Commit | Line | Data |
---|---|---|
2874c5fd | 1 | // SPDX-License-Identifier: GPL-2.0-or-later |
2ad7bf36 | 2 | /* Copyright (c) 2014 Mahesh Bandewar <maheshb@google.com> |
2ad7bf36 MB |
3 | */ |
4 | ||
5 | #include "ipvlan.h" | |
6 | ||
207895fd | 7 | static u32 ipvlan_jhash_secret __read_mostly; |
2ad7bf36 MB |
8 | |
9 | void ipvlan_init_secret(void) | |
10 | { | |
11 | net_get_random_once(&ipvlan_jhash_secret, sizeof(ipvlan_jhash_secret)); | |
12 | } | |
13 | ||
235a9d89 | 14 | void ipvlan_count_rx(const struct ipvl_dev *ipvlan, |
2ad7bf36 MB |
15 | unsigned int len, bool success, bool mcast) |
16 | { | |
2ad7bf36 MB |
17 | if (likely(success)) { |
18 | struct ipvl_pcpu_stats *pcptr; | |
19 | ||
20 | pcptr = this_cpu_ptr(ipvlan->pcpu_stats); | |
21 | u64_stats_update_begin(&pcptr->syncp); | |
22 | pcptr->rx_pkts++; | |
23 | pcptr->rx_bytes += len; | |
24 | if (mcast) | |
25 | pcptr->rx_mcast++; | |
26 | u64_stats_update_end(&pcptr->syncp); | |
27 | } else { | |
28 | this_cpu_inc(ipvlan->pcpu_stats->rx_errs); | |
29 | } | |
30 | } | |
235a9d89 | 31 | EXPORT_SYMBOL_GPL(ipvlan_count_rx); |
2ad7bf36 | 32 | |
94333fac | 33 | #if IS_ENABLED(CONFIG_IPV6) |
2ad7bf36 MB |
34 | static u8 ipvlan_get_v6_hash(const void *iaddr) |
35 | { | |
36 | const struct in6_addr *ip6_addr = iaddr; | |
37 | ||
38 | return __ipv6_addr_jhash(ip6_addr, ipvlan_jhash_secret) & | |
39 | IPVLAN_HASH_MASK; | |
40 | } | |
94333fac MC |
41 | #else |
42 | static u8 ipvlan_get_v6_hash(const void *iaddr) | |
43 | { | |
44 | return 0; | |
45 | } | |
46 | #endif | |
2ad7bf36 MB |
47 | |
48 | static u8 ipvlan_get_v4_hash(const void *iaddr) | |
49 | { | |
50 | const struct in_addr *ip4_addr = iaddr; | |
51 | ||
52 | return jhash_1word(ip4_addr->s_addr, ipvlan_jhash_secret) & | |
53 | IPVLAN_HASH_MASK; | |
54 | } | |
55 | ||
94333fac MC |
56 | static bool addr_equal(bool is_v6, struct ipvl_addr *addr, const void *iaddr) |
57 | { | |
58 | if (!is_v6 && addr->atype == IPVL_IPV4) { | |
59 | struct in_addr *i4addr = (struct in_addr *)iaddr; | |
60 | ||
61 | return addr->ip4addr.s_addr == i4addr->s_addr; | |
62 | #if IS_ENABLED(CONFIG_IPV6) | |
63 | } else if (is_v6 && addr->atype == IPVL_IPV6) { | |
64 | struct in6_addr *i6addr = (struct in6_addr *)iaddr; | |
65 | ||
66 | return ipv6_addr_equal(&addr->ip6addr, i6addr); | |
67 | #endif | |
68 | } | |
69 | ||
70 | return false; | |
71 | } | |
72 | ||
ab5b7013 MB |
73 | static struct ipvl_addr *ipvlan_ht_addr_lookup(const struct ipvl_port *port, |
74 | const void *iaddr, bool is_v6) | |
2ad7bf36 MB |
75 | { |
76 | struct ipvl_addr *addr; | |
77 | u8 hash; | |
78 | ||
79 | hash = is_v6 ? ipvlan_get_v6_hash(iaddr) : | |
80 | ipvlan_get_v4_hash(iaddr); | |
94333fac MC |
81 | hlist_for_each_entry_rcu(addr, &port->hlhead[hash], hlnode) |
82 | if (addr_equal(is_v6, addr, iaddr)) | |
2ad7bf36 | 83 | return addr; |
2ad7bf36 MB |
84 | return NULL; |
85 | } | |
86 | ||
87 | void ipvlan_ht_addr_add(struct ipvl_dev *ipvlan, struct ipvl_addr *addr) | |
88 | { | |
89 | struct ipvl_port *port = ipvlan->port; | |
90 | u8 hash; | |
91 | ||
92 | hash = (addr->atype == IPVL_IPV6) ? | |
93 | ipvlan_get_v6_hash(&addr->ip6addr) : | |
94 | ipvlan_get_v4_hash(&addr->ip4addr); | |
27705f70 JB |
95 | if (hlist_unhashed(&addr->hlnode)) |
