| 1 |
// SPDX-License-Identifier: GPL-2.0-or-later |
| 2 |
/* |
| 3 |
* Handle firewalling |
| 4 |
* Linux ethernet bridge |
| 5 |
* |
| 6 |
* Authors: |
| 7 |
* Lennert Buytenhek <[email protected]> |
| 8 |
* Bart De Schuymer <[email protected]> |
| 9 |
* |
| 10 |
* Lennert dedicates this file to Kerstin Wurdinger. |
| 11 |
*/ |
| 12 |
|
| 13 |
#include <linux/module.h> |
| 14 |
#include <linux/kernel.h> |
| 15 |
#include <linux/slab.h> |
| 16 |
#include <linux/ip.h> |
| 17 |
#include <linux/netdevice.h> |
| 18 |
#include <linux/skbuff.h> |
| 19 |
#include <linux/if_arp.h> |
| 20 |
#include <linux/if_ether.h> |
| 21 |
#include <linux/if_vlan.h> |
| 22 |
#include <linux/if_pppox.h> |
| 23 |
#include <linux/ppp_defs.h> |
| 24 |
#include <linux/netfilter_bridge.h> |
| 25 |
#include <uapi/linux/netfilter_bridge.h> |
| 26 |
#include <linux/netfilter_ipv4.h> |
| 27 |
#include <linux/netfilter_ipv6.h> |
| 28 |
#include <linux/netfilter_arp.h> |
| 29 |
#include <linux/in_route.h> |
| 30 |
#include <linux/rculist.h> |
| 31 |
#include <linux/inetdevice.h> |
| 32 |
|
| 33 |
#include <net/ip.h> |
| 34 |
#include <net/ipv6.h> |
| 35 |
#include <net/addrconf.h> |
| 36 |
#include <net/dst_metadata.h> |
| 37 |
#include <net/route.h> |
| 38 |
#include <net/netfilter/br_netfilter.h> |
| 39 |
#include <net/netns/generic.h> |
| 40 |
#include <net/inet_dscp.h> |
| 41 |
|
| 42 |
#include <linux/uaccess.h> |
| 43 |
#include "br_private.h" |
| 44 |
#ifdef CONFIG_SYSCTL |
| 45 |
#include <linux/sysctl.h> |
| 46 |
#endif |
| 47 |
|
| 48 |
#if IS_ENABLED(CONFIG_NF_CONNTRACK) |
| 49 |
#include <net/netfilter/nf_conntrack_core.h> |
| 50 |
#endif |
| 51 |
|
| 52 |
static unsigned int brnf_net_id __read_mostly; |
| 53 |
|
| 54 |
struct brnf_net { |
| 55 |
bool enabled; |
| 56 |
|
| 57 |
#ifdef CONFIG_SYSCTL |
| 58 |
struct ctl_table_header *ctl_hdr; |
| 59 |
#endif |
| 60 |
|
| 61 |
/* default value is 1 */ |
| 62 |
int call_iptables; |
| 63 |
int call_ip6tables; |
| 64 |
int call_arptables; |
| 65 |
|
| 66 |
/* default value is 0 */ |
| 67 |
int filter_vlan_tagged; |
| 68 |
int filter_pppoe_tagged; |
| 69 |
int pass_vlan_indev; |
| 70 |
}; |
| 71 |
|
| 72 |
#define IS_IP(skb) \ |
| 73 |
(!skb_vlan_tag_present(skb) && skb->protocol == htons(ETH_P_IP)) |
| 74 |
|
| 75 |
#define IS_IPV6(skb) \ |
| 76 |
(!skb_vlan_tag_present(skb) && skb->protocol == htons(ETH_P_IPV6)) |
| 77 |
|
| 78 |
#define IS_ARP(skb) \ |
| 79 |
(!skb_vlan_tag_present(skb) && skb->protocol == htons(ETH_P_ARP)) |
| 80 |
|
| 81 |
static inline __be16 vlan_proto(const struct sk_buff *skb) |
| 82 |
{ |
| 83 |
if (skb_vlan_tag_present(skb)) |
| 84 |
return skb->protocol; |
| 85 |
else if (skb->protocol == htons(ETH_P_8021Q)) |
| 86 |
return vlan_eth_hdr(skb)->h_vlan_encapsulated_proto; |
| 87 |
else |
| 88 |
return 0; |
| 89 |
} |
| 90 |
|
| 91 |
static inline bool is_vlan_ip(const struct sk_buff *skb, const struct net *net) |
| 92 |
{ |
| 93 |
struct brnf_net *brnet = net_generic(net, brnf_net_id); |
| 94 |
|
| 95 |
return vlan_proto(skb) == htons(ETH_P_IP) && brnet->filter_vlan_tagged; |
| 96 |
} |
| 97 |
|
| 98 |
static inline bool is_vlan_ipv6(const struct sk_buff *skb, |
| 99 |
const struct net *net) |
| 100 |
{ |
| 101 |
struct brnf_net *brnet = net_generic(net, brnf_net_id); |
| 102 |
|
| 103 |
return vlan_proto(skb) == htons(ETH_P_IPV6) && |
| 104 |
brnet->filter_vlan_tagged; |
| 105 |
} |
| 106 |
|
| 107 |
static inline bool is_vlan_arp(const struct sk_buff *skb, const struct net *net) |
| 108 |
{ |
| 109 |
struct brnf_net *brnet = net_generic(net, brnf_net_id); |
| 110 |
|
| 111 |
return vlan_proto(skb) == htons(ETH_P_ARP) && brnet->filter_vlan_tagged; |
| 112 |
} |
| 113 |
|
| 114 |
static inline __be16 pppoe_proto(const struct sk_buff *skb) |
| 115 |
{ |
| 116 |
return *((__be16 *)(skb_mac_header(skb) + ETH_HLEN + |
| 117 |
sizeof(struct pppoe_hdr))); |
| 118 |
} |
| 119 |
|
| 120 |
static inline bool is_pppoe_ip(const struct sk_buff *skb, const struct net *net) |
| 121 |
{ |
| 122 |
struct brnf_net *brnet = net_generic(net, brnf_net_id); |
| 123 |
|
| 124 |
return skb->protocol == htons(ETH_P_PPP_SES) && |
| 125 |
pppoe_proto(skb) == htons(PPP_IP) && brnet->filter_pppoe_tagged; |
| 126 |
} |
| 127 |
|
| 128 |
static inline bool is_pppoe_ipv6(const struct sk_buff *skb, |
| 129 |
const struct net *net) |
| 130 |
{ |
| 131 |
struct brnf_net *brnet = net_generic(net, brnf_net_id); |
| 132 |
|
| 133 |
return skb->protocol == htons(ETH_P_PPP_SES) && |
| 134 |
pppoe_proto(skb) == htons(PPP_IPV6) && |
| 135 |
brnet->filter_pppoe_tagged; |
| 136 |
} |
| 137 |
|
| 138 |
/* largest possible L2 header, see br_nf_dev_queue_xmit() */ |
| 139 |
#define NF_BRIDGE_MAX_MAC_HEADER_LENGTH (PPPOE_SES_HLEN + ETH_HLEN) |
| 140 |
|
| 141 |
struct brnf_frag_data { |
| 142 |
local_lock_t bh_lock; |
| 143 |
char mac[NF_BRIDGE_MAX_MAC_HEADER_LENGTH]; |
| 144 |
u8 encap_size; |
| 145 |
u8 size; |
| 146 |
u16 vlan_tci; |
| 147 |
__be16 vlan_proto; |
| 148 |
}; |
| 149 |
|
| 150 |
static DEFINE_PER_CPU(struct brnf_frag_data, brnf_frag_data_storage) = { |
| 151 |
.bh_lock = INIT_LOCAL_LOCK(bh_lock), |
| 152 |
}; |
| 153 |
|
| 154 |
static void nf_bridge_info_free(struct sk_buff *skb) |
| 155 |
{ |
| 156 |
skb_ext_del(skb, SKB_EXT_BRIDGE_NF); |
| 157 |
} |
| 158 |
|
| 159 |
static inline struct net_device *bridge_parent(const struct net_device *dev) |
| 160 |
{ |
| 161 |
struct net_bridge_port *port; |
| 162 |
|
| 163 |
port = br_port_get_rcu(dev); |
| 164 |
return port ? port->br->dev : NULL; |
| 165 |
} |
| 166 |
|
| 167 |
static inline struct nf_bridge_info *nf_bridge_unshare(struct sk_buff *skb) |
| 168 |
{ |
| 169 |
return skb_ext_add(skb, SKB_EXT_BRIDGE_NF); |
| 170 |
} |
| 171 |
|
| 172 |
unsigned int nf_bridge_encap_header_len(const struct sk_buff *skb) |
| 173 |
{ |
| 174 |
switch (skb->protocol) { |
| 175 |
case __cpu_to_be16(ETH_P_8021Q): |
| 176 |
return VLAN_HLEN; |
| 177 |
case __cpu_to_be16(ETH_P_PPP_SES): |
