]>
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> | |
efb6de9b | 37 | #include <net/addrconf.h> |
14c85021 | 38 | #include <net/route.h> |
56768644 | 39 | #include <net/netfilter/br_netfilter.h> |
14c85021 | 40 | |
1da177e4 | 41 | #include <asm/uaccess.h> |
1da177e4 LT |
42 | #include "br_private.h" |
43 | #ifdef CONFIG_SYSCTL | |
44 | #include <linux/sysctl.h> | |
45 | #endif | |
46 | ||
1da177e4 LT |
47 | #ifdef CONFIG_SYSCTL |
48 | static struct ctl_table_header *brnf_sysctl_header; | |
9c1ea148 BH |
49 | static int brnf_call_iptables __read_mostly = 1; |
50 | static int brnf_call_ip6tables __read_mostly = 1; | |
51 | static int brnf_call_arptables __read_mostly = 1; | |
47e0e1ca HX |
52 | static int brnf_filter_vlan_tagged __read_mostly = 0; |
53 | static int brnf_filter_pppoe_tagged __read_mostly = 0; | |
4981682c | 54 | static int brnf_pass_vlan_indev __read_mostly = 0; |
1da177e4 | 55 | #else |
4df53d8b PM |
56 | #define brnf_call_iptables 1 |
57 | #define brnf_call_ip6tables 1 | |
58 | #define brnf_call_arptables 1 | |
47e0e1ca HX |
59 | #define brnf_filter_vlan_tagged 0 |
60 | #define brnf_filter_pppoe_tagged 0 | |
4981682c | 61 | #define brnf_pass_vlan_indev 0 |
1da177e4 LT |
62 | #endif |
63 | ||
739e4505 | 64 | #define IS_IP(skb) \ |
df8a39de | 65 | (!skb_vlan_tag_present(skb) && skb->protocol == htons(ETH_P_IP)) |
739e4505 FW |
66 | |
67 | #define IS_IPV6(skb) \ | |
df8a39de | 68 | (!skb_vlan_tag_present(skb) && skb->protocol == htons(ETH_P_IPV6)) |
739e4505 FW |
69 | |
70 | #define IS_ARP(skb) \ | |
df8a39de | 71 | (!skb_vlan_tag_present(skb) && skb->protocol == htons(ETH_P_ARP)) |
739e4505 | 72 | |
b6f99a21 | 73 | static inline __be16 vlan_proto(const struct sk_buff *skb) |
8b42ec39 | 74 | { |
df8a39de | 75 | if (skb_vlan_tag_present(skb)) |
13937911 JG |
76 | return skb->protocol; |
77 | else if (skb->protocol == htons(ETH_P_8021Q)) | |
78 | return vlan_eth_hdr(skb)->h_vlan_encapsulated_proto; | |
79 | else | |
80 | return 0; | |
8b42ec39 SH |
81 | } |
82 | ||
83 | #define IS_VLAN_IP(skb) \ | |
13937911 | 84 | (vlan_proto(skb) == htons(ETH_P_IP) && \ |
8b42ec39 SH |
85 | brnf_filter_vlan_tagged) |
86 | ||
87 | #define IS_VLAN_IPV6(skb) \ | |
13937911 | 88 | (vlan_proto(skb) == htons(ETH_P_IPV6) && \ |
8b42ec39 SH |
89 | brnf_filter_vlan_tagged) |
90 | ||
91 | #define IS_VLAN_ARP(skb) \ | |
13937911 | 92 | (vlan_proto(skb) == htons(ETH_P_ARP) && \ |
8b42ec39 | 93 | brnf_filter_vlan_tagged) |
1da177e4 | 94 | |
516299d2 MM |
95 | static inline __be16 pppoe_proto(const struct sk_buff *skb) |
96 | { | |
97 | return *((__be16 *)(skb_mac_header(skb) + ETH_HLEN + | |
98 | sizeof(struct pppoe_hdr))); | |
99 | } | |
100 | ||
101 | #define IS_PPPOE_IP(skb) \ | |
102 | (skb->protocol == htons(ETH_P_PPP_SES) && \ | |
103 | pppoe_proto(skb) == htons(PPP_IP) && \ | |
104 | brnf_filter_pppoe_tagged) | |
105 | ||
106 | #define IS_PPPOE_IPV6(skb) \ | |
107 | (skb->protocol == htons(ETH_P_PPP_SES) && \ | |
108 | pppoe_proto(skb) == htons(PPP_IPV6) && \ | |
109 | brnf_filter_pppoe_tagged) | |
110 | ||
e70deecb FW |
111 | /* largest possible L2 header, see br_nf_dev_queue_xmit() */ |
112 | #define NF_BRIDGE_MAX_MAC_HEADER_LENGTH (PPPOE_SES_HLEN + ETH_HLEN) | |
113 | ||
114 | #if IS_ENABLED(CONFIG_NF_DEFRAG_IPV4) | |
115 | struct brnf_frag_data { | |
116 | char mac[NF_BRIDGE_MAX_MAC_HEADER_LENGTH]; | |
117 | u8 encap_size; | |
118 | u8 size; | |
119 | }; | |
120 | ||
121 | static DEFINE_PER_CPU(struct brnf_frag_data, brnf_frag_data_storage); | |
122 | #endif | |
123 | ||
38330783 FW |
124 | static struct nf_bridge_info *nf_bridge_info_get(const struct sk_buff *skb) |
125 | { | |
126 | return skb->nf_bridge; | |
127 | } | |
128 | ||
a9fcc6a4 FW |
129 | static void nf_bridge_info_free(struct sk_buff *skb) |
130 | { | |
131 | if (skb->nf_bridge) { | |
132 | nf_bridge_put(skb->nf_bridge); | |
133 | skb->nf_bridge = NULL; | |
134 | } | |
135 | } | |
136 | ||
4adf0af6 SW |
137 | static inline struct rtable *bridge_parent_rtable(const struct net_device *dev) |
138 | { | |
b5ed54e9 | 139 | struct net_bridge_port *port; |
140 | ||
141 | port = br_port_get_rcu(dev); | |
142 | return port ? &port->br->fake_rtable : NULL; | |
4adf0af6 | 143 | } |
1da177e4 | 144 | |
5dce971a SH |
145 | static inline struct net_device *bridge_parent(const struct net_device *dev) |
146 | { | |
b5ed54e9 | 147 | struct net_bridge_port *port; |
5dce971a | 148 | |
b5ed54e9 | 149 | port = br_port_get_rcu(dev); |
150 | return port ? port->br->dev : NULL; | |
5dce971a | 151 | } |
1da177e4 | 152 | |
fdeabdef SH |
153 | static inline struct nf_bridge_info *nf_bridge_alloc(struct sk_buff *skb) |
154 | { | |
155 | skb->nf_bridge = kzalloc(sizeof(struct nf_bridge_info), GFP_ATOMIC); | |
156 | if (likely(skb->nf_bridge)) | |
157 | atomic_set(&(skb->nf_bridge->use), 1); | |
158 | ||
159 | return skb->nf_bridge; | |
160 | } | |
161 | ||
2dc2f207 PM |
162 | static inline struct nf_bridge_info *nf_bridge_unshare(struct sk_buff *skb) |
163 | { | |
164 | struct nf_bridge_info *nf_bridge = skb->nf_bridge; | |
165 | ||
166 | if (atomic_read(&nf_bridge->use) > 1) { | |
167 | struct nf_bridge_info *tmp = nf_bridge_alloc(skb); | |
168 | ||
169 | if (tmp) { | |
170 | memcpy(tmp, nf_bridge, sizeof(struct nf_bridge_info)); | |
171 | atomic_set(&tmp->use, 1); | |
2dc2f207 | 172 | } |
4c3a76ab | 173 | nf_bridge_put(nf_bridge); |
2dc2f207 PM |
174 | nf_bridge = tmp; |
175 | } | |
176 | return nf_bridge; | |
177 | } | |
178 | ||
8d045163 FW |
179 | static unsigned int nf_bridge_encap_header_len(const struct sk_buff *skb) |
180 | { | |
181 | switch (skb->protocol) { | |
182 | case __cpu_to_be16(ETH_P_8021Q): | |
183 | return VLAN_HLEN; | |
184 | case __cpu_to_be16(ETH_P_PPP_SES): | |
185 | return PPPOE_SES_HLEN; | |
186 | default: | |
187 | return 0; | |
188 | } | |
189 | } | |
190 | ||
fc38582d PM |
191 | static inline void nf_bridge_push_encap_header(struct sk_buff *skb) |
192 | { | |
193 | unsigned int len = nf_bridge_encap_header_len(skb); | |
194 | ||
195 | skb_push(skb, len); | |
196 | skb->network_header -= len; | |