96 | hlist_add_head_rcu(&addr->hlnode, &port->hlhead[hash]); | |
2ad7bf36 MB |
97 | } |
98 | ||
6640e673 | 99 | void ipvlan_ht_addr_del(struct ipvl_addr *addr) |
2ad7bf36 | 100 | { |
27705f70 | 101 | hlist_del_init_rcu(&addr->hlnode); |
2ad7bf36 MB |
102 | } |
103 | ||
e9997c29 JB |
104 | struct ipvl_addr *ipvlan_find_addr(const struct ipvl_dev *ipvlan, |
105 | const void *iaddr, bool is_v6) | |
2ad7bf36 | 106 | { |
82308194 PA |
107 | struct ipvl_addr *addr, *ret = NULL; |
108 | ||
109 | rcu_read_lock(); | |
110 | list_for_each_entry_rcu(addr, &ipvlan->addrs, anode) { | |
111 | if (addr_equal(is_v6, addr, iaddr)) { | |
112 | ret = addr; | |
113 | break; | |
114 | } | |
115 | } | |
116 | rcu_read_unlock(); | |
117 | return ret; | |
e9997c29 | 118 | } |
2ad7bf36 | 119 | |
e9997c29 JB |
120 | bool ipvlan_addr_busy(struct ipvl_port *port, void *iaddr, bool is_v6) |
121 | { | |
122 | struct ipvl_dev *ipvlan; | |
82308194 PA |
123 | bool ret = false; |
124 | ||
125 | rcu_read_lock(); | |
126 | list_for_each_entry_rcu(ipvlan, &port->ipvlans, pnode) { | |
127 | if (ipvlan_find_addr(ipvlan, iaddr, is_v6)) { | |
128 | ret = true; | |
129 | break; | |
130 | } | |
e9997c29 | 131 | } |
82308194 PA |
132 | rcu_read_unlock(); |
133 | return ret; | |
2ad7bf36 MB |
134 | } |
135 | ||
c675e06a | 136 | void *ipvlan_get_L3_hdr(struct ipvl_port *port, struct sk_buff *skb, int *type) |
2ad7bf36 MB |
137 | { |
138 | void *lyr3h = NULL; | |
139 | ||
140 | switch (skb->protocol) { | |
141 | case htons(ETH_P_ARP): { | |
142 | struct arphdr *arph; | |
143 | ||
5fc9220a | 144 | if (unlikely(!pskb_may_pull(skb, arp_hdr_len(port->dev)))) |
2ad7bf36 MB |
145 | return NULL; |
146 | ||
147 | arph = arp_hdr(skb); | |
148 | *type = IPVL_ARP; | |
149 | lyr3h = arph; | |
150 | break; | |
151 | } | |
152 | case htons(ETH_P_IP): { | |
153 | u32 pktlen; | |
154 | struct iphdr *ip4h; | |
155 | ||
156 | if (unlikely(!pskb_may_pull(skb, sizeof(*ip4h)))) | |
157 | return NULL; | |
158 | ||
159 | ip4h = ip_hdr(skb); | |
160 | pktlen = ntohs(ip4h->tot_len); | |
161 | if (ip4h->ihl < 5 || ip4h->version != 4) | |
162 | return NULL; | |
163 | if (skb->len < pktlen || pktlen < (ip4h->ihl * 4)) | |
164 | return NULL; | |
165 | ||
166 | *type = IPVL_IPV4; | |
167 | lyr3h = ip4h; | |
168 | break; | |
169 | } | |
94333fac | 170 | #if IS_ENABLED(CONFIG_IPV6) |
2ad7bf36 MB |
171 | case htons(ETH_P_IPV6): { |
172 | struct ipv6hdr *ip6h; | |
173 | ||
174 | if (unlikely(!pskb_may_pull(skb, sizeof(*ip6h)))) | |
175 | return NULL; | |
176 | ||
177 | ip6h = ipv6_hdr(skb); | |
178 | if (ip6h->version != 6) | |
179 | return NULL; | |
180 | ||
181 | *type = IPVL_IPV6; | |
182 | lyr3h = ip6h; | |
183 | /* Only Neighbour Solicitation pkts need different treatment */ | |
184 | if (ipv6_addr_any(&ip6h->saddr) && | |
185 | ip6h->nexthdr == NEXTHDR_ICMP) { | |
747a7135 GF |
186 | struct icmp6hdr *icmph; |
187 | ||
188 | if (unlikely(!pskb_may_pull(skb, sizeof(*ip6h) + sizeof(*icmph)))) | |
189 | return NULL; | |
190 | ||
191 | ip6h = ipv6_hdr(skb); | |
192 | icmph = (struct icmp6hdr *)(ip6h + 1); | |
193 | ||
194 | if (icmph->icmp6_type == NDISC_NEIGHBOUR_SOLICITATION) { | |
195 | /* Need to access the ipv6 address in body */ | |