| 178 |
return PPPOE_SES_HLEN; |
| 179 |
default: |
| 180 |
return 0; |
| 181 |
} |
| 182 |
} |
| 183 |
|
| 184 |
static inline void nf_bridge_pull_encap_header(struct sk_buff *skb) |
| 185 |
{ |
| 186 |
unsigned int len = nf_bridge_encap_header_len(skb); |
| 187 |
|
| 188 |
skb_pull(skb, len); |
| 189 |
skb->network_header += len; |
| 190 |
} |
| 191 |
|
| 192 |
static inline void nf_bridge_pull_encap_header_rcsum(struct sk_buff *skb) |
| 193 |
{ |
| 194 |
unsigned int len = nf_bridge_encap_header_len(skb); |
| 195 |
|
| 196 |
skb_pull_rcsum(skb, len); |
| 197 |
skb->network_header += len; |
| 198 |
} |
| 199 |
|
| 200 |
/* When handing a packet over to the IP layer |
| 201 |
* check whether we have a skb that is in the |
| 202 |
* expected format |
| 203 |
*/ |
| 204 |
|
| 205 |
static int br_validate_ipv4(struct net *net, struct sk_buff *skb) |
| 206 |
{ |
| 207 |
const struct iphdr *iph; |
| 208 |
u32 len; |
| 209 |
|
| 210 |
if (!pskb_may_pull(skb, sizeof(struct iphdr))) |
| 211 |
goto inhdr_error; |
| 212 |
|
| 213 |
iph = ip_hdr(skb); |
| 214 |
|
| 215 |
/* Basic sanity checks */ |
| 216 |
if (iph->ihl < 5 || iph->version != 4) |
| 217 |
goto inhdr_error; |
| 218 |
|
| 219 |
if (!pskb_may_pull(skb, iph->ihl*4)) |
| 220 |
goto inhdr_error; |
| 221 |
|
| 222 |
iph = ip_hdr(skb); |
| 223 |
if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl))) |
| 224 |
goto csum_error; |
| 225 |
|
| 226 |
len = skb_ip_totlen(skb); |
| 227 |
if (skb->len < len) { |
| 228 |
__IP_INC_STATS(net, IPSTATS_MIB_INTRUNCATEDPKTS); |
| 229 |
goto drop; |
| 230 |
} else if (len < (iph->ihl*4)) |
| 231 |
goto inhdr_error; |
| 232 |
|
| 233 |
if (pskb_trim_rcsum(skb, len)) { |
| 234 |
__IP_INC_STATS(net, IPSTATS_MIB_INDISCARDS); |
| 235 |
goto drop; |
| 236 |
} |
| 237 |
|
| 238 |
memset(IPCB(skb), 0, sizeof(struct inet_skb_parm)); |
| 239 |
/* We should really parse IP options here but until |
| 240 |
* somebody who actually uses IP options complains to |
| 241 |
* us we'll just silently ignore the options because |
| 242 |
* we're lazy! |
| 243 |
*/ |
| 244 |
return 0; |
| 245 |
|
| 246 |
csum_error: |
| 247 |
__IP_INC_STATS(net, IPSTATS_MIB_CSUMERRORS); |
| 248 |
inhdr_error: |
| 249 |
__IP_INC_STATS(net, IPSTATS_MIB_INHDRERRORS); |
| 250 |
drop: |
| 251 |
return -1; |
| 252 |
} |
| 253 |
|
| 254 |
void nf_bridge_update_protocol(struct sk_buff *skb) |
| 255 |
{ |
| 256 |
const struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb); |
| 257 |
|
| 258 |
switch (nf_bridge->orig_proto) { |
| 259 |
case BRNF_PROTO_8021Q: |
| 260 |
skb->protocol = htons(ETH_P_8021Q); |
| 261 |
break; |
| 262 |
case BRNF_PROTO_PPPOE: |
| 263 |
skb->protocol = htons(ETH_P_PPP_SES); |
| 264 |
break; |
| 265 |
case BRNF_PROTO_UNCHANGED: |
| 266 |
break; |
| 267 |
} |
| 268 |
} |
| 269 |
|
| 270 |
/* Obtain the correct destination MAC address, while preserving the original |
| 271 |
* source MAC address. If we already know this address, we just copy it. If we |
| 272 |
* don't, we use the neighbour framework to find out. In both cases, we make |
| 273 |
* sure that br_handle_frame_finish() is called afterwards. |
| 274 |
*/ |
| 275 |
int br_nf_pre_routing_finish_bridge(struct net *net, struct sock *sk, struct sk_buff *skb) |
| 276 |
{ |
| 277 |
struct neighbour *neigh; |
| 278 |
struct dst_entry *dst; |
| 279 |
|
| 280 |
skb->dev = bridge_parent(skb->dev); |
| 281 |
if (!skb->dev) |
| 282 |
goto free_skb; |
| 283 |
dst = skb_dst(skb); |
| 284 |
neigh = dst_neigh_lookup_skb(dst, skb); |
| 285 |
if (neigh) { |
| 286 |
struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb); |
| 287 |
int ret; |
| 288 |
|
| 289 |
if ((READ_ONCE(neigh->nud_state) & NUD_CONNECTED) && |
| 290 |
READ_ONCE(neigh->hh.hh_len)) { |
| 291 |
struct net_device *br_indev; |
| 292 |
|
| 293 |
br_indev = nf_bridge_get_physindev(skb, net); |
| 294 |
if (!br_indev) { |
| 295 |
neigh_release(neigh); |
| 296 |
goto free_skb; |
| 297 |
} |
| 298 |
|
| 299 |
neigh_hh_bridge(&neigh->hh, skb); |
| 300 |
skb->dev = br_indev; |
| 301 |
|
| 302 |
ret = br_handle_frame_finish(net, sk, skb); |
| 303 |
} else { |
| 304 |
/* the neighbour function below overwrites the complete |
| 305 |
* MAC header, so we save the Ethernet source address and |
| 306 |
* protocol number. |
| 307 |
*/ |
| 308 |
skb_copy_from_linear_data_offset(skb, |
| 309 |
-(ETH_HLEN-ETH_ALEN), |
| 310 |
nf_bridge->neigh_header, |
| 311 |
ETH_HLEN-ETH_ALEN); |
| 312 |
/* tell br_dev_xmit to continue with forwarding */ |
| 313 |
nf_bridge->bridged_dnat = 1; |
| 314 |
/* FIXME Need to refragment */ |
| 315 |
ret = READ_ONCE(neigh->output)(neigh, skb); |
| 316 |
} |
| 317 |
neigh_release(neigh); |
| 318 |
return ret; |
| 319 |
} |
| 320 |
free_skb: |
| 321 |
kfree_skb(skb); |
| 322 |
return 0; |
| 323 |
} |
| 324 |
|
| 325 |
static inline bool |
| 326 |
br_nf_ipv4_daddr_was_changed(const struct sk_buff *skb, |
| 327 |
const struct nf_bridge_info *nf_bridge) |
| 328 |
{ |
| 329 |
return ip_hdr(skb)->daddr != nf_bridge->ipv4_daddr; |
| 330 |
} |
| 331 |
|
| 332 |
/* This requires some explaining. If DNAT has taken place, |
| 333 |
* we will need to fix up the destination Ethernet address. |
| 334 |
* This is also true when SNAT takes place (for the reply direction). |
| 335 |
* |
| 336 |
* There are two cases to consider: |
| 337 |
* 1. The packet was DNAT'ed to a device in the same bridge |
| 338 |
* port group as it was received on. We can still bridge |
| 339 |
* the packet. |
| 340 |
* 2. The packet was DNAT'ed to a different device, either |
| 341 |
* a non-bridged device or another bridge port group. |
| 342 |
* The packet will need to be routed. |
| 343 |
* |
| 344 |
* The correct way of distinguishing between these two cases is to |
| 345 |
* call ip_route_input() and to look at skb->dst->dev, which is |
| 346 |
* changed to the destination device if ip_route_input() succeeds. |
| 347 |
* |
| 348 |
* Let's first consider the case that ip_route_input() succeeds: |
