_eth_hdr(skb)->h_vlan_encapsulated_proto; | |
82 | else | |
83 | return 0; | |
8b42ec39 SH |
84 | } |
85 | ||
86 | #define IS_VLAN_IP(skb) \ | |
13937911 | 87 | (vlan_proto(skb) == htons(ETH_P_IP) && \ |
8b42ec39 SH |
88 | brnf_filter_vlan_tagged) |
89 | ||
90 | #define IS_VLAN_IPV6(skb) \ | |
13937911 | 91 | (vlan_proto(skb) == htons(ETH_P_IPV6) && \ |
8b42ec39 SH |
92 | brnf_filter_vlan_tagged) |
93 | ||
94 | #define IS_VLAN_ARP(skb) \ | |
13937911 | 95 | (vlan_proto(skb) == htons(ETH_P_ARP) && \ |
8b42ec39 | 96 | brnf_filter_vlan_tagged) |
1da177e4 | 97 | |
516299d2 MM |
98 | static inline __be16 pppoe_proto(const struct sk_buff *skb) |
99 | { | |
100 | return *((__be16 *)(skb_mac_header(skb) + ETH_HLEN + | |
101 | sizeof(struct pppoe_hdr))); | |
102 | } | |
103 | ||
104 | #define IS_PPPOE_IP(skb) \ | |
105 | (skb->protocol == htons(ETH_P_PPP_SES) && \ | |
106 | pppoe_proto(skb) == htons(PPP_IP) && \ | |
107 | brnf_filter_pppoe_tagged) | |
108 | ||
109 | #define IS_PPPOE_IPV6(skb) \ | |
110 | (skb->protocol == htons(ETH_P_PPP_SES) && \ | |
111 | pppoe_proto(skb) == htons(PPP_IPV6) && \ | |
112 | brnf_filter_pppoe_tagged) | |
113 | ||
e70deecb FW |
114 | /* largest possible L2 header, see br_nf_dev_queue_xmit() */ |
115 | #define NF_BRIDGE_MAX_MAC_HEADER_LENGTH (PPPOE_SES_HLEN + ETH_HLEN) | |
116 | ||
117 | #if IS_ENABLED(CONFIG_NF_DEFRAG_IPV4) | |
118 | struct brnf_frag_data { | |
119 | char mac[NF_BRIDGE_MAX_MAC_HEADER_LENGTH]; | |
120 | u8 encap_size; | |
121 | u8 size; | |
122 | }; | |
123 | ||
124 | static DEFINE_PER_CPU(struct brnf_frag_data, brnf_frag_data_storage); | |
125 | #endif | |
126 | ||
38330783 FW |
127 | static struct nf_bridge_info *nf_bridge_info_get(const struct sk_buff *skb) |
128 | { | |
129 | return skb->nf_bridge; | |
130 | } | |
131 | ||
a9fcc6a4 FW |
132 | static void nf_bridge_info_free(struct sk_buff *skb) |
133 | { | |
134 | if (skb->nf_bridge) { | |
135 | nf_bridge_put(skb->nf_bridge); | |
136 | skb->nf_bridge = NULL; | |
137 | } | |
138 | } | |
139 | ||
4adf0af6 SW |
140 | static inline struct rtable *bridge_parent_rtable(const struct net_device *dev) |
141 | { | |
b5ed54e9 | 142 | struct net_bridge_port *port; |
143 | ||
144 | port = br_port_get_rcu(dev); | |
145 | return port ? &port->br->fake_rtable : NULL; | |
4adf0af6 | 146 | } |
1da177e4 | 147 | |
5dce971a SH |
148 | static inline struct net_device *bridge_parent(const struct net_device *dev) |
149 | { | |
b5ed54e9 | 150 | struct net_bridge_port *port; |
5dce971a | 151 | |
b5ed54e9 | 152 | port = br_port_get_rcu(dev); |
153 | return port ? port->br->dev : NULL; | |
5dce971a | 154 | } |
1da177e4 | 155 | |
fdeabdef SH |
156 | static inline struct nf_bridge_info *nf_bridge_alloc(struct sk_buff *skb) |
157 | { | |
158 | skb->nf_bridge = kzalloc(sizeof(struct nf_bridge_info), GFP_ATOMIC); | |
159 | if (likely(skb->nf_bridge)) | |
160 | atomic_set(&(skb->nf_bridge->use), 1); | |
161 | ||
162 | return skb->nf_bridge; | |
163 | } | |
164 | ||
2dc2f207 PM |
165 | static inline struct nf_bridge_info *nf_bridge_unshare(struct sk_buff *skb) |
166 | { | |
167 | struct nf_bridge_info *nf_bridge = skb->nf_bridge; | |
168 | ||
169 | if (atomic_read(&nf_bridge->use) > 1) { | |
170 | struct nf_bridge_info *tmp = nf_bridge_alloc(skb); | |
171 | ||
172 | if (tmp) { | |
173 | memcpy(tmp, nf_bridge, sizeof(struct nf_bridge_info)); | |
174 | atomic_set(&tmp->use, 1); | |
2dc2f207 | 175 | } |
4c3a76ab | 176 | nf_bridge_put(nf_bridge); |
2dc2f207 PM |
177 | nf_bridge = tmp; |
178 | } | |
179 | return nf_bridge; | |
180 | } | |
181 | ||
8d045163 FW |
182 | static unsigned int nf_bridge_encap_header_len(const struct sk_buff *skb) |
183 | { | |
184 | switch (skb->protocol) { | |
185 | case __cpu_to_be16(ETH_P_8021Q): | |
186 | return VLAN_HLEN; | |
187 | case __cpu_to_be16(ETH_P_PPP_SES): | |
188 | return PPPOE_SES_HLEN; | |
189 | default: | |
190 | return 0; | |
191 | } | |
192 | } | |
193 | ||
fc38582d PM |
194 | static inline void nf_bridge_push_encap_header(struct sk_buff *skb) |
195 | { | |
196 | unsigned int len = nf_bridge_encap_header_len(skb); | |
197 | ||
198 | skb_push(skb, len); | |
199 | skb->network_header -= len; | |
200 | } | |
201 | ||
202 | static inline void nf_bridge_pull_encap_header(struct sk_buff *skb) | |
fdeabdef | 203 | { |
fc38582d PM |
204 | unsigned int len = nf_bridge_encap_header_len(skb); |
205 | ||
206 | skb_pull(skb, len); | |
207 | skb->network_header += len; | |
208 | } | |
fdeabdef | 209 | |
fc38582d PM |
210 | static inline void nf_bridge_pull_encap_header_rcsum(struct sk_buff *skb) |
211 | { | |
212 | unsigned int len = nf_bridge_encap_header_len(skb); | |
213 | ||
214 | skb_pull_rcsum(skb, len); | |
215 | skb->network_header += len; | |
216 | } | |
217 | ||
462fb2af BD |
218 | /* When handing a packet over to the IP layer |
219 | * check whether we have a skb that is in the | |
220 | * expected format | |
221 | */ | |
222 | ||
d0280232 | 223 | static int br_parse_ip_options(struct sk_buff *skb) |
462fb2af | 224 | { |
b71d1d42 | 225 | const struct iphdr *iph; |
462fb2af BD |
226 | struct net_device *dev = skb->dev; |
227 | u32 len; | |
228 | ||
6caab7b0 SB |
229 | if (!pskb_may_pull(skb, sizeof(struct iphdr))) |
230 | goto inhdr_error; | |
231 | ||
462fb2af | 232 | iph = ip_hdr(skb); |
462fb2af BD |
233 | |
234 | /* Basic sanity checks */ | |