196 | if (unlikely(!pskb_may_pull(skb, sizeof(*ip6h) + sizeof(*icmph) | |
197 | + sizeof(struct in6_addr)))) | |
198 | return NULL; | |
199 | ||
200 | ip6h = ipv6_hdr(skb); | |
201 | icmph = (struct icmp6hdr *)(ip6h + 1); | |
202 | } | |
203 | ||
2ad7bf36 | 204 | *type = IPVL_ICMPV6; |
747a7135 | 205 | lyr3h = icmph; |
2ad7bf36 MB |
206 | } |
207 | break; | |
208 | } | |
94333fac | 209 | #endif |
2ad7bf36 MB |
210 | default: |
211 | return NULL; | |
212 | } | |
213 | ||
214 | return lyr3h; | |
215 | } | |
216 | ||
217 | unsigned int ipvlan_mac_hash(const unsigned char *addr) | |
218 | { | |
219 | u32 hash = jhash_1word(__get_unaligned_cpu32(addr+2), | |
220 | ipvlan_jhash_secret); | |
221 | ||
222 | return hash & IPVLAN_MAC_FILTER_MASK; | |
223 | } | |
224 | ||
ba35f858 | 225 | void ipvlan_process_multicast(struct work_struct *work) |
2ad7bf36 | 226 | { |
ba35f858 MB |
227 | struct ipvl_port *port = container_of(work, struct ipvl_port, wq); |
228 | struct ethhdr *ethh; | |
2ad7bf36 | 229 | struct ipvl_dev *ipvlan; |
ba35f858 MB |
230 | struct sk_buff *skb, *nskb; |
231 | struct sk_buff_head list; | |
2ad7bf36 MB |
232 | unsigned int len; |
233 | unsigned int mac_hash; | |
234 | int ret; | |
ba35f858 | 235 | u8 pkt_type; |
e2525360 | 236 | bool tx_pkt; |
2ad7bf36 | 237 | |
ba35f858 | 238 | __skb_queue_head_init(&list); |
2ad7bf36 | 239 | |
ba35f858 MB |
240 | spin_lock_bh(&port->backlog.lock); |
241 | skb_queue_splice_tail_init(&port->backlog, &list); | |
242 | spin_unlock_bh(&port->backlog.lock); | |
2ad7bf36 | 243 | |
ba35f858 | 244 | while ((skb = __skb_dequeue(&list)) != NULL) { |
b1227d01 ED |
245 | struct net_device *dev = skb->dev; |
246 | bool consumed = false; | |
247 | ||
ba35f858 | 248 | ethh = eth_hdr(skb); |
e2525360 | 249 | tx_pkt = IPVL_SKB_CB(skb)->tx_pkt; |
ba35f858 | 250 | mac_hash = ipvlan_mac_hash(ethh->h_dest); |
2ad7bf36 | 251 | |
ba35f858 MB |
252 | if (ether_addr_equal(ethh->h_dest, port->dev->broadcast)) |
253 | pkt_type = PACKET_BROADCAST; | |
2ad7bf36 | 254 | else |
ba35f858 MB |
255 | pkt_type = PACKET_MULTICAST; |
256 | ||
ba35f858 MB |
257 | rcu_read_lock(); |
258 | list_for_each_entry_rcu(ipvlan, &port->ipvlans, pnode) { | |
e2525360 | 259 | if (tx_pkt && (ipvlan->dev == skb->dev)) |
ba35f858 | 260 | continue; |
ba35f858 MB |
261 | if (!test_bit(mac_hash, ipvlan->mac_filters)) |
262 | continue; | |
b1227d01 ED |
263 | if (!(ipvlan->dev->flags & IFF_UP)) |
264 | continue; | |
ba35f858 MB |
265 | ret = NET_RX_DROP; |
266 | len = skb->len + ETH_HLEN; | |
267 | nskb = skb_clone(skb, GFP_ATOMIC); | |
b1227d01 ED |
268 | local_bh_disable(); |
269 | if (nskb) { | |
270 | consumed = true; | |
271 | nskb->pkt_type = pkt_type; | |
272 | nskb->dev = ipvlan->dev; | |
e2525360 | 273 | if (tx_pkt) |
b1227d01 ED |
274 | ret = dev_forward_skb(ipvlan->dev, nskb); |
275 | else | |
276 | ret = netif_rx(nskb); | |
277 | } | |
ba35f858 | 278 | ipvlan_count_rx(ipvlan, len, ret == NET_RX_SUCCESS, true); |
b1227d01 | 279 | local_bh_enable(); |
ba35f858 MB |
280 | } |
281 | rcu_read_unlock(); | |
282 | ||
e2525360 | 283 | if (tx_pkt) { |
ba35f858 MB |
284 | /* If the packet originated here, send it out. */ |
285 | skb->dev = port->dev; | |
286 | skb->pkt_type = pkt_type; | |
287 | dev_queue_xmit(skb); | |