| 349 |
* |
| 350 |
* If the output device equals the logical bridge device the packet |
| 351 |
* came in on, we can consider this bridging. The corresponding MAC |
| 352 |
* address will be obtained in br_nf_pre_routing_finish_bridge. |
| 353 |
* Otherwise, the packet is considered to be routed and we just |
| 354 |
* change the destination MAC address so that the packet will |
| 355 |
* later be passed up to the IP stack to be routed. For a redirected |
| 356 |
* packet, ip_route_input() will give back the localhost as output device, |
| 357 |
* which differs from the bridge device. |
| 358 |
* |
| 359 |
* Let's now consider the case that ip_route_input() fails: |
| 360 |
* |
| 361 |
* This can be because the destination address is martian, in which case |
| 362 |
* the packet will be dropped. |
| 363 |
* If IP forwarding is disabled, ip_route_input() will fail, while |
| 364 |
* ip_route_output_key() can return success. The source |
| 365 |
* address for ip_route_output_key() is set to zero, so ip_route_output_key() |
| 366 |
* thinks we're handling a locally generated packet and won't care |
| 367 |
* if IP forwarding is enabled. If the output device equals the logical bridge |
| 368 |
* device, we proceed as if ip_route_input() succeeded. If it differs from the |
| 369 |
* logical bridge port or if ip_route_output_key() fails we drop the packet. |
| 370 |
*/ |
| 371 |
static int br_nf_pre_routing_finish(struct net *net, struct sock *sk, struct sk_buff *skb) |
| 372 |
{ |
| 373 |
struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb); |
| 374 |
struct net_device *dev = skb->dev, *br_indev; |
| 375 |
const struct iphdr *iph = ip_hdr(skb); |
| 376 |
enum skb_drop_reason reason; |
| 377 |
struct rtable *rt; |
| 378 |
|
| 379 |
br_indev = nf_bridge_get_physindev(skb, net); |
| 380 |
if (!br_indev) { |
| 381 |
kfree_skb(skb); |
| 382 |
return 0; |
| 383 |
} |
| 384 |
|
| 385 |
nf_bridge->frag_max_size = IPCB(skb)->frag_max_size; |
| 386 |
|
| 387 |
if (nf_bridge->pkt_otherhost) { |
| 388 |
skb->pkt_type = PACKET_OTHERHOST; |
| 389 |
nf_bridge->pkt_otherhost = false; |
| 390 |
} |
| 391 |
nf_bridge->in_prerouting = 0; |
| 392 |
if (br_nf_ipv4_daddr_was_changed(skb, nf_bridge)) { |
| 393 |
reason = ip_route_input(skb, iph->daddr, iph->saddr, |
| 394 |
ip4h_dscp(iph), dev); |
| 395 |
if (reason) { |
| 396 |
kfree_skb_reason(skb, reason); |
| 397 |
return 0; |
| 398 |
} else { |
| 399 |
if (skb_dst(skb)->dev == dev) { |
| 400 |
skb->dev = br_indev; |
| 401 |
nf_bridge_update_protocol(skb); |
| 402 |
nf_bridge_push_encap_header(skb); |
| 403 |
br_nf_hook_thresh(NF_BR_PRE_ROUTING, |
| 404 |
net, sk, skb, skb->dev, |
| 405 |
NULL, |
| 406 |
br_nf_pre_routing_finish_bridge); |
| 407 |
return 0; |
| 408 |
} |
| 409 |
ether_addr_copy(eth_hdr(skb)->h_dest, dev->dev_addr); |
| 410 |
skb->pkt_type = PACKET_HOST; |
| 411 |
} |
| 412 |
} else { |
| 413 |
rt = bridge_parent_rtable(br_indev); |
| 414 |
if (!rt) { |
| 415 |
kfree_skb(skb); |
| 416 |
return 0; |
| 417 |
} |
| 418 |
skb_dst_drop(skb); |
| 419 |
skb_dst_set_noref(skb, &rt->dst); |
| 420 |
} |
| 421 |
|
| 422 |
skb->dev = br_indev; |
| 423 |
nf_bridge_update_protocol(skb); |
| 424 |
nf_bridge_push_encap_header(skb); |
| 425 |
br_nf_hook_thresh(NF_BR_PRE_ROUTING, net, sk, skb, skb->dev, NULL, |
| 426 |
br_handle_frame_finish); |
| 427 |
return 0; |
| 428 |
} |
| 429 |
|
| 430 |
static struct net_device *brnf_get_logical_dev(struct sk_buff *skb, |
| 431 |
const struct net_device *dev, |
| 432 |
const struct net *net) |
| 433 |
{ |
| 434 |
struct net_device *vlan, *br; |
| 435 |
struct brnf_net *brnet = net_generic(net, brnf_net_id); |
| 436 |
|
| 437 |
br = bridge_parent(dev); |
| 438 |
|
| 439 |
if (brnet->pass_vlan_indev == 0 || !skb_vlan_tag_present(skb)) |
| 440 |
return br; |
| 441 |
|
| 442 |
vlan = __vlan_find_dev_deep_rcu(br, skb->vlan_proto, |
| 443 |
skb_vlan_tag_get(skb) & VLAN_VID_MASK); |
| 444 |
|
| 445 |
return vlan ? vlan : br; |
| 446 |
} |
| 447 |
|
| 448 |
/* Some common code for IPv4/IPv6 */ |
| 449 |
struct net_device *setup_pre_routing(struct sk_buff *skb, const struct net *net) |
| 450 |
{ |
| 451 |
struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb); |
| 452 |
|
| 453 |
if (skb->pkt_type == PACKET_OTHERHOST) { |
| 454 |
skb->pkt_type = PACKET_HOST; |
| 455 |
nf_bridge->pkt_otherhost = true; |
| 456 |
} |
| 457 |
|
| 458 |
nf_bridge->in_prerouting = 1; |
| 459 |
nf_bridge->physinif = skb->dev->ifindex; |
| 460 |
skb->dev = brnf_get_logical_dev(skb, skb->dev, net); |
| 461 |
|
| 462 |
if (skb->protocol == htons(ETH_P_8021Q)) |
| 463 |
nf_bridge->orig_proto = BRNF_PROTO_8021Q; |
| 464 |
else if (skb->protocol == htons(ETH_P_PPP_SES)) |
| 465 |
nf_bridge->orig_proto = BRNF_PROTO_PPPOE; |
| 466 |
|
| 467 |
/* Must drop socket now because of tproxy. */ |
| 468 |
skb_orphan(skb); |
| 469 |
return skb->dev; |
| 470 |
} |
| 471 |
|
| 472 |
/* Direct IPv6 traffic to br_nf_pre_routing_ipv6. |
| 473 |
* Replicate the checks that IPv4 does on packet reception. |
| 474 |
* Set skb->dev to the bridge device (i.e. parent of the |
| 475 |
* receiving device) to make netfilter happy, the REDIRECT |
| 476 |
* target in particular. Save the original destination IP |
| 477 |
* address to be able to detect DNAT afterwards. */ |
| 478 |
static unsigned int br_nf_pre_routing(void *priv, |
| 479 |
struct sk_buff *skb, |
| 480 |
const struct nf_hook_state *state) |
| 481 |
{ |
| 482 |
struct nf_bridge_info *nf_bridge; |
| 483 |
struct net_bridge_port *p; |
| 484 |
struct net_bridge *br; |
| 485 |
__u32 len = nf_bridge_encap_header_len(skb); |
| 486 |
struct brnf_net *brnet; |
| 487 |
|
| 488 |
if (unlikely(!pskb_may_pull(skb, len))) |
| 489 |
return NF_DROP_REASON(skb, SKB_DROP_REASON_PKT_TOO_SMALL, 0); |
| 490 |
|
| 491 |
p = br_port_get_rcu(state->in); |
| 492 |
if (p == NULL) |
| 493 |
return NF_DROP_REASON(skb, SKB_DROP_REASON_DEV_READY, 0); |
| 494 |