197 | } | |
198 | ||
199 | static inline void nf_bridge_pull_encap_header(struct sk_buff *skb) | |
fdeabdef | 200 | { |
fc38582d PM |
201 | unsigned int len = nf_bridge_encap_header_len(skb); |
202 | ||
203 | skb_pull(skb, len); | |
204 | skb->network_header += len; | |
205 | } | |
fdeabdef | 206 | |
fc38582d PM |
207 | static inline void nf_bridge_pull_encap_header_rcsum(struct sk_buff *skb) |
208 | { | |
209 | unsigned int len = nf_bridge_encap_header_len(skb); | |
210 | ||
211 | skb_pull_rcsum(skb, len); | |
212 | skb->network_header += len; | |
213 | } | |
214 | ||
462fb2af BD |
215 | /* When handing a packet over to the IP layer |
216 | * check whether we have a skb that is in the | |
217 | * expected format | |
218 | */ | |
219 | ||
77d574e7 | 220 | static int br_validate_ipv4(struct sk_buff *skb) |
462fb2af | 221 | { |
b71d1d42 | 222 | const struct iphdr *iph; |
462fb2af BD |
223 | struct net_device *dev = skb->dev; |
224 | u32 len; | |
225 | ||
6caab7b0 SB |
226 | if (!pskb_may_pull(skb, sizeof(struct iphdr))) |
227 | goto inhdr_error; | |
228 | ||
462fb2af | 229 | iph = ip_hdr(skb); |
462fb2af BD |
230 | |
231 | /* Basic sanity checks */ | |
232 | if (iph->ihl < 5 || iph->version != 4) | |
233 | goto inhdr_error; | |
234 | ||
235 | if (!pskb_may_pull(skb, iph->ihl*4)) | |
236 | goto inhdr_error; | |
237 | ||
238 | iph = ip_hdr(skb); | |
239 | if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl))) | |
240 | goto inhdr_error; | |
241 | ||
242 | len = ntohs(iph->tot_len); | |
243 | if (skb->len < len) { | |
244 | IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_INTRUNCATEDPKTS); | |
245 | goto drop; | |
246 | } else if (len < (iph->ihl*4)) | |
247 | goto inhdr_error; | |
248 | ||
249 | if (pskb_trim_rcsum(skb, len)) { | |
250 | IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_INDISCARDS); | |
251 | goto drop; | |
252 | } | |
253 | ||
f8e9881c | 254 | memset(IPCB(skb), 0, sizeof(struct inet_skb_parm)); |
7677e868 HX |
255 | /* We should really parse IP options here but until |
256 | * somebody who actually uses IP options complains to | |
257 | * us we'll just silently ignore the options because | |
258 | * we're lazy! | |
259 | */ | |
462fb2af BD |
260 | return 0; |
261 | ||
262 | inhdr_error: | |
263 | IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_INHDRERRORS); | |
264 | drop: | |
265 | return -1; | |
266 | } | |
267 | ||
a4611d3b BT |
268 | /* We only check the length. A bridge shouldn't do any hop-by-hop stuff |
269 | * anyway | |
270 | */ | |
271 | static int check_hbh_len(struct sk_buff *skb) | |
272 | { | |
273 | unsigned char *raw = (u8 *)(ipv6_hdr(skb) + 1); | |
274 | u32 pkt_len; | |
275 | const unsigned char *nh = skb_network_header(skb); | |
276 | int off = raw - nh; | |
277 | int len = (raw[1] + 1) << 3; | |
278 | ||
279 | if ((raw + len) - skb->data > skb_headlen(skb)) | |
280 | goto bad; | |
281 | ||
282 | off += 2; | |
283 | len -= 2; | |
284 | ||
285 | while (len > 0) { | |
286 | int optlen = nh[off + 1] + 2; | |
287 | ||
288 | switch (nh[off]) { | |
289 | case IPV6_TLV_PAD1: | |
290 | optlen = 1; | |
291 | break; | |
292 | ||
293 | case IPV6_TLV_PADN: | |
294 | break; | |
295 | ||
296 | case IPV6_TLV_JUMBO: | |
297 | if (nh[off + 1] != 4 || (off & 3) != 2) | |
298 | goto bad; | |
299 | pkt_len = ntohl(*(__be32 *)(nh + off + 2)); | |
300 | if (pkt_len <= IPV6_MAXPLEN || | |
301 | ipv6_hdr(skb)->payload_len) | |
302 | goto bad; | |
303 | if (pkt_len > skb->len - sizeof(struct ipv6hdr)) | |
304 | goto bad; | |
305 | if (pskb_trim_rcsum(skb, | |
306 | pkt_len + sizeof(struct ipv6hdr))) | |
307 | goto bad; | |
308 | nh = skb_network_header(skb); | |
309 | break; | |
310 | default: | |
311 | if (optlen > len) | |
312 | goto bad; | |
313 | break; | |
314 | } | |
315 | off += optlen; | |
316 | len -= optlen; | |
317 | } | |
318 | if (len == 0) | |
319 | return 0; | |
320 | bad: | |
321 | return -1; | |
322 | } | |
323 | ||
efb6de9b BT |
324 | /* Equivalent to br_validate_ipv4 for IPv6 */ |
325 | static int br_validate_ipv6(struct sk_buff *skb) | |
326 | { | |
327 | const struct ipv6hdr *hdr; | |
328 | struct net_device *dev = skb->dev; | |
329 | struct inet6_dev *idev = in6_dev_get(skb->dev); | |
330 | u32 pkt_len; | |
331 | u8 ip6h_len = sizeof(struct ipv6hdr); | |
332 | ||
333 | if (!pskb_may_pull(skb, ip6h_len)) | |
334 | goto inhdr_error; | |
335 | ||
336 | if (skb->len < ip6h_len) | |
337 | goto drop; | |
338 | ||
339 | hdr = ipv6_hdr(skb); | |
340 | ||
341 | if (hdr->version != 6) | |
342 | goto inhdr_error; | |
343 | ||
344 | pkt_len = ntohs(hdr->payload_len); | |
345 | ||
346 | if (pkt_len || hdr->nexthdr != NEXTHDR_HOP) { | |
347 | if (pkt_len + ip6h_len > skb->len) { | |
348 | IP6_INC_STATS_BH(dev_net(dev), idev, | |
349 | IPSTATS_MIB_INTRUNCATEDPKTS); | |
350 | goto drop; | |
351 | } | |
352 | if (pskb_trim_rcsum(skb, pkt_len + ip6h_len)) { | |
353 | IP6_INC_STATS_BH(dev_net(dev), idev, | |
354 | IPSTATS_MIB_INDISCARDS); | |
355 | goto drop; | |
356 | } | |
357 | } | |
358 | if (hdr->nexthdr == NEXTHDR_HOP && check_hbh_len(skb)) | |
359 | goto drop; | |
360 | ||
361 | memset(IP6CB(skb), 0, sizeof(struct inet6_skb_parm)); | |
362 | /* No IP options in IPv6 header; however it should be | |
363 | * checked if some next headers need special treatment | |
364 | */ | |
365 | return 0; | |
366 | ||
367 | inhdr_error: | |
368 | IP6_INC_STATS_BH(dev_net(dev), idev, IPSTATS_MIB_INHDRERRORS); | |
369 | drop: | |
370 | return -1; | |
371 | } | |
372 | ||
4a9d2f20 FW |
373 | static void nf_bridge_update_protocol(struct sk_buff *skb) |
374 | { | |
3eaf4025 FW |
375 | switch (skb->nf_bridge->orig_proto) { |
376 | case BRNF_PROTO_8021Q: | |
4a9d2f20 | 377 | skb->protocol = htons(ETH_P_8021Q); |
3eaf4025 FW |
378 | break; |
379 | case BRNF_PROTO_PPPOE: | |
4a9d2f20 | 380 | skb->protocol = htons(ETH_P_PPP_SES); |
3eaf4025 FW |
381 | break; |
382 | case BRNF_PROTO_UNCHANGED: | |
383 | break; | |
384 | } | |
4a9d2f20 FW |
385 | } |
386 | ||
e179e632 BDS |
387 | /* Obtain the correct destination MAC address, while preserving the original |