235 | if (iph->ihl < 5 || iph->version != 4) | |
236 | goto inhdr_error; | |
237 | ||
238 | if (!pskb_may_pull(skb, iph->ihl*4)) | |
239 | goto inhdr_error; | |
240 | ||
241 | iph = ip_hdr(skb); | |
242 | if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl))) | |
243 | goto inhdr_error; | |
244 | ||
245 | len = ntohs(iph->tot_len); | |
246 | if (skb->len < len) { | |
247 | IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_INTRUNCATEDPKTS); | |
248 | goto drop; | |
249 | } else if (len < (iph->ihl*4)) | |
250 | goto inhdr_error; | |
251 | ||
252 | if (pskb_trim_rcsum(skb, len)) { | |
253 | IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_INDISCARDS); | |
254 | goto drop; | |
255 | } | |
256 | ||
f8e9881c | 257 | memset(IPCB(skb), 0, sizeof(struct inet_skb_parm)); |
7677e868 HX |
258 | /* We should really parse IP options here but until |
259 | * somebody who actually uses IP options complains to | |
260 | * us we'll just silently ignore the options because | |
261 | * we're lazy! | |
262 | */ | |
462fb2af BD |
263 | return 0; |
264 | ||
265 | inhdr_error: | |
266 | IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_INHDRERRORS); | |
267 | drop: | |
268 | return -1; | |
269 | } | |
270 | ||
4a9d2f20 FW |
271 | static void nf_bridge_update_protocol(struct sk_buff *skb) |
272 | { | |
3eaf4025 FW |
273 | switch (skb->nf_bridge->orig_proto) { |
274 | case BRNF_PROTO_8021Q: | |
4a9d2f20 | 275 | skb->protocol = htons(ETH_P_8021Q); |
3eaf4025 FW |
276 | break; |
277 | case BRNF_PROTO_PPPOE: | |
4a9d2f20 | 278 | skb->protocol = htons(ETH_P_PPP_SES); |
3eaf4025 FW |
279 | break; |
280 | case BRNF_PROTO_UNCHANGED: | |
281 | break; | |
282 | } | |
4a9d2f20 FW |
283 | } |
284 | ||
1da177e4 LT |
285 | /* PF_BRIDGE/PRE_ROUTING *********************************************/ |
286 | /* Undo the changes made for ip6tables PREROUTING and continue the | |
287 | * bridge PRE_ROUTING hook. */ | |
7026b1dd | 288 | static int br_nf_pre_routing_finish_ipv6(struct sock *sk, struct sk_buff *skb) |
1da177e4 | 289 | { |
38330783 | 290 | struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb); |
511c3f92 | 291 | struct rtable *rt; |
1da177e4 | 292 | |
a1e67951 | 293 | if (nf_bridge->pkt_otherhost) { |
1da177e4 | 294 | skb->pkt_type = PACKET_OTHERHOST; |
a1e67951 | 295 | nf_bridge->pkt_otherhost = false; |
1da177e4 LT |
296 | } |
297 | nf_bridge->mask ^= BRNF_NF_BRIDGE_PREROUTING; | |
298 | ||
511c3f92 ED |
299 | rt = bridge_parent_rtable(nf_bridge->physindev); |
300 | if (!rt) { | |
4adf0af6 SW |
301 | kfree_skb(skb); |
302 | return 0; | |
303 | } | |
f9181f4f | 304 | skb_dst_set_noref(skb, &rt->dst); |
1da177e4 LT |
305 | |
306 | skb->dev = nf_bridge->physindev; | |
e179e632 | 307 | nf_bridge_update_protocol(skb); |
fc38582d | 308 | nf_bridge_push_encap_header(skb); |
7026b1dd DM |
309 | NF_HOOK_THRESH(NFPROTO_BRIDGE, NF_BR_PRE_ROUTING, sk, skb, |
310 | skb->dev, NULL, | |
1da177e4 LT |
311 | br_handle_frame_finish, 1); |
312 | ||
313 | return 0; | |
314 | } | |
315 | ||
e179e632 BDS |
316 | /* Obtain the correct destination MAC address, while preserving the original |
317 | * source MAC address. If we already know this address, we just copy it. If we | |
318 | * don't, we use the neighbour framework to find out. In both cases, we make | |
319 | * sure that br_handle_frame_finish() is called afterwards. | |
320 | */ | |
7026b1dd | 321 | static int br_nf_pre_routing_finish_bridge(struct sock *sk, struct sk_buff *skb) |
e179e632 | 322 | { |
f6b72b62 | 323 | struct neighbour *neigh; |
e179e632 BDS |
324 | struct dst_entry *dst; |
325 | ||
326 | skb->dev = bridge_parent(skb->dev); | |
327 | if (!skb->dev) | |
328 | goto free_skb; | |
329 | dst = skb_dst(skb); | |
f9d75166 DM |
330 | neigh = dst_neigh_lookup_skb(dst, skb); |
331 | if (neigh) { | |
38330783 | 332 | struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb); |
f9d75166 DM |
333 | int ret; |
334 | ||
335 | if (neigh->hh.hh_len) { | |
336 | neigh_hh_bridge(&neigh->hh, skb); | |
337 | skb->dev = nf_bridge->physindev; | |
7026b1dd | 338 | ret = br_handle_frame_finish(sk, skb); |
f9d75166 DM |
339 | } else { |
340 | /* the neighbour function below overwrites the complete | |
341 | * MAC header, so we save the Ethernet source address and | |
342 | * protocol number. | |
343 | */ | |
344 | skb_copy_from_linear_data_offset(skb, | |
345 | -(ETH_HLEN-ETH_ALEN), | |
e70deecb | 346 | nf_bridge->neigh_header, |
f9d75166 DM |
347 | ETH_HLEN-ETH_ALEN); |
348 | /* tell br_dev_xmit to continue with forwarding */ | |
349 | nf_bridge->mask |= BRNF_BRIDGED_DNAT; | |
93fdd47e | 350 | /* FIXME Need to refragment */ |
f9d75166 DM |
351 | ret = neigh->output(neigh, skb); |
352 | } | |
353 | neigh_release(neigh); | |
354 | return ret; | |
e179e632 BDS |
355 | } |
356 | free_skb: | |
357 | kfree_skb(skb); | |
358 | return 0; | |
359 | } | |
360 | ||
c055d5b0 FW |
361 | static bool dnat_took_place(const struct sk_buff *skb) |
362 | { | |
363 | #if IS_ENABLED(CONFIG_NF_CONNTRACK) | |
364 | enum ip_conntrack_info ctinfo; | |
365 | struct nf_conn *ct; | |
366 | ||
367 | ct = nf_ct_get(skb, &ctinfo); | |
368 | if (!ct || nf_ct_is_untracked(ct)) | |
369 | return false; | |
370 | ||
371 | return test_bit(IPS_DST_NAT_BIT, &ct->status); | |
372 | #else | |
373 | return false; | |
374 | #endif | |
375 | } | |
376 | ||