288 | } else { | |
b1227d01 ED |
289 | if (consumed) |
290 | consume_skb(skb); | |
291 | else | |
292 | kfree_skb(skb); | |
2ad7bf36 | 293 | } |
b1227d01 ED |
294 | if (dev) |
295 | dev_put(dev); | |
2ad7bf36 MB |
296 | } |
297 | } | |
298 | ||
b93dd49c MB |
299 | static void ipvlan_skb_crossing_ns(struct sk_buff *skb, struct net_device *dev) |
300 | { | |
301 | bool xnet = true; | |
302 | ||
303 | if (dev) | |
304 | xnet = !net_eq(dev_net(skb->dev), dev_net(dev)); | |
305 | ||
306 | skb_scrub_packet(skb, xnet); | |
307 | if (dev) | |
308 | skb->dev = dev; | |
309 | } | |
310 | ||
cf554ada | 311 | static int ipvlan_rcv_frame(struct ipvl_addr *addr, struct sk_buff **pskb, |
2ad7bf36 MB |
312 | bool local) |
313 | { | |
314 | struct ipvl_dev *ipvlan = addr->master; | |
315 | struct net_device *dev = ipvlan->dev; | |
316 | unsigned int len; | |
317 | rx_handler_result_t ret = RX_HANDLER_CONSUMED; | |
318 | bool success = false; | |
cf554ada | 319 | struct sk_buff *skb = *pskb; |
2ad7bf36 MB |
320 | |
321 | len = skb->len + ETH_HLEN; | |
ab5b7013 MB |
322 | /* Only packets exchanged between two local slaves need to have |
323 | * device-up check as well as skb-share check. | |
324 | */ | |
325 | if (local) { | |
326 | if (unlikely(!(dev->flags & IFF_UP))) { | |
327 | kfree_skb(skb); | |
328 | goto out; | |
329 | } | |
2ad7bf36 | 330 | |
ab5b7013 MB |
331 | skb = skb_share_check(skb, GFP_ATOMIC); |
332 | if (!skb) | |
333 | goto out; | |
2ad7bf36 | 334 | |
ab5b7013 MB |
335 | *pskb = skb; |
336 | } | |
2ad7bf36 MB |
337 | |
338 | if (local) { | |
ab5b7013 | 339 | skb->pkt_type = PACKET_HOST; |
2ad7bf36 MB |
340 | if (dev_forward_skb(ipvlan->dev, skb) == NET_RX_SUCCESS) |
341 | success = true; | |
342 | } else { | |
c0d451c8 | 343 | skb->dev = dev; |
2ad7bf36 MB |
344 | ret = RX_HANDLER_ANOTHER; |
345 | success = true; | |
346 | } | |
347 | ||
348 | out: | |
349 | ipvlan_count_rx(ipvlan, len, success, false); | |
350 | return ret; | |
351 | } | |
352 | ||
c675e06a DB |
353 | struct ipvl_addr *ipvlan_addr_lookup(struct ipvl_port *port, void *lyr3h, |
354 | int addr_type, bool use_dest) | |
2ad7bf36 MB |
355 | { |
356 | struct ipvl_addr *addr = NULL; | |
357 | ||
94333fac MC |
358 | switch (addr_type) { |
359 | #if IS_ENABLED(CONFIG_IPV6) | |
360 | case IPVL_IPV6: { | |
2ad7bf36 MB |
361 | struct ipv6hdr *ip6h; |
362 | struct in6_addr *i6addr; | |
363 | ||
364 | ip6h = (struct ipv6hdr *)lyr3h; | |
365 | i6addr = use_dest ? &ip6h->daddr : &ip6h->saddr; | |
366 | addr = ipvlan_ht_addr_lookup(port, i6addr, true); | |
94333fac MC |
367 | break; |
368 | } | |
369 | case IPVL_ICMPV6: { | |
2ad7bf36 MB |
370 | struct nd_msg *ndmh; |
371 | struct in6_addr *i6addr; | |
372 | ||
373 | /* Make sure that the NeighborSolicitation ICMPv6 packets | |
374 | * are handled to avoid DAD issue. | |
375 | */ | |
376 | ndmh = (struct nd_msg *)lyr3h; | |
377 | if (ndmh->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION) { | |
378 | i6addr = &ndmh->target; | |
379 | addr = ipvlan_ht_addr_lookup(port, i6addr, true); | |
380 | } | |
94333fac MC |
381 | break; |
382 | } | |
383 | #endif | |
384 | case IPVL_IPV4: { | |
2ad7bf36 MB |
385 | struct iphdr *ip4h; |
386 | __be32 *i4addr; | |
387 | ||
388 | ip4h = (struct iphdr *)lyr3h; | |