br = p->br; |
| 495 |
|
| 496 |
brnet = net_generic(state->net, brnf_net_id); |
| 497 |
if (IS_IPV6(skb) || is_vlan_ipv6(skb, state->net) || |
| 498 |
is_pppoe_ipv6(skb, state->net)) { |
| 499 |
if (!brnet->call_ip6tables && |
| 500 |
!br_opt_get(br, BROPT_NF_CALL_IP6TABLES)) |
| 501 |
return NF_ACCEPT; |
| 502 |
if (!ipv6_mod_enabled()) { |
| 503 |
pr_warn_once("Module ipv6 is disabled, so call_ip6tables is not supported."); |
| 504 |
return NF_DROP_REASON(skb, SKB_DROP_REASON_IPV6DISABLED, 0); |
| 505 |
} |
| 506 |
|
| 507 |
nf_bridge_pull_encap_header_rcsum(skb); |
| 508 |
return br_nf_pre_routing_ipv6(priv, skb, state); |
| 509 |
} |
| 510 |
|
| 511 |
if (!brnet->call_iptables && !br_opt_get(br, BROPT_NF_CALL_IPTABLES)) |
| 512 |
return NF_ACCEPT; |
| 513 |
|
| 514 |
if (!IS_IP(skb) && !is_vlan_ip(skb, state->net) && |
| 515 |
!is_pppoe_ip(skb, state->net)) |
| 516 |
return NF_ACCEPT; |
| 517 |
|
| 518 |
nf_bridge_pull_encap_header_rcsum(skb); |
| 519 |
|
| 520 |
if (br_validate_ipv4(state->net, skb)) |
| 521 |
return NF_DROP_REASON(skb, SKB_DROP_REASON_IP_INHDR, 0); |
| 522 |
|
| 523 |
if (!nf_bridge_alloc(skb)) |
| 524 |
return NF_DROP_REASON(skb, SKB_DROP_REASON_NOMEM, 0); |
| 525 |
if (!setup_pre_routing(skb, state->net)) |
| 526 |
return NF_DROP_REASON(skb, SKB_DROP_REASON_DEV_READY, 0); |
| 527 |
|
| 528 |
nf_bridge = nf_bridge_info_get(skb); |
| 529 |
nf_bridge->ipv4_daddr = ip_hdr(skb)->daddr; |
| 530 |
|
| 531 |
skb->protocol = htons(ETH_P_IP); |
| 532 |
skb->transport_header = skb->network_header + ip_hdr(skb)->ihl * 4; |
| 533 |
|
| 534 |
NF_HOOK(NFPROTO_IPV4, NF_INET_PRE_ROUTING, state->net, state->sk, skb, |
| 535 |
skb->dev, NULL, |
| 536 |
br_nf_pre_routing_finish); |
| 537 |
|
| 538 |
return NF_STOLEN; |
| 539 |
} |
| 540 |
|
| 541 |
#if IS_ENABLED(CONFIG_NF_CONNTRACK) |
| 542 |
/* conntracks' nf_confirm logic cannot handle cloned skbs referencing |
| 543 |
* the same nf_conn entry, which will happen for multicast (broadcast) |
| 544 |
* Frames on bridges. |
| 545 |
* |
| 546 |
* Example: |
| 547 |
* macvlan0 |
| 548 |
* br0 |
| 549 |
* ethX ethY |
| 550 |
* |
| 551 |
* ethX (or Y) receives multicast or broadcast packet containing |
| 552 |
* an IP packet, not yet in conntrack table. |
| 553 |
* |
| 554 |
* 1. skb passes through bridge and fake-ip (br_netfilter)Prerouting. |
| 555 |
* -> skb->_nfct now references a unconfirmed entry |
| 556 |
* 2. skb is broad/mcast packet. bridge now passes clones out on each bridge |
| 557 |
* interface. |
| 558 |
* 3. skb gets passed up the stack. |
| 559 |
* 4. In macvlan case, macvlan driver retains clone(s) of the mcast skb |
| 560 |
* and schedules a work queue to send them out on the lower devices. |
| 561 |
* |
| 562 |
* The clone skb->_nfct is not a copy, it is the same entry as the |
| 563 |
* original skb. The macvlan rx handler then returns RX_HANDLER_PASS. |
| 564 |
* 5. Normal conntrack hooks (in NF_INET_LOCAL_IN) confirm the orig skb. |
| 565 |
* |
| 566 |
* The Macvlan broadcast worker and normal confirm path will race. |
| 567 |
* |
| 568 |
* This race will not happen if step 2 already confirmed a clone. In that |
| 569 |
* case later steps perform skb_clone() with skb->_nfct already confirmed (in |
| 570 |
* hash table). This works fine. |
| 571 |
* |
| 572 |
* But such confirmation won't happen when eb/ip/nftables rules dropped the |
| 573 |
* packets before they reached the nf_confirm step in postrouting. |
| 574 |
* |
| 575 |
* Work around this problem by explicit confirmation of the entry at |
| 576 |
* LOCAL_IN time, before upper layer has a chance to clone the unconfirmed |
| 577 |
* entry. |
| 578 |
* |
| 579 |
*/ |
| 580 |
static unsigned int br_nf_local_in(void *priv, |
| 581 |
struct sk_buff *skb, |
| 582 |
const struct nf_hook_state *state) |
| 583 |
{ |
| 584 |
bool promisc = BR_INPUT_SKB_CB(skb)->promisc; |
| 585 |
struct nf_conntrack *nfct = skb_nfct(skb); |
| 586 |
const struct nf_ct_hook *ct_hook; |
| 587 |
struct nf_conn *ct; |
| 588 |
int ret; |
| 589 |
|
| 590 |
if (promisc) { |
| 591 |
nf_reset_ct(skb); |
| 592 |
return NF_ACCEPT; |
| 593 |
} |
| 594 |
|
| 595 |
if (!nfct || skb->pkt_type == PACKET_HOST) |
| 596 |
return NF_ACCEPT; |
| 597 |
|
| 598 |
ct = container_of(nfct, struct nf_conn, ct_general); |
| 599 |
if (likely(nf_ct_is_confirmed(ct))) |
| 600 |
return NF_ACCEPT; |
| 601 |
|
| 602 |
if (WARN_ON_ONCE(refcount_read(&nfct->use) != 1)) { |
| 603 |
nf_reset_ct(skb); |
| 604 |
return NF_ACCEPT; |
| 605 |
} |
| 606 |
|
| 607 |
WARN_ON_ONCE(skb_shared(skb)); |
| 608 |
|
| 609 |
/* We can't call nf_confirm here, it would create a dependency |
| 610 |
* on nf_conntrack module. |
| 611 |
*/ |
| 612 |
ct_hook = rcu_dereference(nf_ct_hook); |
| 613 |
if (!ct_hook) { |
| 614 |
skb->_nfct = 0ul; |
| 615 |
nf_conntrack_put(nfct); |
| 616 |
return NF_ACCEPT; |
| 617 |
} |
| 618 |
|
| 619 |
nf_bridge_pull_encap_header(skb); |
| 620 |
ret = ct_hook->confirm(skb); |
| 621 |
switch (ret & NF_VERDICT_MASK) { |
| 622 |
case NF_STOLEN: |
| 623 |
return NF_STOLEN; |
| 624 |
default: |
| 625 |
nf_bridge_push_encap_header(skb); |
| 626 |
break; |
| 627 |
} |
| 628 |
|
| 629 |
return ret; |
| 630 |
} |
| 631 |
#endif |
| 632 |
|
| 633 |
/* PF_BRIDGE/FORWARD *************************************************/ |
| 634 |
static int br_nf_forward_finish(struct net *net, struct sock *sk, struct sk_buff *skb) |
| 635 |
{ |
| 636 |
struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb); |
| 637 |
struct net_device *in; |
| 638 |
|
| 639 |
if (!IS_ARP(skb) && !is_vlan_arp(skb, net)) { |
| 640 |
|
| 641 |
if (skb->protocol == htons(ETH_P_IP)) |
| 642 |
nf_bridge->frag_max_size = IPCB(skb)->frag_max_size; |
| 643 |
|
| 644 |
if (skb->protocol == htons(ETH_P_IPV6)) |
| 645 |
nf_bridge->frag_max_size = IP6CB(skb)->frag_max_size; |
| 646 |
|
| 647 |
in = nf_bridge_get_physindev(skb, net); |
| 648 |
if (!in) { |
| 649 |
kfree_skb(skb); |
| 650 |
return 0; |
| 651 |