1da177e4 | 377 | /* This requires some explaining. If DNAT has taken place, |
ea2d9b41 | 378 | * we will need to fix up the destination Ethernet address. |
1da177e4 LT |
379 | * |
380 | * There are two cases to consider: | |
381 | * 1. The packet was DNAT'ed to a device in the same bridge | |
382 | * port group as it was received on. We can still bridge | |
383 | * the packet. | |
384 | * 2. The packet was DNAT'ed to a different device, either | |
385 | * a non-bridged device or another bridge port group. | |
386 | * The packet will need to be routed. | |
387 | * | |
388 | * The correct way of distinguishing between these two cases is to | |
389 | * call ip_route_input() and to look at skb->dst->dev, which is | |
390 | * changed to the destination device if ip_route_input() succeeds. | |
391 | * | |
ea2d9b41 | 392 | * Let's first consider the case that ip_route_input() succeeds: |
1da177e4 | 393 | * |
ea2d9b41 BDS |
394 | * If the output device equals the logical bridge device the packet |
395 | * came in on, we can consider this bridging. The corresponding MAC | |
396 | * address will be obtained in br_nf_pre_routing_finish_bridge. | |
1da177e4 LT |
397 | * Otherwise, the packet is considered to be routed and we just |
398 | * change the destination MAC address so that the packet will | |
f216f082 BDS |
399 | * later be passed up to the IP stack to be routed. For a redirected |
400 | * packet, ip_route_input() will give back the localhost as output device, | |
401 | * which differs from the bridge device. | |
1da177e4 | 402 | * |
ea2d9b41 | 403 | * Let's now consider the case that ip_route_input() fails: |
1da177e4 | 404 | * |
f216f082 BDS |
405 | * This can be because the destination address is martian, in which case |
406 | * the packet will be dropped. | |
ea2d9b41 BDS |
407 | * If IP forwarding is disabled, ip_route_input() will fail, while |
408 | * ip_route_output_key() can return success. The source | |
409 | * address for ip_route_output_key() is set to zero, so ip_route_output_key() | |
1da177e4 | 410 | * thinks we're handling a locally generated packet and won't care |
ea2d9b41 BDS |
411 | * if IP forwarding is enabled. If the output device equals the logical bridge |
412 | * device, we proceed as if ip_route_input() succeeded. If it differs from the | |
413 | * logical bridge port or if ip_route_output_key() fails we drop the packet. | |
414 | */ | |
7026b1dd | 415 | static int br_nf_pre_routing_finish(struct sock *sk, struct sk_buff *skb) |
1da177e4 LT |
416 | { |
417 | struct net_device *dev = skb->dev; | |
eddc9ec5 | 418 | struct iphdr *iph = ip_hdr(skb); |
38330783 | 419 | struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb); |
511c3f92 | 420 | struct rtable *rt; |
f216f082 | 421 | int err; |
93fdd47e HX |
422 | int frag_max_size; |
423 | ||
424 | frag_max_size = IPCB(skb)->frag_max_size; | |
425 | BR_INPUT_SKB_CB(skb)->frag_max_size = frag_max_size; | |
1da177e4 | 426 | |
a1e67951 | 427 | if (nf_bridge->pkt_otherhost) { |
1da177e4 | 428 | skb->pkt_type = PACKET_OTHERHOST; |
a1e67951 | 429 | nf_bridge->pkt_otherhost = false; |
1da177e4 LT |
430 | } |
431 | nf_bridge->mask ^= BRNF_NF_BRIDGE_PREROUTING; | |
1da177e4 | 432 | if (dnat_took_place(skb)) { |
f216f082 | 433 | if ((err = ip_route_input(skb, iph->daddr, iph->saddr, iph->tos, dev))) { |
f3abc9b9 | 434 | struct in_device *in_dev = __in_dev_get_rcu(dev); |
f216f082 BDS |
435 | |
436 | /* If err equals -EHOSTUNREACH the error is due to a | |
437 | * martian destination or due to the fact that | |
438 | * forwarding is disabled. For most martian packets, | |
439 | * ip_route_output_key() will fail. It won't fail for 2 types of | |
440 | * martian destinations: loopback destinations and destination | |
441 | * 0.0.0.0. In both cases the packet will be dropped because the | |
442 | * destination is the loopback device and not the bridge. */ | |
443 | if (err != -EHOSTUNREACH || !in_dev || IN_DEV_FORWARD(in_dev)) | |
444 | goto free_skb; | |
1da177e4 | 445 | |
78fbfd8a DM |
446 | rt = ip_route_output(dev_net(dev), iph->daddr, 0, |
447 | RT_TOS(iph->tos), 0); | |
b23dd4fe | 448 | if (!IS_ERR(rt)) { |
1c011bed | 449 | /* - Bridged-and-DNAT'ed traffic doesn't |
f216f082 | 450 | * require ip_forwarding. */ |
b23dd4fe DM |
451 | if (rt->dst.dev == dev) { |
452 | skb_dst_set(skb, &rt->dst); | |
1da177e4 LT |
453 | goto bridged_dnat; |
454 | } | |
b23dd4fe | 455 | ip_rt_put(rt); |
1da177e4 | 456 | } |
f216f082 | 457 | free_skb: |
1da177e4 LT |
458 | kfree_skb(skb); |
459 | return 0; | |
460 | } else { | |
adf30907 | 461 | if (skb_dst(skb)->dev == dev) { |
1da177e4 | 462 | bridged_dnat: |
1da177e4 | 463 | skb->dev = nf_bridge->physindev; |
e179e632 | 464 | nf_bridge_update_protocol(skb); |
fc38582d | 465 | nf_bridge_push_encap_header(skb); |
713aefa3 JE |
466 | NF_HOOK_THRESH(NFPROTO_BRIDGE, |
467 | NF_BR_PRE_ROUTING, | |
7026b1dd | 468 | sk, skb, skb->dev, NULL, |
1da177e4 LT |
469 | br_nf_pre_routing_finish_bridge, |
470 | 1); | |
471 | return 0; | |
472 | } | |
04091142 | 473 | ether_addr_copy(eth_hdr(skb)->h_dest, dev->dev_addr); |
1da177e4 LT |
474 | skb->pkt_type = PACKET_HOST; |
475 | } | |
476 | } else { | |
511c3f92 ED |
477 | rt = bridge_parent_rtable(nf_bridge->physindev); |
478 | if (!rt) { | |
4adf0af6 SW |
479 | kfree_skb(skb); |
480 | return 0; | |
481 | } | |