389 | i4addr = use_dest ? &ip4h->daddr : &ip4h->saddr; | |
390 | addr = ipvlan_ht_addr_lookup(port, i4addr, false); | |
94333fac MC |
391 | break; |
392 | } | |
393 | case IPVL_ARP: { | |
2ad7bf36 MB |
394 | struct arphdr *arph; |
395 | unsigned char *arp_ptr; | |
396 | __be32 dip; | |
397 | ||
398 | arph = (struct arphdr *)lyr3h; | |
399 | arp_ptr = (unsigned char *)(arph + 1); | |
400 | if (use_dest) | |
401 | arp_ptr += (2 * port->dev->addr_len) + 4; | |
402 | else | |
403 | arp_ptr += port->dev->addr_len; | |
404 | ||
405 | memcpy(&dip, arp_ptr, 4); | |
406 | addr = ipvlan_ht_addr_lookup(port, &dip, false); | |
94333fac MC |
407 | break; |
408 | } | |
2ad7bf36 MB |
409 | } |
410 | ||
411 | return addr; | |
412 | } | |
413 | ||
b93dd49c | 414 | static int ipvlan_process_v4_outbound(struct sk_buff *skb) |
2ad7bf36 MB |
415 | { |
416 | const struct iphdr *ip4h = ip_hdr(skb); | |
417 | struct net_device *dev = skb->dev; | |
57c4bf85 | 418 | struct net *net = dev_net(dev); |
2ad7bf36 MB |
419 | struct rtable *rt; |
420 | int err, ret = NET_XMIT_DROP; | |
421 | struct flowi4 fl4 = { | |
63b11e75 | 422 | .flowi4_oif = dev->ifindex, |
2ad7bf36 MB |
423 | .flowi4_tos = RT_TOS(ip4h->tos), |
424 | .flowi4_flags = FLOWI_FLAG_ANYSRC, | |
a98a4ebc | 425 | .flowi4_mark = skb->mark, |
2ad7bf36 MB |
426 | .daddr = ip4h->daddr, |
427 | .saddr = ip4h->saddr, | |
428 | }; | |
429 | ||
57c4bf85 | 430 | rt = ip_route_output_flow(net, &fl4, NULL); |
2ad7bf36 MB |
431 | if (IS_ERR(rt)) |
432 | goto err; | |
433 | ||
434 | if (rt->rt_type != RTN_UNICAST && rt->rt_type != RTN_LOCAL) { | |
435 | ip_rt_put(rt); | |
436 | goto err; | |
437 | } | |
2ad7bf36 | 438 | skb_dst_set(skb, &rt->dst); |
33224b16 | 439 | err = ip_local_out(net, skb->sk, skb); |
2ad7bf36 MB |
440 | if (unlikely(net_xmit_eval(err))) |
441 | dev->stats.tx_errors++; | |
442 | else | |
443 | ret = NET_XMIT_SUCCESS; | |
444 | goto out; | |
445 | err: | |
446 | dev->stats.tx_errors++; | |
447 | kfree_skb(skb); | |
448 | out: | |
449 | return ret; | |
450 | } | |
451 | ||
94333fac | 452 | #if IS_ENABLED(CONFIG_IPV6) |
b93dd49c | 453 | static int ipvlan_process_v6_outbound(struct sk_buff *skb) |
2ad7bf36 MB |
454 | { |
455 | const struct ipv6hdr *ip6h = ipv6_hdr(skb); | |
456 | struct net_device *dev = skb->dev; | |
57c4bf85 | 457 | struct net *net = dev_net(dev); |
2ad7bf36 MB |
458 | struct dst_entry *dst; |
459 | int err, ret = NET_XMIT_DROP; | |
460 | struct flowi6 fl6 = { | |
ca29fd7c | 461 | .flowi6_oif = dev->ifindex, |
2ad7bf36 MB |
462 | .daddr = ip6h->daddr, |
463 | .saddr = ip6h->saddr, | |
464 | .flowi6_flags = FLOWI_FLAG_ANYSRC, | |
465 | .flowlabel = ip6_flowinfo(ip6h), | |
466 | .flowi6_mark = skb->mark, | |
467 | .flowi6_proto = ip6h->nexthdr, | |
468 | }; | |
469 | ||
57c4bf85 | 470 | dst = ip6_route_output(net, NULL, &fl6); |
2aab9525 MB |
471 | if (dst->error) { |
472 | ret = dst->error; | |
473 | dst_release(dst); | |
2ad7bf36 | 474 | goto err; |
2aab9525 | 475 | } |
2ad7bf36 | 476 | skb_dst_set(skb, dst); |
33224b16 | 477 | err = ip6_local_out(net, skb->sk, skb); |
2ad7bf36 MB |
478 | if (unlikely(net_xmit_eval(err))) |
479 | dev->stats.tx_errors++; | |
480 | else | |
481 | ret = NET_XMIT_SUCCESS; | |
482 | goto out; | |