} |
| 652 |
if (nf_bridge->pkt_otherhost) { |
| 653 |
skb->pkt_type = PACKET_OTHERHOST; |
| 654 |
nf_bridge->pkt_otherhost = false; |
| 655 |
} |
| 656 |
nf_bridge_update_protocol(skb); |
| 657 |
} else { |
| 658 |
in = *((struct net_device **)(skb->cb)); |
| 659 |
} |
| 660 |
nf_bridge_push_encap_header(skb); |
| 661 |
|
| 662 |
br_nf_hook_thresh(NF_BR_FORWARD, net, sk, skb, in, skb->dev, |
| 663 |
br_forward_finish); |
| 664 |
return 0; |
| 665 |
} |
| 666 |
|
| 667 |
|
| 668 |
static unsigned int br_nf_forward_ip(struct sk_buff *skb, |
| 669 |
const struct nf_hook_state *state, |
| 670 |
u8 pf) |
| 671 |
{ |
| 672 |
struct nf_bridge_info *nf_bridge; |
| 673 |
struct net_device *parent; |
| 674 |
|
| 675 |
nf_bridge = nf_bridge_info_get(skb); |
| 676 |
if (!nf_bridge) |
| 677 |
return NF_ACCEPT; |
| 678 |
|
| 679 |
/* Need exclusive nf_bridge_info since we might have multiple |
| 680 |
* different physoutdevs. */ |
| 681 |
if (!nf_bridge_unshare(skb)) |
| 682 |
return NF_DROP_REASON(skb, SKB_DROP_REASON_NOMEM, 0); |
| 683 |
|
| 684 |
nf_bridge = nf_bridge_info_get(skb); |
| 685 |
if (!nf_bridge) |
| 686 |
return NF_DROP_REASON(skb, SKB_DROP_REASON_NOMEM, 0); |
| 687 |
|
| 688 |
parent = bridge_parent(state->out); |
| 689 |
if (!parent) |
| 690 |
return NF_DROP_REASON(skb, SKB_DROP_REASON_DEV_READY, 0); |
| 691 |
|
| 692 |
nf_bridge_pull_encap_header(skb); |
| 693 |
|
| 694 |
if (skb->pkt_type == PACKET_OTHERHOST) { |
| 695 |
skb->pkt_type = PACKET_HOST; |
| 696 |
nf_bridge->pkt_otherhost = true; |
| 697 |
} |
| 698 |
|
| 699 |
if (pf == NFPROTO_IPV4) { |
| 700 |
if (br_validate_ipv4(state->net, skb)) |
| 701 |
return NF_DROP_REASON(skb, SKB_DROP_REASON_IP_INHDR, 0); |
| 702 |
IPCB(skb)->frag_max_size = nf_bridge->frag_max_size; |
| 703 |
skb->protocol = htons(ETH_P_IP); |
| 704 |
} else if (pf == NFPROTO_IPV6) { |
| 705 |
if (br_validate_ipv6(state->net, skb)) |
| 706 |
return NF_DROP_REASON(skb, SKB_DROP_REASON_IP_INHDR, 0); |
| 707 |
IP6CB(skb)->frag_max_size = nf_bridge->frag_max_size; |
| 708 |
skb->protocol = htons(ETH_P_IPV6); |
| 709 |
} else { |
| 710 |
WARN_ON_ONCE(1); |
| 711 |
return NF_DROP; |
| 712 |
} |
| 713 |
|
| 714 |
nf_bridge->physoutdev = skb->dev; |
| 715 |
|
| 716 |
NF_HOOK(pf, NF_INET_FORWARD, state->net, NULL, skb, |
| 717 |
brnf_get_logical_dev(skb, state->in, state->net), |
| 718 |
parent, br_nf_forward_finish); |
| 719 |
|
| 720 |
return NF_STOLEN; |
| 721 |
} |
| 722 |
|
| 723 |
static unsigned int br_nf_forward_arp(struct sk_buff *skb, |
| 724 |
const struct nf_hook_state *state) |
| 725 |
{ |
| 726 |
struct net_bridge_port *p; |
| 727 |
struct net_bridge *br; |
| 728 |
struct net_device **d = (struct net_device **)(skb->cb); |
| 729 |
struct brnf_net *brnet; |
| 730 |
|
| 731 |
p = br_port_get_rcu(state->out); |
| 732 |
if (p == NULL) |
| 733 |
return NF_ACCEPT; |
| 734 |
br = p->br; |
| 735 |
|
| 736 |
brnet = net_generic(state->net, brnf_net_id); |
| 737 |
if (!brnet->call_arptables && !br_opt_get(br, BROPT_NF_CALL_ARPTABLES)) |
| 738 |
return NF_ACCEPT; |
| 739 |
|
| 740 |
if (is_vlan_arp(skb, state->net)) |
| 741 |
nf_bridge_pull_encap_header(skb); |
| 742 |
|
| 743 |
if (unlikely(!pskb_may_pull(skb, sizeof(struct arphdr)))) |
| 744 |
return NF_DROP_REASON(skb, SKB_DROP_REASON_PKT_TOO_SMALL, 0); |
| 745 |
|
| 746 |
if (arp_hdr(skb)->ar_pln != 4) { |
| 747 |
if (is_vlan_arp(skb, state->net)) |
| 748 |
nf_bridge_push_encap_header(skb); |
| 749 |
return NF_ACCEPT; |
| 750 |
} |
| 751 |
*d = state->in; |
| 752 |
NF_HOOK(NFPROTO_ARP, NF_ARP_FORWARD, state->net, state->sk, skb, |
| 753 |
state->in, state->out, br_nf_forward_finish); |
| 754 |
|
| 755 |
return NF_STOLEN; |
| 756 |
} |
| 757 |
|
| 758 |
/* This is the 'purely bridged' case. For IP, we pass the packet to |
| 759 |
* netfilter with indev and outdev set to the bridge device, |
| 760 |
* but we are still able to filter on the 'real' indev/outdev |
| 761 |
* because of the physdev module. For ARP, indev and outdev are the |
| 762 |
* bridge ports. |
| 763 |
*/ |
| 764 |
static unsigned int br_nf_forward(void *priv, |
| 765 |
struct sk_buff *skb, |
| 766 |
const struct nf_hook_state *state) |
| 767 |
{ |
| 768 |
if (IS_IP(skb) || is_vlan_ip(skb, state->net) || |
| 769 |
is_pppoe_ip(skb, state->net)) |
| 770 |
return br_nf_forward_ip(skb, state, NFPROTO_IPV4); |
| 771 |
if (IS_IPV6(skb) || is_vlan_ipv6(skb, state->net) || |
| 772 |
is_pppoe_ipv6(skb, state->net)) |
| 773 |
return br_nf_forward_ip(skb, state, NFPROTO_IPV6); |
| 774 |
if (IS_ARP(skb) || is_vlan_arp(skb, state->net)) |
| 775 |
return br_nf_forward_arp(skb, state); |
| 776 |
|
| 777 |
return NF_ACCEPT; |
| 778 |
} |
| 779 |
|
| 780 |
static int br_nf_push_frag_xmit(struct net *net, struct sock *sk, struct sk_buff *skb) |
| 781 |
{ |
| 782 |
struct brnf_frag_data *data; |
| 783 |
int err; |
| 784 |
|
| 785 |
data = this_cpu_ptr(&brnf_frag_data_storage); |
| 786 |
err = skb_cow_head(skb, data->size); |
| 787 |
|
| 788 |
if (err) { |
| 789 |
kfree_skb(skb); |
| 790 |
return 0; |
| 791 |
} |
| 792 |
|
| 793 |
if (data->vlan_proto) |
| 794 |
__vlan_hwaccel_put_tag(skb, data->vlan_proto, data->vlan_tci); |
| 795 |
|
| 796 |
skb_copy_to_linear_data_offset(skb, -data->size, data->mac, data->size); |
| 797 |
__skb_push(skb, data->encap_size); |
| 798 |
|
| 799 |
nf_bridge_info_free(skb); |
| 800 |
return br_dev_queue_push_xmit(net, sk, skb); |
| 801 |
} |
| 802 |
|
| 803 |
static int |
| 804 |
br_nf_ip_fragment(struct net *net, struct sock *sk, struct sk_buff *skb, |
| 805 |
int (*output)(struct net *, struct sock *, struct sk_buff *)) |
| 806 |
{ |
| 807 |
unsigned int mtu = ip_skb_dst_mtu(sk, skb); |
| 808 |
struct iphdr *iph = ip_hdr(skb); |
| 809 |
|
| 810 |
if (unlikely(((iph->frag_off & htons(IP_DF)) && !skb->ignore_df) || |
| 811 |
(IPCB(skb)->frag_max_size && |
| 812 |
IPCB(skb)->frag_max_size > mtu))) { |
| 813 |
IP_INC_STATS(net, IPSTATS_MIB_FRAGFAILS); |
| 814 |
kfree_skb(skb); |