f9181f4f | 482 | skb_dst_set_noref(skb, &rt->dst); |
1da177e4 LT |
483 | } |
484 | ||
485 | skb->dev = nf_bridge->physindev; | |
e179e632 | 486 | nf_bridge_update_protocol(skb); |
fc38582d | 487 | nf_bridge_push_encap_header(skb); |
7026b1dd DM |
488 | NF_HOOK_THRESH(NFPROTO_BRIDGE, NF_BR_PRE_ROUTING, sk, skb, |
489 | skb->dev, NULL, | |
1da177e4 LT |
490 | br_handle_frame_finish, 1); |
491 | ||
492 | return 0; | |
493 | } | |
494 | ||
4981682c PNA |
495 | static struct net_device *brnf_get_logical_dev(struct sk_buff *skb, const struct net_device *dev) |
496 | { | |
497 | struct net_device *vlan, *br; | |
498 | ||
499 | br = bridge_parent(dev); | |
df8a39de | 500 | if (brnf_pass_vlan_indev == 0 || !skb_vlan_tag_present(skb)) |
4981682c PNA |
501 | return br; |
502 | ||
f06c7f9f | 503 | vlan = __vlan_find_dev_deep_rcu(br, skb->vlan_proto, |
df8a39de | 504 | skb_vlan_tag_get(skb) & VLAN_VID_MASK); |
4981682c PNA |
505 | |
506 | return vlan ? vlan : br; | |
507 | } | |
508 | ||
1da177e4 | 509 | /* Some common code for IPv4/IPv6 */ |
5dce971a | 510 | static struct net_device *setup_pre_routing(struct sk_buff *skb) |
1da177e4 | 511 | { |
38330783 | 512 | struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb); |
1da177e4 LT |
513 | |
514 | if (skb->pkt_type == PACKET_OTHERHOST) { | |
515 | skb->pkt_type = PACKET_HOST; | |
a1e67951 | 516 | nf_bridge->pkt_otherhost = true; |
1da177e4 LT |
517 | } |
518 | ||
519 | nf_bridge->mask |= BRNF_NF_BRIDGE_PREROUTING; | |
520 | nf_bridge->physindev = skb->dev; | |
4981682c | 521 | skb->dev = brnf_get_logical_dev(skb, skb->dev); |
3eaf4025 | 522 | |
e179e632 | 523 | if (skb->protocol == htons(ETH_P_8021Q)) |
3eaf4025 | 524 | nf_bridge->orig_proto = BRNF_PROTO_8021Q; |
e179e632 | 525 | else if (skb->protocol == htons(ETH_P_PPP_SES)) |
3eaf4025 | 526 | nf_bridge->orig_proto = BRNF_PROTO_PPPOE; |
5dce971a | 527 | |
6b8dbcf2 FW |
528 | /* Must drop socket now because of tproxy. */ |
529 | skb_orphan(skb); | |
5dce971a | 530 | return skb->dev; |
1da177e4 LT |
531 | } |
532 | ||
533 | /* We only check the length. A bridge shouldn't do any hop-by-hop stuff anyway */ | |
534 | static int check_hbh_len(struct sk_buff *skb) | |
535 | { | |
0660e03f | 536 | unsigned char *raw = (u8 *)(ipv6_hdr(skb) + 1); |
1da177e4 | 537 | u32 pkt_len; |
d56f90a7 ACM |
538 | const unsigned char *nh = skb_network_header(skb); |
539 | int off = raw - nh; | |
789bc3e5 | 540 | int len = (raw[1] + 1) << 3; |
1da177e4 LT |
541 | |
542 | if ((raw + len) - skb->data > skb_headlen(skb)) | |
543 | goto bad; | |
544 | ||
545 | off += 2; | |
546 | len -= 2; | |
547 | ||
548 | while (len > 0) { | |
d56f90a7 | 549 | int optlen = nh[off + 1] + 2; |
1da177e4 | 550 | |
d56f90a7 | 551 | switch (nh[off]) { |
1de5a71c | 552 | case IPV6_TLV_PAD1: |
1da177e4 LT |
553 | optlen = 1; |
554 | break; | |
555 | ||
556 | case IPV6_TLV_PADN: | |
557 | break; | |
558 | ||
559 | case IPV6_TLV_JUMBO: | |
d56f90a7 | 560 | if (nh[off + 1] != 4 || (off & 3) != 2) |
1da177e4 | 561 | goto bad; |
d56f90a7 | 562 | pkt_len = ntohl(*(__be32 *) (nh + off + 2)); |
b0366486 | 563 | if (pkt_len <= IPV6_MAXPLEN || |
0660e03f | 564 | ipv6_hdr(skb)->payload_len) |
b0366486 | 565 | goto bad; |
1da177e4 LT |
566 | if (pkt_len > skb->len - sizeof(struct ipv6hdr)) |
567 | goto bad; | |
b0366486 | 568 | if (pskb_trim_rcsum(skb, |
789bc3e5 | 569 | pkt_len + sizeof(struct ipv6hdr))) |
b0366486 | 570 | goto bad; |
d56f90a7 | 571 | nh = skb_network_header(skb); |
1da177e4 LT |
572 | break; |
573 | default: | |
574 | if (optlen > len) | |
575 | goto bad; | |
576 | break; | |
577 | } | |
578 | off += optlen; | |
579 | len -= optlen; | |
580 | } | |
581 | if (len == 0) | |
582 | return 0; | |
583 | bad: | |
584 | return -1; | |
585 | ||
586 | } | |
587 | ||
588 | /* Replicate the checks that IPv6 does on packet reception and pass the packet | |
589 | * to ip6tables, which doesn't support NAT, so things are fairly simple. */ | |
795aa6ef | 590 | static unsigned int br_nf_pre_routing_ipv6(const struct nf_hook_ops *ops, |
789bc3e5 | 591 | struct sk_buff *skb, |
238e54c9 | 592 | const struct nf_hook_state *state) |
1da177e4 | 593 | { |
b71d1d42 | 594 | const struct ipv6hdr *hdr; |
1da177e4 | 595 | u32 pkt_len; |
1da177e4 LT |
596 | |
597 | if (skb->len < sizeof(struct ipv6hdr)) | |
deef4b52 | 598 | return NF_DROP; |
1da177e4 LT |
599 | |
600 | if (!pskb_may_pull(skb, sizeof(struct ipv6hdr))) | |
deef4b52 | 601 | return NF_DROP; |
1da177e4 | 602 | |
0660e03f | 603 | hdr = ipv6_hdr(skb); |
1da177e4 LT |
604 | |
605 | if (hdr->version != 6) | |
deef4b52 | 606 | return NF_DROP; |
1da177e4 LT |
607 | |
608 | pkt_len = ntohs(hdr->payload_len); | |
609 | ||
610 | if (pkt_len || hdr->nexthdr != NEXTHDR_HOP) { | |
611 | if (pkt_len + sizeof(struct ipv6hdr) > skb->len) | |
deef4b52 | 612 | return NF_DROP; |
b38dfee3 | 613 | if (pskb_trim_rcsum(skb, pkt_len + sizeof(struct ipv6hdr))) |
deef4b52 | 614 | return NF_DROP; |
1da177e4 LT |
615 | } |
616 | if (hdr->nexthdr == NEXTHDR_HOP && check_hbh_len(skb)) | |
deef4b52 | 617 | return NF_DROP; |
1da177e4 | 618 | |
789bc3e5 | 619 | nf_bridge_put(skb->nf_bridge); |
fdeabdef | 620 | if (!nf_bridge_alloc(skb)) |
1da177e4 | 621 | return NF_DROP; |
5dce971a SH |
622 | if (!setup_pre_routing(skb)) |
623 | return NF_DROP; | |
1da177e4 | 624 | |