388 | * source MAC address. If we already know this address, we just copy it. If we | |
389 | * don't, we use the neighbour framework to find out. In both cases, we make | |
390 | * sure that br_handle_frame_finish() is called afterwards. | |
391 | */ | |
7026b1dd | 392 | static int br_nf_pre_routing_finish_bridge(struct sock *sk, struct sk_buff *skb) |
e179e632 | 393 | { |
f6b72b62 | 394 | struct neighbour *neigh; |
e179e632 BDS |
395 | struct dst_entry *dst; |
396 | ||
397 | skb->dev = bridge_parent(skb->dev); | |
398 | if (!skb->dev) | |
399 | goto free_skb; | |
400 | dst = skb_dst(skb); | |
f9d75166 DM |
401 | neigh = dst_neigh_lookup_skb(dst, skb); |
402 | if (neigh) { | |
38330783 | 403 | struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb); |
f9d75166 DM |
404 | int ret; |
405 | ||
406 | if (neigh->hh.hh_len) { | |
407 | neigh_hh_bridge(&neigh->hh, skb); | |
408 | skb->dev = nf_bridge->physindev; | |
7026b1dd | 409 | ret = br_handle_frame_finish(sk, skb); |
f9d75166 DM |
410 | } else { |
411 | /* the neighbour function below overwrites the complete | |
412 | * MAC header, so we save the Ethernet source address and | |
413 | * protocol number. | |
414 | */ | |
415 | skb_copy_from_linear_data_offset(skb, | |
416 | -(ETH_HLEN-ETH_ALEN), | |
e70deecb | 417 | nf_bridge->neigh_header, |
f9d75166 DM |
418 | ETH_HLEN-ETH_ALEN); |
419 | /* tell br_dev_xmit to continue with forwarding */ | |
420 | nf_bridge->mask |= BRNF_BRIDGED_DNAT; | |
93fdd47e | 421 | /* FIXME Need to refragment */ |
f9d75166 DM |
422 | ret = neigh->output(neigh, skb); |
423 | } | |
424 | neigh_release(neigh); | |
425 | return ret; | |
e179e632 BDS |
426 | } |
427 | free_skb: | |
428 | kfree_skb(skb); | |
429 | return 0; | |
430 | } | |
431 | ||
faecbb45 FW |
432 | static bool daddr_was_changed(const struct sk_buff *skb, |
433 | const struct nf_bridge_info *nf_bridge) | |
c055d5b0 | 434 | { |
72b31f72 BT |
435 | switch (skb->protocol) { |
436 | case htons(ETH_P_IP): | |
437 | return ip_hdr(skb)->daddr != nf_bridge->ipv4_daddr; | |
438 | case htons(ETH_P_IPV6): | |
439 | return memcmp(&nf_bridge->ipv6_daddr, &ipv6_hdr(skb)->daddr, | |
440 | sizeof(ipv6_hdr(skb)->daddr)) != 0; | |
441 | default: | |
442 | return false; | |
443 | } | |
c055d5b0 FW |
444 | } |
445 | ||
72b31f72 BT |
446 | /* PF_BRIDGE/PRE_ROUTING: Undo the changes made for ip6tables |
447 | * PREROUTING and continue the bridge PRE_ROUTING hook. See comment | |
448 | * for br_nf_pre_routing_finish(), same logic is used here but | |
449 | * equivalent IPv6 function ip6_route_input() called indirectly. | |
8cae308d BT |
450 | */ |
451 | static int br_nf_pre_routing_finish_ipv6(struct sock *sk, struct sk_buff *skb) | |
452 | { | |
453 | struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb); | |
454 | struct rtable *rt; | |
72b31f72 BT |
455 | struct net_device *dev = skb->dev; |
456 | const struct nf_ipv6_ops *v6ops = nf_get_ipv6_ops(); | |
8cae308d | 457 | |
efb6de9b BT |
458 | nf_bridge->frag_max_size = IP6CB(skb)->frag_max_size; |
459 | ||
8cae308d BT |
460 | if (nf_bridge->pkt_otherhost) { |
461 | skb->pkt_type = PACKET_OTHERHOST; | |
462 | nf_bridge->pkt_otherhost = false; | |
463 | } | |
464 | nf_bridge->mask &= ~BRNF_NF_BRIDGE_PREROUTING; | |
72b31f72 BT |
465 | if (daddr_was_changed(skb, nf_bridge)) { |
466 | skb_dst_drop(skb); | |
467 | v6ops->route_input(skb); | |
8cae308d | 468 | |
72b31f72 BT |
469 | if (skb_dst(skb)->error) { |
470 | kfree_skb(skb); | |
471 | return 0; | |
472 | } | |
473 | ||
474 | if (skb_dst(skb)->dev == dev) { | |
475 | skb->dev = nf_bridge->physindev; | |
476 | nf_bridge_update_protocol(skb); | |
477 | nf_bridge_push_encap_header(skb); | |
478 | NF_HOOK_THRESH(NFPROTO_BRIDGE, NF_BR_PRE_ROUTING, | |
479 | sk, skb, skb->dev, NULL, | |
480 | br_nf_pre_routing_finish_bridge, | |
481 | 1); | |
482 | return 0; | |
483 | } | |
484 | ether_addr_copy(eth_hdr(skb)->h_dest, dev->dev_addr); | |
485 | skb->pkt_type = PACKET_HOST; | |
486 | } else { | |
487 | rt = bridge_parent_rtable(nf_bridge->physindev); | |
488 | if (!rt) { | |
489 | kfree_skb(skb); | |
490 | return 0; | |
491 | } | |
492 | skb_dst_set_noref(skb, &rt->dst); | |
8cae308d | 493 | } |
8cae308d BT |
494 | |
495 | skb->dev = nf_bridge->physindev; | |
496 | nf_bridge_update_protocol(skb); | |
497 | nf_bridge_push_encap_header(skb); | |
498 | NF_HOOK_THRESH(NFPROTO_BRIDGE, NF_BR_PRE_ROUTING, sk, skb, | |
499 | skb->dev, NULL, | |
500 | br_handle_frame_finish, 1); | |
501 | ||
502 | return 0; | |
503 | } | |
504 | ||
1da177e4 | 505 | /* This requires some explaining. If DNAT has taken place, |
ea2d9b41 | 506 | * we will need to fix up the destination Ethernet address. |
faecbb45 | 507 | * This is also true when SNAT takes place (for the reply direction). |
1da177e4 LT |
508 | * |
509 | * There are two cases to consider: | |
510 | * 1. The packet was DNAT'ed to a device in the same bridge | |
511 | * port group as it was received on. We can still bridge | |
512 | * the packet. | |
513 | * 2. The packet was DNAT'ed to a different device, either | |
514 | * a non-bridged device or another bridge port group. | |
515 | * The packet will need to be routed. | |
516 | * | |
517 | * The correct way of distinguishing between these two cases is to | |
518 | * call ip_route_input() and to look at skb->dst->dev, which is | |
519 | * changed to the destination device if ip_route_input() succeeds. | |
520 | * | |
ea2d9b41 | 521 | * Let's first consider the case that ip_route_input() succeeds: |
1da177e4 | 522 | * |
ea2d9b41 BDS |
523 | * If the output device equals the logical bridge device the packet |
524 | * came in on, we can consider this bridging. The corresponding MAC | |
525 | * address will be obtained in br_nf_pre_routing_finish_bridge. | |
1da177e4 LT |
526 | * Otherwise, the packet is considered to be routed and we just |
527 | * change the destination MAC address so that the packet will | |
f216f082 BDS |
528 | * later be passed up to the IP stack to be routed. For a redirected |