483 | err: | |
484 | dev->stats.tx_errors++; | |
485 | kfree_skb(skb); | |
486 | out: | |
487 | return ret; | |
488 | } | |
94333fac MC |
489 | #else |
490 | static int ipvlan_process_v6_outbound(struct sk_buff *skb) | |
491 | { | |
492 | return NET_XMIT_DROP; | |
493 | } | |
494 | #endif | |
2ad7bf36 | 495 | |
b93dd49c | 496 | static int ipvlan_process_outbound(struct sk_buff *skb) |
2ad7bf36 MB |
497 | { |
498 | struct ethhdr *ethh = eth_hdr(skb); | |
499 | int ret = NET_XMIT_DROP; | |
500 | ||
501 | /* In this mode we dont care about multicast and broadcast traffic */ | |
502 | if (is_multicast_ether_addr(ethh->h_dest)) { | |
cccc200f PA |
503 | pr_debug_ratelimited("Dropped {multi|broad}cast of type=[%x]\n", |
504 | ntohs(skb->protocol)); | |
2ad7bf36 MB |
505 | kfree_skb(skb); |
506 | goto out; | |
507 | } | |
508 | ||
509 | /* The ipvlan is a pseudo-L2 device, so the packets that we receive | |
510 | * will have L2; which need to discarded and processed further | |
511 | * in the net-ns of the main-device. | |
512 | */ | |
513 | if (skb_mac_header_was_set(skb)) { | |
514 | skb_pull(skb, sizeof(*ethh)); | |
515 | skb->mac_header = (typeof(skb->mac_header))~0U; | |
516 | skb_reset_network_header(skb); | |
517 | } | |
518 | ||
519 | if (skb->protocol == htons(ETH_P_IPV6)) | |
b93dd49c | 520 | ret = ipvlan_process_v6_outbound(skb); |
2ad7bf36 | 521 | else if (skb->protocol == htons(ETH_P_IP)) |
b93dd49c | 522 | ret = ipvlan_process_v4_outbound(skb); |
2ad7bf36 MB |
523 | else { |
524 | pr_warn_ratelimited("Dropped outbound packet type=%x\n", | |
525 | ntohs(skb->protocol)); | |
526 | kfree_skb(skb); | |
527 | } | |
528 | out: | |
529 | return ret; | |
530 | } | |
531 | ||
ba35f858 | 532 | static void ipvlan_multicast_enqueue(struct ipvl_port *port, |
e2525360 | 533 | struct sk_buff *skb, bool tx_pkt) |
ba35f858 MB |
534 | { |
535 | if (skb->protocol == htons(ETH_P_PAUSE)) { | |
536 | kfree_skb(skb); | |
537 | return; | |
538 | } | |
539 | ||
e2525360 MB |
540 | /* Record that the deferred packet is from TX or RX path. By |
541 | * looking at mac-addresses on packet will lead to erronus decisions. | |
542 | * (This would be true for a loopback-mode on master device or a | |
543 | * hair-pin mode of the switch.) | |
544 | */ | |
545 | IPVL_SKB_CB(skb)->tx_pkt = tx_pkt; | |
546 | ||
ba35f858 MB |
547 | spin_lock(&port->backlog.lock); |
548 | if (skb_queue_len(&port->backlog) < IPVLAN_QBACKLOG_LIMIT) { | |
b1227d01 ED |
549 | if (skb->dev) |
550 | dev_hold(skb->dev); | |
ba35f858 MB |
551 | __skb_queue_tail(&port->backlog, skb); |
552 | spin_unlock(&port->backlog.lock); | |
553 | schedule_work(&port->wq); | |
554 | } else { | |
555 | spin_unlock(&port->backlog.lock); | |
556 | atomic_long_inc(&skb->dev->rx_dropped); | |
557 | kfree_skb(skb); | |
558 | } | |
559 | } | |
560 | ||
2ad7bf36 MB |
561 | static int ipvlan_xmit_mode_l3(struct sk_buff *skb, struct net_device *dev) |
562 | { | |
563 | const struct ipvl_dev *ipvlan = netdev_priv(dev); | |
564 | void *lyr3h; | |
565 | struct ipvl_addr *addr; | |
566 | int addr_type; | |
567 | ||
5fc9220a | 568 | lyr3h = ipvlan_get_L3_hdr(ipvlan->port, skb, &addr_type); |
2ad7bf36 MB |
569 | if (!lyr3h) |
570 | goto out; | |
571 | ||
fe89aa6b MB |