e179e632 | 625 | skb->protocol = htons(ETH_P_IPV6); |
7026b1dd DM |
626 | NF_HOOK(NFPROTO_IPV6, NF_INET_PRE_ROUTING, state->sk, skb, |
627 | skb->dev, NULL, | |
1da177e4 LT |
628 | br_nf_pre_routing_finish_ipv6); |
629 | ||
630 | return NF_STOLEN; | |
1da177e4 LT |
631 | } |
632 | ||
633 | /* Direct IPv6 traffic to br_nf_pre_routing_ipv6. | |
634 | * Replicate the checks that IPv4 does on packet reception. | |
635 | * Set skb->dev to the bridge device (i.e. parent of the | |
636 | * receiving device) to make netfilter happy, the REDIRECT | |
637 | * target in particular. Save the original destination IP | |
638 | * address to be able to detect DNAT afterwards. */ | |
795aa6ef PM |
639 | static unsigned int br_nf_pre_routing(const struct nf_hook_ops *ops, |
640 | struct sk_buff *skb, | |
238e54c9 | 641 | const struct nf_hook_state *state) |
1da177e4 | 642 | { |
4df53d8b PM |
643 | struct net_bridge_port *p; |
644 | struct net_bridge *br; | |
e7c243c9 EP |
645 | __u32 len = nf_bridge_encap_header_len(skb); |
646 | ||
e7c243c9 | 647 | if (unlikely(!pskb_may_pull(skb, len))) |
deef4b52 | 648 | return NF_DROP; |
1da177e4 | 649 | |
238e54c9 | 650 | p = br_port_get_rcu(state->in); |
4df53d8b | 651 | if (p == NULL) |
deef4b52 | 652 | return NF_DROP; |
4df53d8b PM |
653 | br = p->br; |
654 | ||
739e4505 | 655 | if (IS_IPV6(skb) || IS_VLAN_IPV6(skb) || IS_PPPOE_IPV6(skb)) { |
4df53d8b | 656 | if (!brnf_call_ip6tables && !br->nf_call_ip6tables) |
1da177e4 | 657 | return NF_ACCEPT; |
4df53d8b | 658 | |
fc38582d | 659 | nf_bridge_pull_encap_header_rcsum(skb); |
238e54c9 | 660 | return br_nf_pre_routing_ipv6(ops, skb, state); |
1da177e4 | 661 | } |
4df53d8b PM |
662 | |
663 | if (!brnf_call_iptables && !br->nf_call_iptables) | |
1da177e4 | 664 | return NF_ACCEPT; |
1da177e4 | 665 | |
739e4505 | 666 | if (!IS_IP(skb) && !IS_VLAN_IP(skb) && !IS_PPPOE_IP(skb)) |
1da177e4 LT |
667 | return NF_ACCEPT; |
668 | ||
fc38582d | 669 | nf_bridge_pull_encap_header_rcsum(skb); |
1da177e4 | 670 | |
462fb2af | 671 | if (br_parse_ip_options(skb)) |
deef4b52 | 672 | return NF_DROP; |
17762060 | 673 | |
789bc3e5 | 674 | nf_bridge_put(skb->nf_bridge); |
fdeabdef | 675 | if (!nf_bridge_alloc(skb)) |
1da177e4 | 676 | return NF_DROP; |
5dce971a SH |
677 | if (!setup_pre_routing(skb)) |
678 | return NF_DROP; | |
c055d5b0 | 679 | |
e179e632 | 680 | skb->protocol = htons(ETH_P_IP); |
1da177e4 | 681 | |
7026b1dd DM |
682 | NF_HOOK(NFPROTO_IPV4, NF_INET_PRE_ROUTING, state->sk, skb, |
683 | skb->dev, NULL, | |
1da177e4 LT |
684 | br_nf_pre_routing_finish); |
685 | ||
686 | return NF_STOLEN; | |
1da177e4 LT |
687 | } |
688 | ||
689 | ||
690 | /* PF_BRIDGE/LOCAL_IN ************************************************/ | |
691 | /* The packet is locally destined, which requires a real | |
692 | * dst_entry, so detach the fake one. On the way up, the | |
693 | * packet would pass through PRE_ROUTING again (which already | |
694 | * took place when the packet entered the bridge), but we | |
695 | * register an IPv4 PRE_ROUTING 'sabotage' hook that will | |
696 | * prevent this from happening. */ | |
795aa6ef PM |
697 | static unsigned int br_nf_local_in(const struct nf_hook_ops *ops, |
698 | struct sk_buff *skb, | |
238e54c9 | 699 | const struct nf_hook_state *state) |
1da177e4 | 700 | { |
a881e963 | 701 | br_drop_fake_rtable(skb); |
1da177e4 LT |
702 | return NF_ACCEPT; |
703 | } | |
704 | ||
1da177e4 | 705 | /* PF_BRIDGE/FORWARD *************************************************/ |
7026b1dd | 706 | static int br_nf_forward_finish(struct sock *sk, struct sk_buff *skb) |
1da177e4 | 707 | { |
38330783 | 708 | struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb); |
1da177e4 | 709 | struct net_device *in; |
1da177e4 | 710 | |
739e4505 | 711 | if (!IS_ARP(skb) && !IS_VLAN_ARP(skb)) { |
0b67c43c FW |
712 | int frag_max_size; |
713 | ||
714 | if (skb->protocol == htons(ETH_P_IP)) { | |
715 | frag_max_size = IPCB(skb)->frag_max_size; | |
716 | BR_INPUT_SKB_CB(skb)->frag_max_size = frag_max_size; | |
717 | } | |
718 | ||
1da177e4 | 719 | in = nf_bridge->physindev; |
a1e67951 | 720 | if (nf_bridge->pkt_otherhost) { |
1da177e4 | 721 | skb->pkt_type = PACKET_OTHERHOST; |
a1e67951 | 722 | nf_bridge->pkt_otherhost = false; |
1da177e4 | 723 | } |
e94c6743 | 724 | nf_bridge_update_protocol(skb); |
1da177e4 LT |
725 | } else { |
726 | in = *((struct net_device **)(skb->cb)); | |
727 | } | |
fc38582d | 728 | nf_bridge_push_encap_header(skb); |
e179e632 | 729 | |
7026b1dd DM |
730 | NF_HOOK_THRESH(NFPROTO_BRIDGE, NF_BR_FORWARD, sk, skb, |
731 | in, skb->dev, br_forward_finish, 1); | |
1da177e4 LT |
732 | return 0; |
733 | } | |
734 | ||
739e4505 | 735 | |
1da177e4 LT |
736 | /* This is the 'purely bridged' case. For IP, we pass the packet to |
737 | * netfilter with indev and outdev set to the bridge device, | |
738 | * but we are still able to filter on the 'real' indev/outdev | |
739 | * because of the physdev module. For ARP, indev and outdev are the | |
740 | * bridge ports. */ | |
795aa6ef PM |
741 | static unsigned int br_nf_forward_ip(const struct nf_hook_ops *ops, |
742 | struct sk_buff *skb, | |
238e54c9 | 743 | const struct nf_hook_state *state) |
1da177e4 | 744 | { |
1da177e4 | 745 | struct nf_bridge_info *nf_bridge; |
5dce971a | 746 | struct net_device *parent; |
76108cea | 747 | u_int8_t pf; |
1da177e4 LT |