529 | * packet, ip_route_input() will give back the localhost as output device, | |
530 | * which differs from the bridge device. | |
1da177e4 | 531 | * |
ea2d9b41 | 532 | * Let's now consider the case that ip_route_input() fails: |
1da177e4 | 533 | * |
f216f082 BDS |
534 | * This can be because the destination address is martian, in which case |
535 | * the packet will be dropped. | |
ea2d9b41 BDS |
536 | * If IP forwarding is disabled, ip_route_input() will fail, while |
537 | * ip_route_output_key() can return success. The source | |
538 | * address for ip_route_output_key() is set to zero, so ip_route_output_key() | |
1da177e4 | 539 | * thinks we're handling a locally generated packet and won't care |
ea2d9b41 BDS |
540 | * if IP forwarding is enabled. If the output device equals the logical bridge |
541 | * device, we proceed as if ip_route_input() succeeded. If it differs from the | |
542 | * logical bridge port or if ip_route_output_key() fails we drop the packet. | |
543 | */ | |
7026b1dd | 544 | static int br_nf_pre_routing_finish(struct sock *sk, struct sk_buff *skb) |
1da177e4 LT |
545 | { |
546 | struct net_device *dev = skb->dev; | |
eddc9ec5 | 547 | struct iphdr *iph = ip_hdr(skb); |
38330783 | 548 | struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb); |
511c3f92 | 549 | struct rtable *rt; |
f216f082 | 550 | int err; |
93fdd47e | 551 | |
411ffb4f | 552 | nf_bridge->frag_max_size = IPCB(skb)->frag_max_size; |
1da177e4 | 553 | |
a1e67951 | 554 | if (nf_bridge->pkt_otherhost) { |
1da177e4 | 555 | skb->pkt_type = PACKET_OTHERHOST; |
a1e67951 | 556 | nf_bridge->pkt_otherhost = false; |
1da177e4 | 557 | } |
d39a33ed | 558 | nf_bridge->mask &= ~BRNF_NF_BRIDGE_PREROUTING; |
faecbb45 | 559 | if (daddr_was_changed(skb, nf_bridge)) { |
f216f082 | 560 | if ((err = ip_route_input(skb, iph->daddr, iph->saddr, iph->tos, dev))) { |
f3abc9b9 | 561 | struct in_device *in_dev = __in_dev_get_rcu(dev); |
f216f082 BDS |
562 | |
563 | /* If err equals -EHOSTUNREACH the error is due to a | |
564 | * martian destination or due to the fact that | |
565 | * forwarding is disabled. For most martian packets, | |
566 | * ip_route_output_key() will fail. It won't fail for 2 types of | |
567 | * martian destinations: loopback destinations and destination | |
568 | * 0.0.0.0. In both cases the packet will be dropped because the | |
569 | * destination is the loopback device and not the bridge. */ | |
570 | if (err != -EHOSTUNREACH || !in_dev || IN_DEV_FORWARD(in_dev)) | |
571 | goto free_skb; | |
1da177e4 | 572 | |
78fbfd8a DM |
573 | rt = ip_route_output(dev_net(dev), iph->daddr, 0, |
574 | RT_TOS(iph->tos), 0); | |
b23dd4fe | 575 | if (!IS_ERR(rt)) { |
1c011bed | 576 | /* - Bridged-and-DNAT'ed traffic doesn't |
f216f082 | 577 | * require ip_forwarding. */ |
b23dd4fe DM |
578 | if (rt->dst.dev == dev) { |
579 | skb_dst_set(skb, &rt->dst); | |
1da177e4 LT |
580 | goto bridged_dnat; |
581 | } | |
b23dd4fe | 582 | ip_rt_put(rt); |
1da177e4 | 583 | } |
f216f082 | 584 | free_skb: |
1da177e4 LT |
585 | kfree_skb(skb); |
586 | return 0; | |
587 | } else { | |
adf30907 | 588 | if (skb_dst(skb)->dev == dev) { |
1da177e4 | 589 | bridged_dnat: |
1da177e4 | 590 | skb->dev = nf_bridge->physindev; |
e179e632 | 591 | nf_bridge_update_protocol(skb); |
fc38582d | 592 | nf_bridge_push_encap_header(skb); |
713aefa3 JE |
593 | NF_HOOK_THRESH(NFPROTO_BRIDGE, |
594 | NF_BR_PRE_ROUTING, | |
7026b1dd | 595 | sk, skb, skb->dev, NULL, |
1da177e4 LT |
596 | br_nf_pre_routing_finish_bridge, |
597 | 1); | |
598 | return 0; | |
599 | } | |
04091142 | 600 | ether_addr_copy(eth_hdr(skb)->h_dest, dev->dev_addr); |
1da177e4 LT |
601 | skb->pkt_type = PACKET_HOST; |
602 | } | |
603 | } else { | |
511c3f92 ED |
604 | rt = bridge_parent_rtable(nf_bridge->physindev); |
605 | if (!rt) { | |
4adf0af6 SW |
606 | kfree_skb(skb); |
607 | return 0; | |
608 | } | |
f9181f4f | 609 | skb_dst_set_noref(skb, &rt->dst); |
1da177e4 LT |
610 | } |
611 | ||
612 | skb->dev = nf_bridge->physindev; | |
e179e632 | 613 | nf_bridge_update_protocol(skb); |
fc38582d | 614 | nf_bridge_push_encap_header(skb); |
7026b1dd DM |
615 | NF_HOOK_THRESH(NFPROTO_BRIDGE, NF_BR_PRE_ROUTING, sk, skb, |
616 | skb->dev, NULL, | |
1da177e4 LT |
617 | br_handle_frame_finish, 1); |
618 | ||
619 | return 0; | |
620 | } | |
621 | ||
4981682c PNA |
622 | static struct net_device *brnf_get_logical_dev(struct sk_buff *skb, const struct net_device *dev) |
623 | { | |
624 | struct net_device *vlan, *br; | |
625 | ||
626 | br = bridge_parent(dev); | |
df8a39de | 627 | if (brnf_pass_vlan_indev == 0 || !skb_vlan_tag_present(skb)) |
4981682c PNA |
628 | return br; |
629 | ||
f06c7f9f | 630 | vlan = __vlan_find_dev_deep_rcu(br, skb->vlan_proto, |
df8a39de | 631 | skb_vlan_tag_get(skb) & VLAN_VID_MASK); |
4981682c PNA |
632 | |
633 | return vlan ? vlan : br; | |
634 | } | |
635 | ||
1da177e4 | 636 | /* Some common code for IPv4/IPv6 */ |
5dce971a | 637 | static struct net_device *setup_pre_routing(struct sk_buff *skb) |
1da177e4 | 638 | { |
38330783 | 639 | struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb); |
1da177e4 LT |
640 | |
641 | if (skb->pkt_type == PACKET_OTHERHOST) { | |
642 | skb->pkt_type = PACKET_HOST; | |
a1e67951 | 643 | nf_bridge->pkt_otherhost = true; |
1da177e4 LT |
644 | } |
645 | ||
646 | nf_bridge->mask |= BRNF_NF_BRIDGE_PREROUTING; | |
647 | nf_bridge->physindev = skb->dev; | |
4981682c | 648 | skb->dev = brnf_get_logical_dev(skb, skb->dev); |
3eaf4025 | 649 | |
e179e632 | 650 | if (skb->protocol == htons(ETH_P_8021Q)) |
3eaf4025 | 651 | nf_bridge->orig_proto = BRNF_PROTO_8021Q; |
e179e632 | 652 | else if (skb->protocol == htons(ETH_P_PPP_SES)) |
3eaf4025 | 653 | nf_bridge->orig_proto = BRNF_PROTO_PPPOE; |
5dce971a | 654 | |
6b8dbcf2 FW |
655 | /* Must drop socket now because of tproxy. */ |
656 | skb_orphan(skb); | |
5dce971a | 657 | return skb->dev; |
1da177e4 LT |
658 | } |
659 | ||
1da177e4 | 660 | /* Replicate the checks that IPv6 does on packet reception and pass the packet |
efb6de9b BT |
661 | * to ip6tables. |
662 | */ | |