572 | if (!ipvlan_is_vepa(ipvlan->port)) { |
573 | addr = ipvlan_addr_lookup(ipvlan->port, lyr3h, addr_type, true); | |
574 | if (addr) { | |
575 | if (ipvlan_is_private(ipvlan->port)) { | |
576 | consume_skb(skb); | |
577 | return NET_XMIT_DROP; | |
578 | } | |
579 | return ipvlan_rcv_frame(addr, &skb, true); | |
a190d04d | 580 | } |
a190d04d | 581 | } |
2ad7bf36 | 582 | out: |
b93dd49c MB |
583 | ipvlan_skb_crossing_ns(skb, ipvlan->phy_dev); |
584 | return ipvlan_process_outbound(skb); | |
2ad7bf36 MB |
585 | } |
586 | ||
587 | static int ipvlan_xmit_mode_l2(struct sk_buff *skb, struct net_device *dev) | |
588 | { | |
589 | const struct ipvl_dev *ipvlan = netdev_priv(dev); | |
590 | struct ethhdr *eth = eth_hdr(skb); | |
591 | struct ipvl_addr *addr; | |
592 | void *lyr3h; | |
593 | int addr_type; | |
594 | ||
fe89aa6b MB |
595 | if (!ipvlan_is_vepa(ipvlan->port) && |
596 | ether_addr_equal(eth->h_dest, eth->h_source)) { | |
5fc9220a | 597 | lyr3h = ipvlan_get_L3_hdr(ipvlan->port, skb, &addr_type); |
2ad7bf36 MB |
598 | if (lyr3h) { |
599 | addr = ipvlan_addr_lookup(ipvlan->port, lyr3h, addr_type, true); | |
a190d04d MB |
600 | if (addr) { |
601 | if (ipvlan_is_private(ipvlan->port)) { | |
602 | consume_skb(skb); | |
603 | return NET_XMIT_DROP; | |
604 | } | |
cf554ada | 605 | return ipvlan_rcv_frame(addr, &skb, true); |
a190d04d | 606 | } |
2ad7bf36 MB |
607 | } |
608 | skb = skb_share_check(skb, GFP_ATOMIC); | |
609 | if (!skb) | |
610 | return NET_XMIT_DROP; | |
611 | ||
612 | /* Packet definitely does not belong to any of the | |
613 | * virtual devices, but the dest is local. So forward | |
614 | * the skb for the main-dev. At the RX side we just return | |
615 | * RX_PASS for it to be processed further on the stack. | |
616 | */ | |
617 | return dev_forward_skb(ipvlan->phy_dev, skb); | |
618 | ||
619 | } else if (is_multicast_ether_addr(eth->h_dest)) { | |
b93dd49c | 620 | ipvlan_skb_crossing_ns(skb, NULL); |
e2525360 | 621 | ipvlan_multicast_enqueue(ipvlan->port, skb, true); |
ba35f858 | 622 | return NET_XMIT_SUCCESS; |
2ad7bf36 MB |
623 | } |
624 | ||
c0d451c8 | 625 | skb->dev = ipvlan->phy_dev; |
2ad7bf36 MB |
626 | return dev_queue_xmit(skb); |
627 | } | |
628 | ||
629 | int ipvlan_queue_xmit(struct sk_buff *skb, struct net_device *dev) | |
630 | { | |
631 | struct ipvl_dev *ipvlan = netdev_priv(dev); | |
0fba37a3 | 632 | struct ipvl_port *port = ipvlan_port_get_rcu_bh(ipvlan->phy_dev); |
2ad7bf36 MB |
633 | |
634 | if (!port) | |
635 | goto out; | |
636 | ||
637 | if (unlikely(!pskb_may_pull(skb, sizeof(struct ethhdr)))) | |
638 | goto out; | |
639 | ||
640 | switch(port->mode) { | |
641 | case IPVLAN_MODE_L2: | |
642 | return ipvlan_xmit_mode_l2(skb, dev); | |
643 | case IPVLAN_MODE_L3: | |
c675e06a | 644 | #ifdef CONFIG_IPVLAN_L3S |
4fbae7d8 | 645 | case IPVLAN_MODE_L3S: |
c675e06a | 646 | #endif |
2ad7bf36 MB |
647 | return ipvlan_xmit_mode_l3(skb, dev); |
648 | } | |
649 | ||
650 | /* Should not reach here */ | |
651 | WARN_ONCE(true, "ipvlan_queue_xmit() called for mode = [%hx]\n", | |
652 | port->mode); | |
653 | out: | |
654 | kfree_skb(skb); | |
655 | return NET_XMIT_DROP; | |
656 | } | |
657 | ||
658 | static bool ipvlan_external_frame(struct sk_buff *skb, struct ipvl_port *port) | |