748 | |
749 | if (!skb->nf_bridge) | |
750 | return NF_ACCEPT; | |
751 | ||
2dc2f207 PM |
752 | /* Need exclusive nf_bridge_info since we might have multiple |
753 | * different physoutdevs. */ | |
754 | if (!nf_bridge_unshare(skb)) | |
755 | return NF_DROP; | |
756 | ||
38330783 FW |
757 | nf_bridge = nf_bridge_info_get(skb); |
758 | if (!nf_bridge) | |
759 | return NF_DROP; | |
760 | ||
238e54c9 | 761 | parent = bridge_parent(state->out); |
5dce971a SH |
762 | if (!parent) |
763 | return NF_DROP; | |
764 | ||
739e4505 | 765 | if (IS_IP(skb) || IS_VLAN_IP(skb) || IS_PPPOE_IP(skb)) |
aa740f46 | 766 | pf = NFPROTO_IPV4; |
739e4505 | 767 | else if (IS_IPV6(skb) || IS_VLAN_IPV6(skb) || IS_PPPOE_IPV6(skb)) |
aa740f46 | 768 | pf = NFPROTO_IPV6; |
a2bd40ad HX |
769 | else |
770 | return NF_ACCEPT; | |
1da177e4 | 771 | |
3db05fea | 772 | nf_bridge_pull_encap_header(skb); |
1da177e4 | 773 | |
1da177e4 LT |
774 | if (skb->pkt_type == PACKET_OTHERHOST) { |
775 | skb->pkt_type = PACKET_HOST; | |
a1e67951 | 776 | nf_bridge->pkt_otherhost = true; |
1da177e4 LT |
777 | } |
778 | ||
0b67c43c FW |
779 | if (pf == NFPROTO_IPV4) { |
780 | int frag_max = BR_INPUT_SKB_CB(skb)->frag_max_size; | |
781 | ||
782 | if (br_parse_ip_options(skb)) | |
783 | return NF_DROP; | |
784 | ||
785 | IPCB(skb)->frag_max_size = frag_max; | |
786 | } | |
6b1e960f | 787 | |
1da177e4 | 788 | nf_bridge->physoutdev = skb->dev; |
aa740f46 | 789 | if (pf == NFPROTO_IPV4) |
e179e632 BDS |
790 | skb->protocol = htons(ETH_P_IP); |
791 | else | |
792 | skb->protocol = htons(ETH_P_IPV6); | |
1da177e4 | 793 | |
7026b1dd DM |
794 | NF_HOOK(pf, NF_INET_FORWARD, NULL, skb, |
795 | brnf_get_logical_dev(skb, state->in), | |
238e54c9 | 796 | parent, br_nf_forward_finish); |
1da177e4 LT |
797 | |
798 | return NF_STOLEN; | |
799 | } | |
800 | ||
795aa6ef PM |
801 | static unsigned int br_nf_forward_arp(const struct nf_hook_ops *ops, |
802 | struct sk_buff *skb, | |
238e54c9 | 803 | const struct nf_hook_state *state) |
1da177e4 | 804 | { |
4df53d8b PM |
805 | struct net_bridge_port *p; |
806 | struct net_bridge *br; | |
1da177e4 LT |
807 | struct net_device **d = (struct net_device **)(skb->cb); |
808 | ||
238e54c9 | 809 | p = br_port_get_rcu(state->out); |
4df53d8b PM |
810 | if (p == NULL) |
811 | return NF_ACCEPT; | |
812 | br = p->br; | |
813 | ||
814 | if (!brnf_call_arptables && !br->nf_call_arptables) | |
1da177e4 | 815 | return NF_ACCEPT; |
1da177e4 | 816 | |
739e4505 | 817 | if (!IS_ARP(skb)) { |
8b42ec39 | 818 | if (!IS_VLAN_ARP(skb)) |
1da177e4 | 819 | return NF_ACCEPT; |
3db05fea | 820 | nf_bridge_pull_encap_header(skb); |
1da177e4 LT |
821 | } |
822 | ||
d0a92be0 | 823 | if (arp_hdr(skb)->ar_pln != 4) { |
fc38582d | 824 | if (IS_VLAN_ARP(skb)) |
3db05fea | 825 | nf_bridge_push_encap_header(skb); |
1da177e4 LT |
826 | return NF_ACCEPT; |
827 | } | |
238e54c9 | 828 | *d = state->in; |
7026b1dd DM |
829 | NF_HOOK(NFPROTO_ARP, NF_ARP_FORWARD, state->sk, skb, |
830 | state->in, state->out, br_nf_forward_finish); | |
1da177e4 LT |
831 | |
832 | return NF_STOLEN; | |
833 | } | |
834 | ||
aff09ce3 | 835 | #if IS_ENABLED(CONFIG_NF_DEFRAG_IPV4) |
aadd51aa | 836 | static int br_nf_push_frag_xmit(struct sock *sk, struct sk_buff *skb) |
8bd63cf1 | 837 | { |
e70deecb | 838 | struct brnf_frag_data *data; |
8bd63cf1 | 839 | int err; |
8bd63cf1 | 840 | |
e70deecb FW |
841 | data = this_cpu_ptr(&brnf_frag_data_storage); |
842 | err = skb_cow_head(skb, data->size); | |
8bd63cf1 | 843 | |
e70deecb | 844 | if (err) { |
8bd63cf1 FW |
845 | kfree_skb(skb); |
846 | return 0; | |
847 | } | |
848 | ||
e70deecb FW |
849 | skb_copy_to_linear_data_offset(skb, -data->size, data->mac, data->size); |
850 | __skb_push(skb, data->encap_size); | |
851 | ||
a9fcc6a4 | 852 | nf_bridge_info_free(skb); |
7026b1dd | 853 | return br_dev_queue_push_xmit(sk, skb); |
8bd63cf1 FW |
854 | } |
855 | ||
7026b1dd | 856 | static int br_nf_dev_queue_xmit(struct sock *sk, struct sk_buff *skb) |
2e2f7aef | 857 | { |
462fb2af | 858 | int ret; |
93fdd47e | 859 | int frag_max_size; |
7a8d831d | 860 | unsigned int mtu_reserved; |
462fb2af | 861 | |
a9fcc6a4 FW |
862 | if (skb_is_gso(skb) || skb->protocol != htons(ETH_P_IP)) { |
863 | nf_bridge_info_free(skb); | |
7026b1dd | 864 | return br_dev_queue_push_xmit(sk, skb); |
a9fcc6a4 | 865 | } |
7a8d831d FW |
866 | |
867 | mtu_reserved = nf_bridge_mtu_reduction(skb); | |
93fdd47e HX |
868 | /* This is wrong! We should preserve the original fragment |
869 | * boundaries by preserving frag_list rather than refragmenting. | |
870 | */ | |
7a8d831d | 871 | if (skb->len + mtu_reserved > skb->dev->mtu) { |
e70deecb FW |
872 | struct brnf_frag_data *data; |
873 | ||
93fdd47e | 874 | frag_max_size = BR_INPUT_SKB_CB(skb)->frag_max_size; |
462fb2af BD |
875 | if (br_parse_ip_options(skb)) |
876 | /* Drop invalid packet */ | |
877 | return NF_DROP; | |
93fdd47e | 878 | IPCB(skb)->frag_max_size = frag_max_size; |
e70deecb FW |
879 | |
880 | nf_bridge_update_protocol(skb); | |
881 | ||
882 | data = this_cpu_ptr(&brnf_frag_data_storage); | |
883 | data->encap_size = nf_bridge_encap_header_len(skb); | |
884 | data->size = ETH_HLEN + data->encap_size; | |
885 | ||
886 | skb_copy_from_linear_data_offset(skb, -data->size, data->mac, | |
887 | data->size); | |
888 | ||
7026b1dd | 889 | ret = ip_fragment(sk, skb, br_nf_push_frag_xmit); |