795aa6ef | 663 | static unsigned int br_nf_pre_routing_ipv6(const struct nf_hook_ops *ops, |
789bc3e5 | 664 | struct sk_buff *skb, |
238e54c9 | 665 | const struct nf_hook_state *state) |
1da177e4 | 666 | { |
72b31f72 | 667 | struct nf_bridge_info *nf_bridge; |
1da177e4 | 668 | |
efb6de9b | 669 | if (br_validate_ipv6(skb)) |
deef4b52 | 670 | return NF_DROP; |
1da177e4 | 671 | |
789bc3e5 | 672 | nf_bridge_put(skb->nf_bridge); |
fdeabdef | 673 | if (!nf_bridge_alloc(skb)) |
1da177e4 | 674 | return NF_DROP; |
5dce971a SH |
675 | if (!setup_pre_routing(skb)) |
676 | return NF_DROP; | |
1da177e4 | 677 | |
72b31f72 BT |
678 | nf_bridge = nf_bridge_info_get(skb); |
679 | nf_bridge->ipv6_daddr = ipv6_hdr(skb)->daddr; | |
680 | ||
e179e632 | 681 | skb->protocol = htons(ETH_P_IPV6); |
7026b1dd DM |
682 | NF_HOOK(NFPROTO_IPV6, NF_INET_PRE_ROUTING, state->sk, skb, |
683 | skb->dev, NULL, | |
1da177e4 LT |
684 | br_nf_pre_routing_finish_ipv6); |
685 | ||
686 | return NF_STOLEN; | |
1da177e4 LT |
687 | } |
688 | ||
689 | /* Direct IPv6 traffic to br_nf_pre_routing_ipv6. | |
690 | * Replicate the checks that IPv4 does on packet reception. | |
691 | * Set skb->dev to the bridge device (i.e. parent of the | |
692 | * receiving device) to make netfilter happy, the REDIRECT | |
693 | * target in particular. Save the original destination IP | |
694 | * address to be able to detect DNAT afterwards. */ | |
795aa6ef PM |
695 | static unsigned int br_nf_pre_routing(const struct nf_hook_ops *ops, |
696 | struct sk_buff *skb, | |
238e54c9 | 697 | const struct nf_hook_state *state) |
1da177e4 | 698 | { |
faecbb45 | 699 | struct nf_bridge_info *nf_bridge; |
4df53d8b PM |
700 | struct net_bridge_port *p; |
701 | struct net_bridge *br; | |
e7c243c9 EP |
702 | __u32 len = nf_bridge_encap_header_len(skb); |
703 | ||
e7c243c9 | 704 | if (unlikely(!pskb_may_pull(skb, len))) |
deef4b52 | 705 | return NF_DROP; |
1da177e4 | 706 | |
238e54c9 | 707 | p = br_port_get_rcu(state->in); |
4df53d8b | 708 | if (p == NULL) |
deef4b52 | 709 | return NF_DROP; |
4df53d8b PM |
710 | br = p->br; |
711 | ||
739e4505 | 712 | if (IS_IPV6(skb) || IS_VLAN_IPV6(skb) || IS_PPPOE_IPV6(skb)) { |
4df53d8b | 713 | if (!brnf_call_ip6tables && !br->nf_call_ip6tables) |
1da177e4 | 714 | return NF_ACCEPT; |
4df53d8b | 715 | |
fc38582d | 716 | nf_bridge_pull_encap_header_rcsum(skb); |
238e54c9 | 717 | return br_nf_pre_routing_ipv6(ops, skb, state); |
1da177e4 | 718 | } |
4df53d8b PM |
719 | |
720 | if (!brnf_call_iptables && !br->nf_call_iptables) | |
1da177e4 | 721 | return NF_ACCEPT; |
1da177e4 | 722 | |
739e4505 | 723 | if (!IS_IP(skb) && !IS_VLAN_IP(skb) && !IS_PPPOE_IP(skb)) |
1da177e4 LT |
724 | return NF_ACCEPT; |
725 | ||
fc38582d | 726 | nf_bridge_pull_encap_header_rcsum(skb); |
1da177e4 | 727 | |
77d574e7 | 728 | if (br_validate_ipv4(skb)) |
deef4b52 | 729 | return NF_DROP; |
17762060 | 730 | |
789bc3e5 | 731 | nf_bridge_put(skb->nf_bridge); |
fdeabdef | 732 | if (!nf_bridge_alloc(skb)) |
1da177e4 | 733 | return NF_DROP; |
5dce971a SH |
734 | if (!setup_pre_routing(skb)) |
735 | return NF_DROP; | |
c055d5b0 | 736 | |
faecbb45 FW |
737 | nf_bridge = nf_bridge_info_get(skb); |
738 | nf_bridge->ipv4_daddr = ip_hdr(skb)->daddr; | |
739 | ||
e179e632 | 740 | skb->protocol = htons(ETH_P_IP); |
1da177e4 | 741 | |
7026b1dd DM |
742 | NF_HOOK(NFPROTO_IPV4, NF_INET_PRE_ROUTING, state->sk, skb, |
743 | skb->dev, NULL, | |
1da177e4 LT |
744 | br_nf_pre_routing_finish); |
745 | ||
746 | return NF_STOLEN; | |
1da177e4 LT |
747 | } |
748 | ||
749 | ||
750 | /* PF_BRIDGE/LOCAL_IN ************************************************/ | |
751 | /* The packet is locally destined, which requires a real | |
752 | * dst_entry, so detach the fake one. On the way up, the | |
753 | * packet would pass through PRE_ROUTING again (which already | |
754 | * took place when the packet entered the bridge), but we | |
755 | * register an IPv4 PRE_ROUTING 'sabotage' hook that will | |
756 | * prevent this from happening. */ | |
795aa6ef PM |
757 | static unsigned int br_nf_local_in(const struct nf_hook_ops *ops, |
758 | struct sk_buff *skb, | |
238e54c9 | 759 | const struct nf_hook_state *state) |
1da177e4 | 760 | { |
a881e963 | 761 | br_drop_fake_rtable(skb); |
1da177e4 LT |
762 | return NF_ACCEPT; |
763 | } | |
764 | ||
1da177e4 | 765 | /* PF_BRIDGE/FORWARD *************************************************/ |
7026b1dd | 766 | static int br_nf_forward_finish(struct sock *sk, struct sk_buff *skb) |
1da177e4 | 767 | { |
38330783 | 768 | struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb); |
1da177e4 | 769 | struct net_device *in; |
1da177e4 | 770 | |
739e4505 | 771 | if (!IS_ARP(skb) && !IS_VLAN_ARP(skb)) { |
0b67c43c | 772 | |
efb6de9b | 773 | if (skb->protocol == htons(ETH_P_IP)) |
411ffb4f | 774 | nf_bridge->frag_max_size = IPCB(skb)->frag_max_size; |
efb6de9b BT |
775 | |
776 | if (skb->protocol == htons(ETH_P_IPV6)) | |
777 | nf_bridge->frag_max_size = IP6CB(skb)->frag_max_size; | |
0b67c43c | 778 | |
1da177e4 | 779 | in = nf_bridge->physindev; |
a1e67951 | 780 | if (nf_bridge->pkt_otherhost) { |
1da177e4 | 781 | skb->pkt_type = PACKET_OTHERHOST; |
a1e67951 | 782 | nf_bridge->pkt_otherhost = false; |
1da177e4 | 783 | } |
e94c6743 | 784 | nf_bridge_update_protocol(skb); |
1da177e4 LT |
785 | } else { |
786 | in = *((struct net_device **)(skb->cb)); | |
787 | } | |
fc38582d | 788 | nf_bridge_push_encap_header(skb); |
e179e632 | 789 | |
7026b1dd DM |
790 | NF_HOOK_THRESH(NFPROTO_BRIDGE, NF_BR_FORWARD, sk, skb, |
791 | in, skb->dev, br_forward_finish, 1); | |
1da177e4 LT |
792 | return 0; |
793 | } | |
794 | ||
739e4505 | 795 | |
1da177e4 LT |
796 | /* This is the 'purely bridged' case. For IP, we pass the packet to |
797 | * netfilter with indev and outdev set to the bridge device, | |
798 | * but we are still able to filter on the 'real' indev/outdev | |
799 | * because of the physdev module. For ARP, indev and outdev are the | |
800 | * bridge ports. */ | |
795aa6ef PM |
801 | static unsigned int br_nf_forward_ip(const struct nf_hook_ops *ops, |