| 815 |
return -EMSGSIZE; |
| 816 |
} |
| 817 |
|
| 818 |
return ip_do_fragment(net, sk, skb, output); |
| 819 |
} |
| 820 |
|
| 821 |
static unsigned int nf_bridge_mtu_reduction(const struct sk_buff *skb) |
| 822 |
{ |
| 823 |
const struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb); |
| 824 |
|
| 825 |
if (nf_bridge->orig_proto == BRNF_PROTO_PPPOE) |
| 826 |
return PPPOE_SES_HLEN; |
| 827 |
return 0; |
| 828 |
} |
| 829 |
|
| 830 |
static int br_nf_dev_queue_xmit(struct net *net, struct sock *sk, struct sk_buff *skb) |
| 831 |
{ |
| 832 |
struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb); |
| 833 |
unsigned int mtu, mtu_reserved; |
| 834 |
int ret; |
| 835 |
|
| 836 |
mtu_reserved = nf_bridge_mtu_reduction(skb); |
| 837 |
mtu = skb->dev->mtu; |
| 838 |
|
| 839 |
if (nf_bridge->pkt_otherhost) { |
| 840 |
skb->pkt_type = PACKET_OTHERHOST; |
| 841 |
nf_bridge->pkt_otherhost = false; |
| 842 |
} |
| 843 |
|
| 844 |
if (nf_bridge->frag_max_size && nf_bridge->frag_max_size < mtu) |
| 845 |
mtu = nf_bridge->frag_max_size; |
| 846 |
|
| 847 |
nf_bridge_update_protocol(skb); |
| 848 |
nf_bridge_push_encap_header(skb); |
| 849 |
|
| 850 |
if (skb_is_gso(skb) || skb->len + mtu_reserved <= mtu) { |
| 851 |
nf_bridge_info_free(skb); |
| 852 |
return br_dev_queue_push_xmit(net, sk, skb); |
| 853 |
} |
| 854 |
|
| 855 |
/* Fragmentation on metadata/template dst is not supported */ |
| 856 |
if (unlikely(!skb_valid_dst(skb))) |
| 857 |
goto drop; |
| 858 |
|
| 859 |
/* This is wrong! We should preserve the original fragment |
| 860 |
* boundaries by preserving frag_list rather than refragmenting. |
| 861 |
*/ |
| 862 |
if (IS_ENABLED(CONFIG_NF_DEFRAG_IPV4) && |
| 863 |
skb->protocol == htons(ETH_P_IP)) { |
| 864 |
struct brnf_frag_data *data; |
| 865 |
|
| 866 |
if (br_validate_ipv4(net, skb)) |
| 867 |
goto drop; |
| 868 |
|
| 869 |
IPCB(skb)->frag_max_size = nf_bridge->frag_max_size; |
| 870 |
|
| 871 |
local_lock_nested_bh(&brnf_frag_data_storage.bh_lock); |
| 872 |
data = this_cpu_ptr(&brnf_frag_data_storage); |
| 873 |
|
| 874 |
if (skb_vlan_tag_present(skb)) { |
| 875 |
data->vlan_tci = skb->vlan_tci; |
| 876 |
data->vlan_proto = skb->vlan_proto; |
| 877 |
} else { |
| 878 |
data->vlan_proto = 0; |
| 879 |
} |
| 880 |
|
| 881 |
data->encap_size = nf_bridge_encap_header_len(skb); |
| 882 |
data->size = ETH_HLEN + data->encap_size; |
| 883 |
|
| 884 |
skb_copy_from_linear_data_offset(skb, -data->size, data->mac, |
| 885 |
data->size); |
| 886 |
|
| 887 |
ret = br_nf_ip_fragment(net, sk, skb, br_nf_push_frag_xmit); |
| 888 |
local_unlock_nested_bh(&brnf_frag_data_storage.bh_lock); |
| 889 |
return ret; |
| 890 |
} |
| 891 |
if (IS_ENABLED(CONFIG_NF_DEFRAG_IPV6) && |
| 892 |
skb->protocol == htons(ETH_P_IPV6)) { |
| 893 |
const struct nf_ipv6_ops *v6ops = nf_get_ipv6_ops(); |
| 894 |
struct brnf_frag_data *data; |
| 895 |
|
| 896 |
if (br_validate_ipv6(net, skb)) |
| 897 |
goto drop; |
| 898 |
|
| 899 |
IP6CB(skb)->frag_max_size = nf_bridge->frag_max_size; |
| 900 |
|
| 901 |
local_lock_nested_bh(&brnf_frag_data_storage.bh_lock); |
| 902 |
data = this_cpu_ptr(&brnf_frag_data_storage); |
| 903 |
data->encap_size = nf_bridge_encap_header_len(skb); |
| 904 |
data->size = ETH_HLEN + data->encap_size; |
| 905 |
|
| 906 |
skb_copy_from_linear_data_offset(skb, -data->size, data->mac, |
| 907 |
data->size); |
| 908 |
|
| 909 |
if (v6ops) { |
| 910 |
ret = v6ops->fragment(net, sk, skb, br_nf_push_frag_xmit); |
| 911 |
local_unlock_nested_bh(&brnf_frag_data_storage.bh_lock); |
| 912 |
return ret; |
| 913 |
} |
| 914 |
local_unlock_nested_bh(&brnf_frag_data_storage.bh_lock); |
| 915 |
|
| 916 |
kfree_skb(skb); |
| 917 |
return -EMSGSIZE; |
| 918 |
} |
| 919 |
nf_bridge_info_free(skb); |
| 920 |
return br_dev_queue_push_xmit(net, sk, skb); |
| 921 |
drop: |
| 922 |
kfree_skb(skb); |
| 923 |
return 0; |
| 924 |
} |
| 925 |
|
| 926 |
/* PF_BRIDGE/POST_ROUTING ********************************************/ |
| 927 |
static unsigned int br_nf_post_routing(void *priv, |
| 928 |
struct sk_buff *skb, |
| 929 |
const struct nf_hook_state *state) |
| 930 |
{ |
| 931 |
struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb); |
| 932 |
struct net_device *realoutdev = bridge_parent(skb->dev); |
| 933 |
u_int8_t pf; |
| 934 |
|
| 935 |
/* if nf_bridge is set, but ->physoutdev is NULL, this packet came in |
| 936 |
* on a bridge, but was delivered locally and is now being routed: |
| 937 |
* |
| 938 |
* POST_ROUTING was already invoked from the ip stack. |
| 939 |
*/ |
| 940 |
if (!nf_bridge || !nf_bridge->physoutdev) |
| 941 |
return NF_ACCEPT; |
| 942 |
|
| 943 |
if (!realoutdev) |
| 944 |
return NF_DROP_REASON(skb, SKB_DROP_REASON_DEV_READY, 0); |
| 945 |
|
| 946 |
if (IS_IP(skb) || is_vlan_ip(skb, state->net) || |
| 947 |
is_pppoe_ip(skb, state->net)) |
| 948 |
pf = NFPROTO_IPV4; |
| 949 |
else if (IS_IPV6(skb) || is_vlan_ipv6(skb, state->net) || |
| 950 |
is_pppoe_ipv6(skb, state->net)) |
| 951 |
pf = NFPROTO_IPV6; |
| 952 |
else |
| 953 |
return NF_ACCEPT; |
| 954 |
|
| 955 |
if (skb->pkt_type == PACKET_OTHERHOST) { |
| 956 |
skb->pkt_type = PACKET_HOST; |
| 957 |
nf_bridge->pkt_otherhost = true; |
| 958 |
} |
| 959 |
|
| 960 |
nf_bridge_pull_encap_header(skb); |
| 961 |
if (pf == NFPROTO_IPV4) |
| 962 |
skb->protocol = htons(ETH_P_IP); |
| 963 |
else |
| 964 |
skb->protocol = htons(ETH_P_IPV6); |
| 965 |
|
| 966 |
NF_HOOK(pf, NF_INET_POST_ROUTING, state->net, state->sk, skb, |
| 967 |
NULL, realoutdev, |
| 968 |
br_nf_dev_queue_xmit); |
| 969 |
|
| 970 |
return NF_STOLEN; |
| 971 |
} |
| 972 |
|
| 973 |
/* IP/SABOTAGE *****************************************************/ |
| 974 |
/* Don't hand locally destined packets to PF_INET(6)/PRE_ROUTING |
| 975 |
* for the second time. */ |
| 976 |
static unsigned int ip_sabotage_in(void *priv, |
| 977 |
struct sk_buff *skb, |
| 978 |
const struct nf_hook_state *state) |
| 979 |
{ |
| 980 |
struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb); |
| 981 |
|
| 982 |
if (nf_bridge) { |
| 983 |
if (nf_bridge->sabotage_in_done) |
| 984 |
return NF_ACCEPT; |
| 985 |
|
| 986 |
if (!nf_bridge->in_prerouting && |
| 987 |
!netif_is_l3_master(skb->dev) && |
| 988 |
!netif_is_l3_slave(skb->dev)) { |
| 989 |
nf_bridge->sabotage_in_done = 1; |
| 990 |
state->okfn(state->net, state->sk, skb); |
| 991 |
return NF_STOLEN; |
| 992 |
} |
| 993 |
} |
| 994 |
|
| 995 |
return NF_ACCEPT; |
| 996 |
} |
| 997 |
|
| 998 |
/* This is called when br_netfilter has called into iptables/netfilter, |
| 999 |
* and DNAT has taken place on a bridge-forwarded packet. |
| 1000 |
* |
| 1001 |
* neigh->output has created a new MAC header, with local br0 MAC |
| 1002 |
* as saddr. |
| 1003 |
* |
| 1004 |
* This restores the original MAC saddr of the bridged packet |
| 1005 |
* before invoking bridge forward logic to transmit the packet. |
| 1006 |
*/ |
| 1007 |
static void br_nf_pre_routing_finish_bridge_slow(struct sk_buff *skb) |
| 1008 |
{ |
| 1009 |
struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb); |
| 1010 |
struct net_device *br_indev; |
| 1011 |
|
| 1012 |
br_indev = nf_bridge_get_physindev(skb, dev_net(skb->dev)); |
| 1013 |
if (!br_indev) { |
| 1014 |
kfree_skb(skb); |
| 1015 |
return; |
| 1016 |
} |
| 1017 |
|
| 1018 |
skb_pull(skb, ETH_HLEN); |
| 1019 |
nf_bridge->bridged_dnat = 0; |
| 1020 |
|
| 1021 |
BUILD_BUG_ON(sizeof(nf_bridge->neigh_header) != (ETH_HLEN - ETH_ALEN)); |
| 1022 |
|
| 1023 |
skb_copy_to_linear_data_offset(skb, -(ETH_HLEN - ETH_ALEN), |
| 1024 |
nf_bridge->neigh_header, |
| 1025 |
ETH_HLEN - ETH_ALEN); |
| 1026 |
skb->dev = br_indev; |
| 1027 |
|
| 1028 |
nf_bridge->physoutdev = NULL; |
| 1029 |
br_handle_frame_finish(dev_net(skb->dev), NULL, skb); |
| 1030 |
} |
| 1031 |
|
| 1032 |
static int br_nf_dev_xmit(struct sk_buff *skb) |
| 1033 |
{ |
| 1034 |
const struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb); |
| 1035 |
|
| 1036 |
if (nf_bridge && nf_bridge->bridged_dnat) { |
| 1037 |
br_nf_pre_routing_finish_bridge_slow(skb); |
| 1038 |
return 1; |
| 1039 |
} |
| 1040 |
return 0; |
| 1041 |
} |
| 1042 |
|
| 1043 |
static const struct nf_br_ops br_ops = { |
| 1044 |
.br_dev_xmit_hook = br_nf_dev_xmit, |
| 1045 |
}; |
| 1046 |
|
| 1047 |
/* For br_nf_post_routing, we need (prio = NF_BR_PRI_LAST), because |
| 1048 |
* br_dev_queue_push_xmit is called afterwards */ |
| 1049 |
static const struct nf_hook_ops br_nf_ops[] = { |
| 1050 |
{ |
| 1051 |
.hook = br_nf_pre_routing, |
| 1052 |
.pf = NFPROTO_BRIDGE, |
| 1053 |
.hooknum = NF_BR_PRE_ROUTING, |
| 1054 |
.priority = NF_BR_PRI_BRNF, |
| 1055 |
}, |
| 1056 |
#if IS_ENABLED(CONFIG_NF_CONNTRACK) |
| 1057 |
{ |
| 1058 |
.hook = br_nf_local_in, |
| 1059 |
.pf = NFPROTO_BRIDGE, |
| 1060 |
.hooknum = NF_BR_LOCAL_IN, |
| 1061 |
.priority = NF_BR_PRI_LAST, |
| 1062 |
}, |
| 1063 |
#endif |
| 1064 |
{ |
| 1065 |
.hook = br_nf_forward, |
| 1066 |
.pf = NFPROTO_BRIDGE, |
| 1067 |
.hooknum = NF_BR_FORWARD, |
| 1068 |
.priority = NF_BR_PRI_BRNF, |
| 1069 |
}, |
| 1070 |
{ |
| 1071 |
.hook = br_nf_post_routing, |
| 1072 |
.pf = NFPROTO_BRIDGE, |
| 1073 |
.hooknum = NF_BR_POST_ROUTING, |
| 1074 |
.priority = NF_BR_PRI_LAST, |
| 1075 |
}, |
| 1076 |
{ |
| 1077 |
.hook = ip_sabotage_in, |
| 1078 |
.pf = NFPROTO_IPV4, |
| 1079 |
.hooknum = NF_INET_PRE_ROUTING, |
| 1080 |
.priority = NF_IP_PRI_FIRST, |
| 1081 |
}, |
| 1082 |
{ |
| 1083 |
.hook = ip_sabotage_in, |
| 1084 |
.pf = NFPROTO_IPV6, |
| 1085 |
.hooknum = NF_INET_PRE_ROUTING, |
| 1086 |
.priority = NF_IP6_PRI_FIRST, |
| 1087 |
}, |
| 1088 |
}; |
| 1089 |
|
| 1090 |
static int brnf_device_event(struct notifier_block *unused, unsigned long event, |
| 1091 |
void *ptr) |
| 1092 |
{ |
| 1093 |
struct net_device *dev = netdev_notifier_info_to_dev(ptr); |
| 1094 |
struct brnf_net *brnet; |
| 1095 |
struct net *net; |
| 1096 |
int ret; |
| 1097 |
|
| 1098 |
if (event != NETDEV_REGISTER || !netif_is_bridge_master(dev)) |
| 1099 |
return NOTIFY_DONE; |
| 1100 |
|
| 1101 |
ASSERT_RTNL(); |
| 1102 |
|
| 1103 |
net = dev_net(dev); |
| 1104 |
brnet = net_generic(net, brnf_net_id); |
| 1105 |
if (brnet->enabled) |
| 1106 |
return NOTIFY_OK; |
| 1107 |
|
| 1108 |
ret = nf_register_net_hooks(net, br_nf_ops, ARRAY_SIZE(br_nf_ops)); |
| 1109 |
if (ret) |
| 1110 |
return NOTIFY_BAD; |
| 1111 |
|
| 1112 |
brnet->enabled = true; |
| 1113 |
return NOTIFY_OK; |
| 1114 |
} |
| 1115 |
|
| 1116 |
static struct notifier_block brnf_notifier __read_mostly = { |
| 1117 |
.notifier_call = brnf_device_event, |
| 1118 |
}; |
| 1119 |
|
| 1120 |
/* recursively invokes nf_hook_slow (again), skipping already-called |
| 1121 |
* hooks (< NF_BR_PRI_BRNF). |
| 1122 |
* |
| 1123 |
* Called with rcu read lock held. |
| 1124 |
*/ |
| 1125 |
int br_nf_hook_thresh(unsigned int hook, struct net *net, |
| 1126 |
struct sock *sk, struct sk_buff *skb, |
| 1127 |
struct net_device *indev, |
| 1128 |
struct net_device *outdev, |
| 1129 |
int (*okfn)(struct net *, struct sock *, |
| 1130 |
struct sk_buff *)) |
| 1131 |
{ |
| 1132 |
const struct nf_hook_entries *e; |
| 1133 |
struct nf_hook_state state; |
| 1134 |
struct nf_hook_ops **ops; |
| 1135 |
unsigned int i; |
| 1136 |
int ret; |
| 1137 |
|
| 1138 |
e = rcu_dereference(net->nf.hooks_bridge[hook]); |
| 1139 |
if (!e) |
| 1140 |
return okfn(net, sk, skb); |
| 1141 |
|
| 1142 |
ops = nf_hook_entries_get_hook_ops(e); |
| 1143 |
for (i = 0; i < e->num_hook_entries; i++) { |
| 1144 |
/* These hooks have already been called */ |
| 1145 |
if (ops[i]->priority < NF_BR_PRI_BRNF) |
| 1146 |
continue; |
| 1147 |
|
| 1148 |
/* These hooks have not been called yet, run them. */ |
| 1149 |
if (ops[i]->priority > NF_BR_PRI_BRNF) |
| 1150 |
break; |
| 1151 |
|
| 1152 |
/* take a closer look at NF_BR_PRI_BRNF. */ |
| 1153 |
if (ops[i]->hook == br_nf_pre_routing) { |
| 1154 |
/* This hook diverted the skb to this function, |
| 1155 |