e70deecb | 890 | } else { |
a9fcc6a4 | 891 | nf_bridge_info_free(skb); |
7026b1dd | 892 | ret = br_dev_queue_push_xmit(sk, skb); |
e70deecb | 893 | } |
462fb2af BD |
894 | |
895 | return ret; | |
2e2f7aef | 896 | } |
c197facc | 897 | #else |
7026b1dd | 898 | static int br_nf_dev_queue_xmit(struct sock *sk, struct sk_buff *skb) |
c197facc | 899 | { |
a9fcc6a4 FW |
900 | nf_bridge_info_free(skb); |
901 | return br_dev_queue_push_xmit(sk, skb); | |
c197facc | 902 | } |
903 | #endif | |
1da177e4 LT |
904 | |
905 | /* PF_BRIDGE/POST_ROUTING ********************************************/ | |
795aa6ef PM |
906 | static unsigned int br_nf_post_routing(const struct nf_hook_ops *ops, |
907 | struct sk_buff *skb, | |
238e54c9 | 908 | const struct nf_hook_state *state) |
1da177e4 | 909 | { |
38330783 | 910 | struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb); |
1da177e4 | 911 | struct net_device *realoutdev = bridge_parent(skb->dev); |
76108cea | 912 | u_int8_t pf; |
1da177e4 | 913 | |
e4bb9bcb FW |
914 | /* if nf_bridge is set, but ->physoutdev is NULL, this packet came in |
915 | * on a bridge, but was delivered locally and is now being routed: | |
916 | * | |
917 | * POST_ROUTING was already invoked from the ip stack. | |
918 | */ | |
919 | if (!nf_bridge || !nf_bridge->physoutdev) | |
81d9ddae PM |
920 | return NF_ACCEPT; |
921 | ||
5dce971a SH |
922 | if (!realoutdev) |
923 | return NF_DROP; | |
924 | ||
739e4505 | 925 | if (IS_IP(skb) || IS_VLAN_IP(skb) || IS_PPPOE_IP(skb)) |
aa740f46 | 926 | pf = NFPROTO_IPV4; |
739e4505 | 927 | else if (IS_IPV6(skb) || IS_VLAN_IPV6(skb) || IS_PPPOE_IPV6(skb)) |
aa740f46 | 928 | pf = NFPROTO_IPV6; |
a2bd40ad HX |
929 | else |
930 | return NF_ACCEPT; | |
1da177e4 | 931 | |
1da177e4 LT |
932 | /* We assume any code from br_dev_queue_push_xmit onwards doesn't care |
933 | * about the value of skb->pkt_type. */ | |
934 | if (skb->pkt_type == PACKET_OTHERHOST) { | |
935 | skb->pkt_type = PACKET_HOST; | |
a1e67951 | 936 | nf_bridge->pkt_otherhost = true; |
1da177e4 LT |
937 | } |
938 | ||
fc38582d | 939 | nf_bridge_pull_encap_header(skb); |
aa740f46 | 940 | if (pf == NFPROTO_IPV4) |
e179e632 BDS |
941 | skb->protocol = htons(ETH_P_IP); |
942 | else | |
943 | skb->protocol = htons(ETH_P_IPV6); | |
1da177e4 | 944 | |
7026b1dd DM |
945 | NF_HOOK(pf, NF_INET_POST_ROUTING, state->sk, skb, |
946 | NULL, realoutdev, | |
2e2f7aef | 947 | br_nf_dev_queue_xmit); |
1da177e4 LT |
948 | |
949 | return NF_STOLEN; | |
1da177e4 LT |
950 | } |
951 | ||
1da177e4 LT |
952 | /* IP/SABOTAGE *****************************************************/ |
953 | /* Don't hand locally destined packets to PF_INET(6)/PRE_ROUTING | |
954 | * for the second time. */ | |
795aa6ef PM |
955 | static unsigned int ip_sabotage_in(const struct nf_hook_ops *ops, |
956 | struct sk_buff *skb, | |
238e54c9 | 957 | const struct nf_hook_state *state) |
1da177e4 | 958 | { |
3db05fea HX |
959 | if (skb->nf_bridge && |
960 | !(skb->nf_bridge->mask & BRNF_NF_BRIDGE_PREROUTING)) { | |
1da177e4 LT |
961 | return NF_STOP; |
962 | } | |
963 | ||
964 | return NF_ACCEPT; | |
965 | } | |
966 | ||
e5de75bf PNA |
967 | /* This is called when br_netfilter has called into iptables/netfilter, |
968 | * and DNAT has taken place on a bridge-forwarded packet. | |
969 | * | |
970 | * neigh->output has created a new MAC header, with local br0 MAC | |
971 | * as saddr. | |
972 | * | |
973 | * This restores the original MAC saddr of the bridged packet | |
974 | * before invoking bridge forward logic to transmit the packet. | |
975 | */ | |
976 | static void br_nf_pre_routing_finish_bridge_slow(struct sk_buff *skb) | |
977 | { | |
38330783 | 978 | struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb); |
e5de75bf PNA |
979 | |
980 | skb_pull(skb, ETH_HLEN); | |
981 | nf_bridge->mask &= ~BRNF_BRIDGED_DNAT; | |
982 | ||
e70deecb FW |
983 | BUILD_BUG_ON(sizeof(nf_bridge->neigh_header) != (ETH_HLEN - ETH_ALEN)); |
984 | ||
985 | skb_copy_to_linear_data_offset(skb, -(ETH_HLEN - ETH_ALEN), | |
986 | nf_bridge->neigh_header, | |
987 | ETH_HLEN - ETH_ALEN); | |
e5de75bf | 988 | skb->dev = nf_bridge->physindev; |
7fb48c5b FW |
989 | |
990 | nf_bridge->physoutdev = NULL; | |
7026b1dd | 991 | br_handle_frame_finish(NULL, skb); |
e5de75bf PNA |
992 | } |
993 | ||
1a4ba64d | 994 | static int br_nf_dev_xmit(struct sk_buff *skb) |
e5de75bf PNA |
995 | { |
996 | if (skb->nf_bridge && (skb->nf_bridge->mask & BRNF_BRIDGED_DNAT)) { | |
997 | br_nf_pre_routing_finish_bridge_slow(skb); | |
998 | return 1; | |
999 | } | |
1000 | return 0; | |
1001 | } | |
1a4ba64d PNA |
1002 | |
1003 | static const struct nf_br_ops br_ops = { | |
1004 | .br_dev_xmit_hook = br_nf_dev_xmit, | |
1005 | }; | |
e5de75bf | 1006 | |
4b7fd5d9 PNA |
1007 | void br_netfilter_enable(void) |
1008 | { | |
1009 | } | |
1010 | EXPORT_SYMBOL_GPL(br_netfilter_enable); | |
1011 | ||
ea2d9b41 BDS |
1012 | /* For br_nf_post_routing, we need (prio = NF_BR_PRI_LAST), because |
1013 | * br_dev_queue_push_xmit is called afterwards */ | |
1999414a | 1014 | static struct nf_hook_ops br_nf_ops[] __read_mostly = { |
1490fd89 CG |
1015 | { |
1016 | .hook = br_nf_pre_routing, | |
1017 | .owner = THIS_MODULE, | |
aa740f46 | 1018 | .pf = NFPROTO_BRIDGE, |
1490fd89 CG |
1019 | .hooknum = NF_BR_PRE_ROUTING, |