802 | struct sk_buff *skb, | |
238e54c9 | 803 | const struct nf_hook_state *state) |
1da177e4 | 804 | { |
1da177e4 | 805 | struct nf_bridge_info *nf_bridge; |
5dce971a | 806 | struct net_device *parent; |
76108cea | 807 | u_int8_t pf; |
1da177e4 LT |
808 | |
809 | if (!skb->nf_bridge) | |
810 | return NF_ACCEPT; | |
811 | ||
2dc2f207 PM |
812 | /* Need exclusive nf_bridge_info since we might have multiple |
813 | * different physoutdevs. */ | |
814 | if (!nf_bridge_unshare(skb)) | |
815 | return NF_DROP; | |
816 | ||
38330783 FW |
817 | nf_bridge = nf_bridge_info_get(skb); |
818 | if (!nf_bridge) | |
819 | return NF_DROP; | |
820 | ||
238e54c9 | 821 | parent = bridge_parent(state->out); |
5dce971a SH |
822 | if (!parent) |
823 | return NF_DROP; | |
824 | ||
739e4505 | 825 | if (IS_IP(skb) || IS_VLAN_IP(skb) || IS_PPPOE_IP(skb)) |
aa740f46 | 826 | pf = NFPROTO_IPV4; |
739e4505 | 827 | else if (IS_IPV6(skb) || IS_VLAN_IPV6(skb) || IS_PPPOE_IPV6(skb)) |
aa740f46 | 828 | pf = NFPROTO_IPV6; |
a2bd40ad HX |
829 | else |
830 | return NF_ACCEPT; | |
1da177e4 | 831 | |
3db05fea | 832 | nf_bridge_pull_encap_header(skb); |
1da177e4 | 833 | |
1da177e4 LT |
834 | if (skb->pkt_type == PACKET_OTHERHOST) { |
835 | skb->pkt_type = PACKET_HOST; | |
a1e67951 | 836 | nf_bridge->pkt_otherhost = true; |
1da177e4 LT |
837 | } |
838 | ||
0b67c43c | 839 | if (pf == NFPROTO_IPV4) { |
77d574e7 | 840 | if (br_validate_ipv4(skb)) |
0b67c43c | 841 | return NF_DROP; |
411ffb4f | 842 | IPCB(skb)->frag_max_size = nf_bridge->frag_max_size; |
0b67c43c | 843 | } |
6b1e960f | 844 | |
efb6de9b BT |
845 | if (pf == NFPROTO_IPV6) { |
846 | if (br_validate_ipv6(skb)) | |
847 | return NF_DROP; | |
848 | IP6CB(skb)->frag_max_size = nf_bridge->frag_max_size; | |
849 | } | |
850 | ||
1da177e4 | 851 | nf_bridge->physoutdev = skb->dev; |
aa740f46 | 852 | if (pf == NFPROTO_IPV4) |
e179e632 BDS |
853 | skb->protocol = htons(ETH_P_IP); |
854 | else | |
855 | skb->protocol = htons(ETH_P_IPV6); | |
1da177e4 | 856 | |
7026b1dd DM |
857 | NF_HOOK(pf, NF_INET_FORWARD, NULL, skb, |
858 | brnf_get_logical_dev(skb, state->in), | |
238e54c9 | 859 | parent, br_nf_forward_finish); |
1da177e4 LT |
860 | |
861 | return NF_STOLEN; | |
862 | } | |
863 | ||
795aa6ef PM |
864 | static unsigned int br_nf_forward_arp(const struct nf_hook_ops *ops, |
865 | struct sk_buff *skb, | |
238e54c9 | 866 | const struct nf_hook_state *state) |
1da177e4 | 867 | { |
4df53d8b PM |
868 | struct net_bridge_port *p; |
869 | struct net_bridge *br; | |
1da177e4 LT |
870 | struct net_device **d = (struct net_device **)(skb->cb); |
871 | ||
238e54c9 | 872 | p = br_port_get_rcu(state->out); |
4df53d8b PM |
873 | if (p == NULL) |
874 | return NF_ACCEPT; | |
875 | br = p->br; | |
876 | ||
877 | if (!brnf_call_arptables && !br->nf_call_arptables) | |
1da177e4 | 878 | return NF_ACCEPT; |
1da177e4 | 879 | |
739e4505 | 880 | if (!IS_ARP(skb)) { |
8b42ec39 | 881 | if (!IS_VLAN_ARP(skb)) |
1da177e4 | 882 | return NF_ACCEPT; |
3db05fea | 883 | nf_bridge_pull_encap_header(skb); |
1da177e4 LT |
884 | } |
885 | ||
d0a92be0 | 886 | if (arp_hdr(skb)->ar_pln != 4) { |
fc38582d | 887 | if (IS_VLAN_ARP(skb)) |
3db05fea | 888 | nf_bridge_push_encap_header(skb); |
1da177e4 LT |
889 | return NF_ACCEPT; |
890 | } | |
238e54c9 | 891 | *d = state->in; |
7026b1dd DM |
892 | NF_HOOK(NFPROTO_ARP, NF_ARP_FORWARD, state->sk, skb, |
893 | state->in, state->out, br_nf_forward_finish); | |
1da177e4 LT |
894 | |
895 | return NF_STOLEN; | |
896 | } | |
897 | ||
efb6de9b | 898 | #if IS_ENABLED(CONFIG_NF_DEFRAG_IPV4) || IS_ENABLED(CONFIG_NF_DEFRAG_IPV6) |
aadd51aa | 899 | static int br_nf_push_frag_xmit(struct sock *sk, struct sk_buff *skb) |
8bd63cf1 | 900 | { |
e70deecb | 901 | struct brnf_frag_data *data; |
8bd63cf1 | 902 | int err; |
8bd63cf1 | 903 | |
e70deecb FW |
904 | data = this_cpu_ptr(&brnf_frag_data_storage); |
905 | err = skb_cow_head(skb, data->size); | |
8bd63cf1 | 906 | |
e70deecb | 907 | if (err) { |
8bd63cf1 FW |
908 | kfree_skb(skb); |
909 | return 0; | |
910 | } | |
911 | ||
e70deecb FW |
912 | skb_copy_to_linear_data_offset(skb, -data->size, data->mac, data->size); |
913 | __skb_push(skb, data->encap_size); | |
914 | ||
a9fcc6a4 | 915 | nf_bridge_info_free(skb); |
7026b1dd | 916 | return br_dev_queue_push_xmit(sk, skb); |
8bd63cf1 | 917 | } |
efb6de9b | 918 | #endif |
8bd63cf1 | 919 | |
49d16b23 AZ |
920 | static int br_nf_ip_fragment(struct sock *sk, struct sk_buff *skb, |
921 | int (*output)(struct sock *, struct sk_buff *)) | |
922 | { | |
923 | unsigned int mtu = ip_skb_dst_mtu(skb); | |
924 | struct iphdr *iph = ip_hdr(skb); | |
925 | struct rtable *rt = skb_rtable(skb); | |
926 | struct net_device *dev = rt->dst.dev; | |
927 | ||
928 | if (unlikely(((iph->frag_off & htons(IP_DF)) && !skb->ignore_df) || | |
929 | (IPCB(skb)->frag_max_size && | |
930 | IPCB(skb)->frag_max_size > mtu))) { | |
931 | IP_INC_STATS(dev_net(dev), IPSTATS_MIB_FRAGFAILS); | |
932 | kfree_skb(skb); | |
933 | return -EMSGSIZE; | |
934 | } | |
935 | ||
936 | return ip_do_fragment(sk, skb, output); | |
937 | } | |
938 | ||
33b1f313 FW |
939 | static unsigned int nf_bridge_mtu_reduction(const struct sk_buff *skb) |
940 | { | |
941 | if (skb->nf_bridge->orig_proto == BRNF_PROTO_PPPOE) | |
942 | return PPPOE_SES_HLEN; | |
943 | return 0; | |
944 | } | |
945 | ||
7026b1dd | 946 | static int br_nf_dev_queue_xmit(struct sock *sk, struct sk_buff *skb) |
2e2f7aef | 947 | { |
411ffb4f | 948 | struct nf_bridge_info *nf_bridge; |
7a8d831d | 949 | unsigned int mtu_reserved; |
462fb2af | 950 | |
efb6de9b BT |
951 | mtu_reserved = nf_bridge_mtu_reduction(skb); |
952 | ||
953 | if (skb_is_gso(skb) || skb->len + mtu_reserved <= skb->dev->mtu) { | |
a9fcc6a4 | 954 | nf_bridge_info_free(skb); |
7026b1dd | 955 | return br_dev_queue_push_xmit(sk, skb); |
a9fcc6a4 | 956 | } |
7a8d831d | 957 | |
411ffb4f | 958 | nf_bridge = nf_bridge_info_get(skb); |
efb6de9b BT |
959 | |
960 | #if IS_ENABLED(CONFIG_NF_DEFRAG_IPV4) | |
93fdd47e HX |
961 | /* This is wrong! We should preserve the original fragment |