* hooks after this have not been run yet. |
| 1156 |
*/ |
| 1157 |
i++; |
| 1158 |
break; |
| 1159 |
} |
| 1160 |
} |
| 1161 |
|
| 1162 |
nf_hook_state_init(&state, hook, NFPROTO_BRIDGE, indev, outdev, |
| 1163 |
sk, net, okfn); |
| 1164 |
|
| 1165 |
ret = nf_hook_slow(skb, &state, e, i); |
| 1166 |
if (ret == 1) |
| 1167 |
ret = okfn(net, sk, skb); |
| 1168 |
|
| 1169 |
return ret; |
| 1170 |
} |
| 1171 |
|
| 1172 |
#ifdef CONFIG_SYSCTL |
| 1173 |
static |
| 1174 |
int brnf_sysctl_call_tables(const struct ctl_table *ctl, int write, |
| 1175 |
void *buffer, size_t *lenp, loff_t *ppos) |
| 1176 |
{ |
| 1177 |
int ret; |
| 1178 |
|
| 1179 |
ret = proc_dointvec(ctl, write, buffer, lenp, ppos); |
| 1180 |
|
| 1181 |
if (write && *(int *)(ctl->data)) |
| 1182 |
*(int *)(ctl->data) = 1; |
| 1183 |
return ret; |
| 1184 |
} |
| 1185 |
|
| 1186 |
static struct ctl_table brnf_table[] = { |
| 1187 |
{ |
| 1188 |
.procname = "bridge-nf-call-arptables", |
| 1189 |
.maxlen = sizeof(int), |
| 1190 |
.mode = 0644, |
| 1191 |
.proc_handler = brnf_sysctl_call_tables, |
| 1192 |
}, |
| 1193 |
{ |
| 1194 |
.procname = "bridge-nf-call-iptables", |
| 1195 |
.maxlen = sizeof(int), |
| 1196 |
.mode = 0644, |
| 1197 |
.proc_handler = brnf_sysctl_call_tables, |
| 1198 |
}, |
| 1199 |
{ |
| 1200 |
.procname = "bridge-nf-call-ip6tables", |
| 1201 |
.maxlen = sizeof(int), |
| 1202 |
.mode = 0644, |
| 1203 |
.proc_handler = brnf_sysctl_call_tables, |
| 1204 |
}, |
| 1205 |
{ |
| 1206 |
.procname = "bridge-nf-filter-vlan-tagged", |
| 1207 |
.maxlen = sizeof(int), |
| 1208 |
.mode = 0644, |
| 1209 |
.proc_handler = brnf_sysctl_call_tables, |
| 1210 |
}, |
| 1211 |
{ |
| 1212 |
.procname = "bridge-nf-filter-pppoe-tagged", |
| 1213 |
.maxlen = sizeof(int), |
| 1214 |
.mode = 0644, |
| 1215 |
.proc_handler = brnf_sysctl_call_tables, |
| 1216 |
}, |
| 1217 |
{ |
| 1218 |
.procname = "bridge-nf-pass-vlan-input-dev", |
| 1219 |
.maxlen = sizeof(int), |
| 1220 |
.mode = 0644, |
| 1221 |
.proc_handler = brnf_sysctl_call_tables, |
| 1222 |
}, |
| 1223 |
}; |
| 1224 |
|
| 1225 |
static inline void br_netfilter_sysctl_default(struct brnf_net *brnf) |
| 1226 |
{ |
| 1227 |
brnf->call_iptables = 1; |
| 1228 |
brnf->call_ip6tables = 1; |
| 1229 |
brnf->call_arptables = 1; |
| 1230 |
brnf->filter_vlan_tagged = 0; |
| 1231 |
brnf->filter_pppoe_tagged = 0; |
| 1232 |
brnf->pass_vlan_indev = 0; |
| 1233 |
} |
| 1234 |
|
| 1235 |
static int br_netfilter_sysctl_init_net(struct net *net) |
| 1236 |
{ |
| 1237 |
struct ctl_table *table = brnf_table; |
| 1238 |
struct brnf_net *brnet; |
| 1239 |
|
| 1240 |
if (!net_eq(net, &init_net)) { |
| 1241 |
table = kmemdup(table, sizeof(brnf_table), GFP_KERNEL); |
| 1242 |
if (!table) |
| 1243 |
return -ENOMEM; |
| 1244 |
} |
| 1245 |
|
| 1246 |
brnet = net_generic(net, brnf_net_id); |
| 1247 |
table[0].data = &brnet->call_arptables; |
| 1248 |
table[1].data = &brnet->call_iptables; |
| 1249 |
table[2].data = &brnet->call_ip6tables; |
| 1250 |
table[3].data = &brnet->filter_vlan_tagged; |
| 1251 |
table[4].data = &brnet->filter_pppoe_tagged; |
| 1252 |
table[5].data = &brnet->pass_vlan_indev; |
| 1253 |
|
| 1254 |
br_netfilter_sysctl_default(brnet); |
| 1255 |
|
| 1256 |
brnet->ctl_hdr = register_net_sysctl_sz(net, "net/bridge", table, |
| 1257 |
ARRAY_SIZE(brnf_table)); |
| 1258 |
if (!brnet->ctl_hdr) { |
| 1259 |
if (!net_eq(net, &init_net)) |
| 1260 |
kfree(table); |
| 1261 |
|
| 1262 |
return -ENOMEM; |
| 1263 |
} |
| 1264 |
|
| 1265 |
return 0; |
| 1266 |
} |
| 1267 |
|
| 1268 |
static void br_netfilter_sysctl_exit_net(struct net *net, |
| 1269 |
struct brnf_net *brnet) |
| 1270 |
{ |
| 1271 |
const struct ctl_table *table = brnet->ctl_hdr->ctl_table_arg; |
| 1272 |
|
| 1273 |
unregister_net_sysctl_table(brnet->ctl_hdr); |
| 1274 |
if (!net_eq(net, &init_net)) |
| 1275 |
kfree(table); |
| 1276 |
} |
| 1277 |
|
| 1278 |
static int __net_init brnf_init_net(struct net *net) |
| 1279 |
{ |
| 1280 |
return br_netfilter_sysctl_init_net(net); |
| 1281 |
} |
| 1282 |
#endif |
| 1283 |
|
| 1284 |
static void __net_exit brnf_exit_net(struct net *net) |
| 1285 |
{ |
| 1286 |
struct brnf_net *brnet; |
| 1287 |
|
| 1288 |
brnet = net_generic(net, brnf_net_id); |
| 1289 |
if (brnet->enabled) { |
| 1290 |
nf_unregister_net_hooks(net, br_nf_ops, ARRAY_SIZE(br_nf_ops)); |
| 1291 |
brnet->enabled = false; |
| 1292 |
} |
| 1293 |
|
| 1294 |
#ifdef CONFIG_SYSCTL |
| 1295 |
br_netfilter_sysctl_exit_net(net, brnet); |
| 1296 |
#endif |
| 1297 |
} |
| 1298 |
|
| 1299 |
static struct pernet_operations brnf_net_ops __read_mostly = { |
| 1300 |
#ifdef CONFIG_SYSCTL |
| 1301 |
.init = brnf_init_net, |
| 1302 |
#endif |
| 1303 |
.exit = brnf_exit_net, |
| 1304 |
.id = &brnf_net_id, |
| 1305 |
.size = sizeof(struct brnf_net), |
| 1306 |
}; |
| 1307 |
|
| 1308 |
static int __init br_netfilter_init(void) |
| 1309 |
{ |
| 1310 |
int ret; |
| 1311 |
|
| 1312 |
ret = register_pernet_subsys(&brnf_net_ops); |
| 1313 |
if (ret < 0) |
| 1314 |
return ret; |
| 1315 |
|
| 1316 |
ret = register_netdevice_notifier(&brnf_notifier); |
| 1317 |
if (ret < 0) { |
| 1318 |
unregister_pernet_subsys(&brnf_net_ops); |
| 1319 |
return ret; |
| 1320 |
} |
| 1321 |
|
| 1322 |
RCU_INIT_POINTER(nf_br_ops, &br_ops); |
| 1323 |
printk(KERN_NOTICE "Bridge firewalling registered\n"); |
| 1324 |
return 0; |
| 1325 |
} |
| 1326 |
|
| 1327 |
static void __exit br_netfilter_fini(void) |
| 1328 |
{ |
| 1329 |
RCU_INIT_POINTER(nf_br_ops, NULL); |
| 1330 |
unregister_netdevice_notifier(&brnf_notifier); |
| 1331 |
unregister_pernet_subsys(&brnf_net_ops); |
| 1332 |
} |
| 1333 |
|
| 1334 |
module_init(br_netfilter_init); |
| 1335 |
module_exit(br_netfilter_fini); |
| 1336 |
|
| 1337 |
MODULE_LICENSE("GPL"); |
| 1338 |
MODULE_AUTHOR("Lennert Buytenhek <[email protected]>"); |
| 1339 |
MODULE_AUTHOR("Bart De Schuymer <[email protected]>"); |
| 1340 |
MODULE_DESCRIPTION("Linux ethernet netfilter firewall bridge"); |