1020 | .priority = NF_BR_PRI_BRNF, | |
1021 | }, | |
1022 | { | |
1023 | .hook = br_nf_local_in, | |
1024 | .owner = THIS_MODULE, | |
aa740f46 | 1025 | .pf = NFPROTO_BRIDGE, |
1490fd89 CG |
1026 | .hooknum = NF_BR_LOCAL_IN, |
1027 | .priority = NF_BR_PRI_BRNF, | |
1028 | }, | |
1029 | { | |
1030 | .hook = br_nf_forward_ip, | |
1031 | .owner = THIS_MODULE, | |
aa740f46 | 1032 | .pf = NFPROTO_BRIDGE, |
1490fd89 CG |
1033 | .hooknum = NF_BR_FORWARD, |
1034 | .priority = NF_BR_PRI_BRNF - 1, | |
1035 | }, | |
1036 | { | |
1037 | .hook = br_nf_forward_arp, | |
1038 | .owner = THIS_MODULE, | |
aa740f46 | 1039 | .pf = NFPROTO_BRIDGE, |
1490fd89 CG |
1040 | .hooknum = NF_BR_FORWARD, |
1041 | .priority = NF_BR_PRI_BRNF, | |
1042 | }, | |
1490fd89 CG |
1043 | { |
1044 | .hook = br_nf_post_routing, | |
1045 | .owner = THIS_MODULE, | |
aa740f46 | 1046 | .pf = NFPROTO_BRIDGE, |
1490fd89 CG |
1047 | .hooknum = NF_BR_POST_ROUTING, |
1048 | .priority = NF_BR_PRI_LAST, | |
1049 | }, | |
1050 | { | |
1051 | .hook = ip_sabotage_in, | |
1052 | .owner = THIS_MODULE, | |
aa740f46 | 1053 | .pf = NFPROTO_IPV4, |
1490fd89 CG |
1054 | .hooknum = NF_INET_PRE_ROUTING, |
1055 | .priority = NF_IP_PRI_FIRST, | |
1056 | }, | |
1057 | { | |
1058 | .hook = ip_sabotage_in, | |
1059 | .owner = THIS_MODULE, | |
aa740f46 | 1060 | .pf = NFPROTO_IPV6, |
1490fd89 CG |
1061 | .hooknum = NF_INET_PRE_ROUTING, |
1062 | .priority = NF_IP6_PRI_FIRST, | |
1063 | }, | |
1da177e4 LT |
1064 | }; |
1065 | ||
1066 | #ifdef CONFIG_SYSCTL | |
1067 | static | |
fe2c6338 | 1068 | int brnf_sysctl_call_tables(struct ctl_table *ctl, int write, |
56b148eb | 1069 | void __user *buffer, size_t *lenp, loff_t *ppos) |
1da177e4 LT |
1070 | { |
1071 | int ret; | |
1072 | ||
8d65af78 | 1073 | ret = proc_dointvec(ctl, write, buffer, lenp, ppos); |
1da177e4 LT |
1074 | |
1075 | if (write && *(int *)(ctl->data)) | |
1076 | *(int *)(ctl->data) = 1; | |
1077 | return ret; | |
1078 | } | |
1079 | ||
fe2c6338 | 1080 | static struct ctl_table brnf_table[] = { |
1da177e4 | 1081 | { |
1da177e4 LT |
1082 | .procname = "bridge-nf-call-arptables", |
1083 | .data = &brnf_call_arptables, | |
1084 | .maxlen = sizeof(int), | |
1085 | .mode = 0644, | |
6d9f239a | 1086 | .proc_handler = brnf_sysctl_call_tables, |
1da177e4 LT |
1087 | }, |
1088 | { | |
1da177e4 LT |
1089 | .procname = "bridge-nf-call-iptables", |
1090 | .data = &brnf_call_iptables, | |
1091 | .maxlen = sizeof(int), | |
1092 | .mode = 0644, | |
6d9f239a | 1093 | .proc_handler = brnf_sysctl_call_tables, |
1da177e4 LT |
1094 | }, |
1095 | { | |
1da177e4 LT |
1096 | .procname = "bridge-nf-call-ip6tables", |
1097 | .data = &brnf_call_ip6tables, | |
1098 | .maxlen = sizeof(int), | |
1099 | .mode = 0644, | |
6d9f239a | 1100 | .proc_handler = brnf_sysctl_call_tables, |
1da177e4 LT |
1101 | }, |
1102 | { | |
1da177e4 LT |
1103 | .procname = "bridge-nf-filter-vlan-tagged", |
1104 | .data = &brnf_filter_vlan_tagged, | |
1105 | .maxlen = sizeof(int), | |
1106 | .mode = 0644, | |
6d9f239a | 1107 | .proc_handler = brnf_sysctl_call_tables, |
516299d2 MM |
1108 | }, |
1109 | { | |
516299d2 MM |
1110 | .procname = "bridge-nf-filter-pppoe-tagged", |
1111 | .data = &brnf_filter_pppoe_tagged, | |
1112 | .maxlen = sizeof(int), | |
1113 | .mode = 0644, | |
6d9f239a | 1114 | .proc_handler = brnf_sysctl_call_tables, |
1da177e4 | 1115 | }, |
4981682c PNA |
1116 | { |
1117 | .procname = "bridge-nf-pass-vlan-input-dev", | |
1118 | .data = &brnf_pass_vlan_indev, | |
1119 | .maxlen = sizeof(int), | |
1120 | .mode = 0644, | |
1121 | .proc_handler = brnf_sysctl_call_tables, | |
1122 | }, | |
f8572d8f | 1123 | { } |
1da177e4 | 1124 | }; |
1da177e4 LT |
1125 | #endif |
1126 | ||
34666d46 | 1127 | static int __init br_netfilter_init(void) |
1da177e4 | 1128 | { |
5eb87f45 | 1129 | int ret; |
1da177e4 | 1130 | |
34666d46 | 1131 | ret = nf_register_hooks(br_nf_ops, ARRAY_SIZE(br_nf_ops)); |
5eb87f45 | 1132 | if (ret < 0) |
1da177e4 | 1133 | return ret; |
fc66f95c | 1134 | |
1da177e4 | 1135 | #ifdef CONFIG_SYSCTL |
ec8f23ce | 1136 | brnf_sysctl_header = register_net_sysctl(&init_net, "net/bridge", brnf_table); |
1da177e4 | 1137 | if (brnf_sysctl_header == NULL) { |
789bc3e5 SH |
1138 | printk(KERN_WARNING |
1139 | "br_netfilter: can't register to sysctl.\n"); | |
96727239 GU |
1140 | nf_unregister_hooks(br_nf_ops, ARRAY_SIZE(br_nf_ops)); |
1141 | return -ENOMEM; | |
1da177e4 LT |
1142 | } |
1143 | #endif | |
1a4ba64d | 1144 | RCU_INIT_POINTER(nf_br_ops, &br_ops); |
1da177e4 | 1145 | printk(KERN_NOTICE "Bridge firewalling registered\n"); |
1da177e4 LT |
1146 | return 0; |
1147 | } | |
1148 | ||
34666d46 | 1149 | static void __exit br_netfilter_fini(void) |
1da177e4 | 1150 | { |
1a4ba64d | 1151 | RCU_INIT_POINTER(nf_br_ops, NULL); |
5eb87f45 | 1152 | nf_unregister_hooks(br_nf_ops, ARRAY_SIZE(br_nf_ops)); |
1da177e4 | 1153 | #ifdef CONFIG_SYSCTL |
5dd3df10 | 1154 | unregister_net_sysctl_table(brnf_sysctl_header); |
1da177e4 LT |
1155 | #endif |
1156 | } | |
34666d46 PNA |
1157 | |
1158 | module_init(br_netfilter_init); | |
1159 | module_exit(br_netfilter_fini); | |
1160 | ||
1161 | MODULE_LICENSE("GPL"); | |
1162 | MODULE_AUTHOR("Lennert Buytenhek <buytenh@gnu.org>"); | |
1163 | MODULE_AUTHOR("Bart De Schuymer <bdschuym@pandora.be>"); | |
1164 | MODULE_DESCRIPTION("Linux ethernet netfilter firewall bridge"); |