962 | * boundaries by preserving frag_list rather than refragmenting. | |
963 | */ | |
efb6de9b | 964 | if (skb->protocol == htons(ETH_P_IP)) { |
e70deecb FW |
965 | struct brnf_frag_data *data; |
966 | ||
77d574e7 | 967 | if (br_validate_ipv4(skb)) |
462fb2af | 968 | return NF_DROP; |
411ffb4f BT |
969 | |
970 | IPCB(skb)->frag_max_size = nf_bridge->frag_max_size; | |
e70deecb FW |
971 | |
972 | nf_bridge_update_protocol(skb); | |
973 | ||
974 | data = this_cpu_ptr(&brnf_frag_data_storage); | |
975 | data->encap_size = nf_bridge_encap_header_len(skb); | |
976 | data->size = ETH_HLEN + data->encap_size; | |
977 | ||
978 | skb_copy_from_linear_data_offset(skb, -data->size, data->mac, | |
979 | data->size); | |
980 | ||
efb6de9b | 981 | return br_nf_ip_fragment(sk, skb, br_nf_push_frag_xmit); |
e70deecb | 982 | } |
efb6de9b BT |
983 | #endif |
984 | #if IS_ENABLED(CONFIG_NF_DEFRAG_IPV6) | |
985 | if (skb->protocol == htons(ETH_P_IPV6)) { | |
986 | const struct nf_ipv6_ops *v6ops = nf_get_ipv6_ops(); | |
987 | struct brnf_frag_data *data; | |
462fb2af | 988 | |
efb6de9b BT |
989 | if (br_validate_ipv6(skb)) |
990 | return NF_DROP; | |
991 | ||
992 | IP6CB(skb)->frag_max_size = nf_bridge->frag_max_size; | |
993 | ||
994 | nf_bridge_update_protocol(skb); | |
995 | ||
996 | data = this_cpu_ptr(&brnf_frag_data_storage); | |
997 | data->encap_size = nf_bridge_encap_header_len(skb); | |
998 | data->size = ETH_HLEN + data->encap_size; | |
999 | ||
1000 | skb_copy_from_linear_data_offset(skb, -data->size, data->mac, | |
1001 | data->size); | |
1002 | ||
1003 | if (v6ops) | |
1004 | return v6ops->fragment(sk, skb, br_nf_push_frag_xmit); | |
1005 | else | |
1006 | return -EMSGSIZE; | |
1007 | } | |
1008 | #endif | |
a9fcc6a4 FW |
1009 | nf_bridge_info_free(skb); |
1010 | return br_dev_queue_push_xmit(sk, skb); | |
c197facc | 1011 | } |
1da177e4 LT |
1012 | |
1013 | /* PF_BRIDGE/POST_ROUTING ********************************************/ | |
795aa6ef PM |
1014 | static unsigned int br_nf_post_routing(const struct nf_hook_ops *ops, |
1015 | struct sk_buff *skb, | |
238e54c9 | 1016 | const struct nf_hook_state *state) |
1da177e4 | 1017 | { |
38330783 | 1018 | struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb); |
1da177e4 | 1019 | struct net_device *realoutdev = bridge_parent(skb->dev); |
76108cea | 1020 | u_int8_t pf; |
1da177e4 | 1021 | |
e4bb9bcb FW |
1022 | /* if nf_bridge is set, but ->physoutdev is NULL, this packet came in |
1023 | * on a bridge, but was delivered locally and is now being routed: | |
1024 | * | |
1025 | * POST_ROUTING was already invoked from the ip stack. | |
1026 | */ | |
1027 | if (!nf_bridge || !nf_bridge->physoutdev) | |
81d9ddae PM |
1028 | return NF_ACCEPT; |
1029 | ||
5dce971a SH |
1030 | if (!realoutdev) |
1031 | return NF_DROP; | |
1032 | ||
739e4505 | 1033 | if (IS_IP(skb) || IS_VLAN_IP(skb) || IS_PPPOE_IP(skb)) |
aa740f46 | 1034 | pf = NFPROTO_IPV4; |
739e4505 | 1035 | else if (IS_IPV6(skb) || IS_VLAN_IPV6(skb) || IS_PPPOE_IPV6(skb)) |
aa740f46 | 1036 | pf = NFPROTO_IPV6; |
a2bd40ad HX |
1037 | else |
1038 | return NF_ACCEPT; | |
1da177e4 | 1039 | |
1da177e4 LT |
1040 | /* We assume any code from br_dev_queue_push_xmit onwards doesn't care |
1041 | * about the value of skb->pkt_type. */ | |
1042 | if (skb->pkt_type == PACKET_OTHERHOST) { | |
1043 | skb->pkt_type = PACKET_HOST; | |
a1e67951 | 1044 | nf_bridge->pkt_otherhost = true; |
1da177e4 LT |
1045 | } |
1046 | ||
fc38582d | 1047 | nf_bridge_pull_encap_header(skb); |
aa740f46 | 1048 | if (pf == NFPROTO_IPV4) |
e179e632 BDS |
1049 | skb->protocol = htons(ETH_P_IP); |
1050 | else | |
1051 | skb->protocol = htons(ETH_P_IPV6); | |
1da177e4 | 1052 | |
7026b1dd DM |
1053 | NF_HOOK(pf, NF_INET_POST_ROUTING, state->sk, skb, |
1054 | NULL, realoutdev, | |
2e2f7aef | 1055 | br_nf_dev_queue_xmit); |
1da177e4 LT |
1056 | |
1057 | return NF_STOLEN; | |
1da177e4 LT |
1058 | } |
1059 | ||
1da177e4 LT |
1060 | /* IP/SABOTAGE *****************************************************/ |
1061 | /* Don't hand locally destined packets to PF_INET(6)/PRE_ROUTING | |
1062 | * for the second time. */ | |
795aa6ef PM |
1063 | static unsigned int ip_sabotage_in(const struct nf_hook_ops *ops, |
1064 | struct sk_buff *skb, | |
238e54c9 | 1065 | const struct nf_hook_state *state) |
1da177e4 | 1066 | { |
3db05fea HX |
1067 | if (skb->nf_bridge && |
1068 | !(skb->nf_bridge->mask & BRNF_NF_BRIDGE_PREROUTING)) { | |
1da177e4 LT |
1069 | return NF_STOP; |
1070 | } | |
1071 | ||
1072 | return NF_ACCEPT; | |
1073 | } | |
1074 | ||
e5de75bf PNA |
1075 | /* This is called when br_netfilter has called into iptables/netfilter, |
1076 | * and DNAT has taken place on a bridge-forwarded packet. | |
1077 | * | |
1078 | * neigh->output has created a new MAC header, with local br0 MAC | |
1079 | * as saddr. | |
1080 | * | |
1081 | * This restores the original MAC saddr of the bridged packet | |
1082 | * before invoking bridge forward logic to transmit the packet. | |
1083 | */ | |
1084 | static void br_nf_pre_routing_finish_bridge_slow(struct sk_buff *skb) | |
1085 | { | |
38330783 | 1086 | struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb); |
e5de75bf PNA |
1087 | |
1088 | skb_pull(skb, ETH_HLEN); | |
1089 | nf_bridge->mask &= ~BRNF_BRIDGED_DNAT; | |
1090 | ||
e70deecb FW |
1091 | BUILD_BUG_ON(sizeof(nf_bridge->neigh_header) != (ETH_HLEN - ETH_ALEN)); |
1092 | ||
1093 | skb_copy_to_linear_data_offset(skb, -(ETH_HLEN - ETH_ALEN), | |
1094 | nf_bridge->neigh_header, | |
1095 | ETH_HLEN - ETH_ALEN); | |
e5de75bf | 1096 | skb->dev = nf_bridge->physindev; |
7fb48c5b FW |
1097 | |
1098 | nf_bridge->physoutdev = NULL; | |
7026b1dd | 1099 | br_handle_frame_finish(NULL, skb); |
e5de75bf PNA |
1100 | } |
1101 | ||
1a4ba64d | 1102 | static int br_nf_dev_xmit(struct sk_buff *skb) |
e5de75bf PNA |
1103 | { |
1104 | if (skb->nf_bridge && (skb->nf_bridge->mask & BRNF_BRIDGED_DNAT)) { | |
1105 | br_nf_pre_routing_finish_bridge_slow(skb); | |
1106 | return 1; | |
1107 | } | |
1108 | return 0; | |
1109 | } | |