659 | { | |
660 | struct ethhdr *eth = eth_hdr(skb); | |
661 | struct ipvl_addr *addr; | |
662 | void *lyr3h; | |
663 | int addr_type; | |
664 | ||
665 | if (ether_addr_equal(eth->h_source, skb->dev->dev_addr)) { | |
5fc9220a | 666 | lyr3h = ipvlan_get_L3_hdr(port, skb, &addr_type); |
2ad7bf36 MB |
667 | if (!lyr3h) |
668 | return true; | |
669 | ||
670 | addr = ipvlan_addr_lookup(port, lyr3h, addr_type, false); | |
671 | if (addr) | |
672 | return false; | |
673 | } | |
674 | ||
675 | return true; | |
676 | } | |
677 | ||
678 | static rx_handler_result_t ipvlan_handle_mode_l3(struct sk_buff **pskb, | |
679 | struct ipvl_port *port) | |
680 | { | |
681 | void *lyr3h; | |
682 | int addr_type; | |
683 | struct ipvl_addr *addr; | |
684 | struct sk_buff *skb = *pskb; | |
685 | rx_handler_result_t ret = RX_HANDLER_PASS; | |
686 | ||
5fc9220a | 687 | lyr3h = ipvlan_get_L3_hdr(port, skb, &addr_type); |
2ad7bf36 MB |
688 | if (!lyr3h) |
689 | goto out; | |
690 | ||
691 | addr = ipvlan_addr_lookup(port, lyr3h, addr_type, true); | |
692 | if (addr) | |
cf554ada | 693 | ret = ipvlan_rcv_frame(addr, pskb, false); |
2ad7bf36 MB |
694 | |
695 | out: | |
696 | return ret; | |
697 | } | |
698 | ||
699 | static rx_handler_result_t ipvlan_handle_mode_l2(struct sk_buff **pskb, | |
700 | struct ipvl_port *port) | |
701 | { | |
702 | struct sk_buff *skb = *pskb; | |
703 | struct ethhdr *eth = eth_hdr(skb); | |
704 | rx_handler_result_t ret = RX_HANDLER_PASS; | |
2ad7bf36 MB |
705 | |
706 | if (is_multicast_ether_addr(eth->h_dest)) { | |
ba35f858 MB |
707 | if (ipvlan_external_frame(skb, port)) { |
708 | struct sk_buff *nskb = skb_clone(skb, GFP_ATOMIC); | |
709 | ||
710 | /* External frames are queued for device local | |
711 | * distribution, but a copy is given to master | |
712 | * straight away to avoid sending duplicates later | |
713 | * when work-queue processes this frame. This is | |
714 | * achieved by returning RX_HANDLER_PASS. | |
715 | */ | |
b93dd49c MB |
716 | if (nskb) { |
717 | ipvlan_skb_crossing_ns(nskb, NULL); | |
e2525360 | 718 | ipvlan_multicast_enqueue(port, nskb, false); |
b93dd49c | 719 | } |
ba35f858 | 720 | } |
2ad7bf36 | 721 | } else { |
24e5992a GF |
722 | /* Perform like l3 mode for non-multicast packet */ |
723 | ret = ipvlan_handle_mode_l3(pskb, port); | |
2ad7bf36 MB |
724 | } |
725 | ||
726 | return ret; | |
727 | } | |
728 | ||
729 | rx_handler_result_t ipvlan_handle_frame(struct sk_buff **pskb) | |
730 | { | |
731 | struct sk_buff *skb = *pskb; | |
732 | struct ipvl_port *port = ipvlan_port_get_rcu(skb->dev); | |
733 | ||
734 | if (!port) | |
735 | return RX_HANDLER_PASS; | |
736 | ||
737 | switch (port->mode) { | |
738 | case IPVLAN_MODE_L2: | |
739 | return ipvlan_handle_mode_l2(pskb, port); | |
740 | case IPVLAN_MODE_L3: | |
741 | return ipvlan_handle_mode_l3(pskb, port); | |
c675e06a | 742 | #ifdef CONFIG_IPVLAN_L3S |
4fbae7d8 MB |
743 | case IPVLAN_MODE_L3S: |
744 | return RX_HANDLER_PASS; | |
c675e06a | 745 | #endif |
2ad7bf36 MB |
746 | } |
747 | ||
748 | /* Should not reach here */ | |
749 | WARN_ONCE(true, "ipvlan_handle_frame() called for mode = [%hx]\n", | |
750 | port->mode); | |
751 | kfree_skb(skb); | |
a534dc52 | 752 | return RX_HANDLER_CONSUMED; |
2ad7bf36 | 753 | } |