1a4ba64d PNA |
1110 | |
1111 | static const struct nf_br_ops br_ops = { | |
1112 | .br_dev_xmit_hook = br_nf_dev_xmit, | |
1113 | }; | |
e5de75bf | 1114 | |
4b7fd5d9 PNA |
1115 | void br_netfilter_enable(void) |
1116 | { | |
1117 | } | |
1118 | EXPORT_SYMBOL_GPL(br_netfilter_enable); | |
1119 | ||
ea2d9b41 BDS |
1120 | /* For br_nf_post_routing, we need (prio = NF_BR_PRI_LAST), because |
1121 | * br_dev_queue_push_xmit is called afterwards */ | |
1999414a | 1122 | static struct nf_hook_ops br_nf_ops[] __read_mostly = { |
1490fd89 CG |
1123 | { |
1124 | .hook = br_nf_pre_routing, | |
1125 | .owner = THIS_MODULE, | |
aa740f46 | 1126 | .pf = NFPROTO_BRIDGE, |
1490fd89 CG |
1127 | .hooknum = NF_BR_PRE_ROUTING, |
1128 | .priority = NF_BR_PRI_BRNF, | |
1129 | }, | |
1130 | { | |
1131 | .hook = br_nf_local_in, | |
1132 | .owner = THIS_MODULE, | |
aa740f46 | 1133 | .pf = NFPROTO_BRIDGE, |
1490fd89 CG |
1134 | .hooknum = NF_BR_LOCAL_IN, |
1135 | .priority = NF_BR_PRI_BRNF, | |
1136 | }, | |
1137 | { | |
1138 | .hook = br_nf_forward_ip, | |
1139 | .owner = THIS_MODULE, | |
aa740f46 | 1140 | .pf = NFPROTO_BRIDGE, |
1490fd89 CG |
1141 | .hooknum = NF_BR_FORWARD, |
1142 | .priority = NF_BR_PRI_BRNF - 1, | |
1143 | }, | |
1144 | { | |
1145 | .hook = br_nf_forward_arp, | |
1146 | .owner = THIS_MODULE, | |
aa740f46 | 1147 | .pf = NFPROTO_BRIDGE, |
1490fd89 CG |
1148 | .hooknum = NF_BR_FORWARD, |
1149 | .priority = NF_BR_PRI_BRNF, | |
1150 | }, | |
1490fd89 CG |
1151 | { |
1152 | .hook = br_nf_post_routing, | |
1153 | .owner = THIS_MODULE, | |
aa740f46 | 1154 | .pf = NFPROTO_BRIDGE, |
1490fd89 CG |
1155 | .hooknum = NF_BR_POST_ROUTING, |
1156 | .priority = NF_BR_PRI_LAST, | |
1157 | }, | |
1158 | { | |
1159 | .hook = ip_sabotage_in, | |
1160 | .owner = THIS_MODULE, | |
aa740f46 | 1161 | .pf = NFPROTO_IPV4, |
1490fd89 CG |
1162 | .hooknum = NF_INET_PRE_ROUTING, |
1163 | .priority = NF_IP_PRI_FIRST, | |
1164 | }, | |
1165 | { | |
1166 | .hook = ip_sabotage_in, | |
1167 | .owner = THIS_MODULE, | |
aa740f46 | 1168 | .pf = NFPROTO_IPV6, |
1490fd89 CG |
1169 | .hooknum = NF_INET_PRE_ROUTING, |
1170 | .priority = NF_IP6_PRI_FIRST, | |
1171 | }, | |
1da177e4 LT |
1172 | }; |
1173 | ||
1174 | #ifdef CONFIG_SYSCTL | |
1175 | static | |
fe2c6338 | 1176 | int brnf_sysctl_call_tables(struct ctl_table *ctl, int write, |
56b148eb | 1177 | void __user *buffer, size_t *lenp, loff_t *ppos) |
1da177e4 LT |
1178 | { |
1179 | int ret; | |
1180 | ||
8d65af78 | 1181 | ret = proc_dointvec(ctl, write, buffer, lenp, ppos); |
1da177e4 LT |
1182 | |
1183 | if (write && *(int *)(ctl->data)) | |
1184 | *(int *)(ctl->data) = 1; | |
1185 | return ret; | |
1186 | } | |
1187 | ||
fe2c6338 | 1188 | static struct ctl_table brnf_table[] = { |
1da177e4 | 1189 | { |
1da177e4 LT |
1190 | .procname = "bridge-nf-call-arptables", |
1191 | .data = &brnf_call_arptables, | |
1192 | .maxlen = sizeof(int), | |
1193 | .mode = 0644, | |
6d9f239a | 1194 | .proc_handler = brnf_sysctl_call_tables, |
1da177e4 LT |
1195 | }, |
1196 | { | |
1da177e4 LT |
1197 | .procname = "bridge-nf-call-iptables", |
1198 | .data = &brnf_call_iptables, | |
1199 | .maxlen = sizeof(int), | |
1200 | .mode = 0644, | |
6d9f239a | 1201 | .proc_handler = brnf_sysctl_call_tables, |
1da177e4 LT |
1202 | }, |
1203 | { | |
1da177e4 LT |
1204 | .procname = "bridge-nf-call-ip6tables", |
1205 | .data = &brnf_call_ip6tables, | |
1206 | .maxlen = sizeof(int), | |
1207 | .mode = 0644, | |
6d9f239a | 1208 | .proc_handler = brnf_sysctl_call_tables, |
1da177e4 LT |
1209 | }, |
1210 | { | |
1da177e4 LT |
1211 | .procname = "bridge-nf-filter-vlan-tagged", |
1212 | .data = &brnf_filter_vlan_tagged, | |
1213 | .maxlen = sizeof(int), | |
1214 | .mode = 0644, | |
6d9f239a | 1215 | .proc_handler = brnf_sysctl_call_tables, |
516299d2 MM |
1216 | }, |
1217 | { | |
516299d2 MM |
1218 | .procname = "bridge-nf-filter-pppoe-tagged", |
1219 | .data = &brnf_filter_pppoe_tagged, | |
1220 | .maxlen = sizeof(int), | |
1221 | .mode = 0644, | |
6d9f239a | 1222 | .proc_handler = brnf_sysctl_call_tables, |
1da177e4 | 1223 | }, |
4981682c PNA |
1224 | { |
1225 | .procname = "bridge-nf-pass-vlan-input-dev", | |
1226 | .data = &brnf_pass_vlan_indev, | |
1227 | .maxlen = sizeof(int), | |
1228 | .mode = 0644, | |
1229 | .proc_handler = brnf_sysctl_call_tables, | |
1230 | }, | |
f8572d8f | 1231 | { } |
1da177e4 | 1232 | }; |
1da177e4 LT |
1233 | #endif |
1234 | ||
34666d46 | 1235 | static int __init br_netfilter_init(void) |
1da177e4 | 1236 | { |
5eb87f45 | 1237 | int ret; |
1da177e4 | 1238 | |
34666d46 | 1239 | ret = nf_register_hooks(br_nf_ops, ARRAY_SIZE(br_nf_ops)); |
5eb87f45 | 1240 | if (ret < 0) |
1da177e4 | 1241 | return ret; |
fc66f95c | 1242 | |
1da177e4 | 1243 | #ifdef CONFIG_SYSCTL |
ec8f23ce | 1244 | brnf_sysctl_header = register_net_sysctl(&init_net, "net/bridge", brnf_table); |
1da177e4 | 1245 | if (brnf_sysctl_header == NULL) { |
789bc3e5 SH |
1246 | printk(KERN_WARNING |
1247 | "br_netfilter: can't register to sysctl.\n"); | |
96727239 GU |
1248 | nf_unregister_hooks(br_nf_ops, ARRAY_SIZE(br_nf_ops)); |
1249 | return -ENOMEM; | |
1da177e4 LT |
1250 | } |
1251 | #endif | |
1a4ba64d | 1252 | RCU_INIT_POINTER(nf_br_ops, &br_ops); |
1da177e4 | 1253 | printk(KERN_NOTICE "Bridge firewalling registered\n"); |
1da177e4 LT |
1254 | return 0; |
1255 | } | |
1256 | ||
34666d46 | 1257 | static void __exit br_netfilter_fini(void) |
1da177e4 | 1258 | { |
1a4ba64d | 1259 | RCU_INIT_POINTER(nf_br_ops, NULL); |
5eb87f45 | 1260 | nf_unregister_hooks(br_nf_ops, ARRAY_SIZE(br_nf_ops)); |
1da177e4 | 1261 | #ifdef CONFIG_SYSCTL |
5dd3df10 | 1262 | unregister_net_sysctl_table(brnf_sysctl_header); |
1da177e4 LT |
1263 | #endif |
1264 | } | |
34666d46 PNA |
1265 | |
1266 | module_init(br_netfilter_init); | |
1267 | module_exit(br_netfilter_fini); | |
1268 | ||
1269 | MODULE_LICENSE("GPL"); | |
1270 | MODULE_AUTHOR("Lennert Buytenhek <buytenh@gnu.org>"); | |
1271 | MODULE_AUTHOR("Bart De Schuymer <bdschuym@pandora.be>"); | |
1272 | MODULE_DESCRIPTION("Linux ethernet netfilter firewall bridge"); |