Commit | Line | Data |
---|---|---|
ccb1352e | 1 | /* |
4572ef52 | 2 | * Copyright (c) 2007-2017 Nicira, Inc. |
ccb1352e JG |
3 | * |
4 | * This program is free software; you can redistribute it and/or | |
5 | * modify it under the terms of version 2 of the GNU General Public | |
6 | * License as published by the Free Software Foundation. | |
7 | * | |
8 | * This program is distributed in the hope that it will be useful, but | |
9 | * WITHOUT ANY WARRANTY; without even the implied warranty of | |
10 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
11 | * General Public License for more details. | |
12 | * | |
13 | * You should have received a copy of the GNU General Public License | |
14 | * along with this program; if not, write to the Free Software | |
15 | * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA | |
16 | * 02110-1301, USA | |
17 | */ | |
18 | ||
19 | #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt | |
20 | ||
21 | #include <linux/skbuff.h> | |
22 | #include <linux/in.h> | |
23 | #include <linux/ip.h> | |
24 | #include <linux/openvswitch.h> | |
7f8a436e | 25 | #include <linux/netfilter_ipv6.h> |
a175a723 | 26 | #include <linux/sctp.h> |
ccb1352e JG |
27 | #include <linux/tcp.h> |
28 | #include <linux/udp.h> | |
29 | #include <linux/in6.h> | |
30 | #include <linux/if_arp.h> | |
31 | #include <linux/if_vlan.h> | |
25cd9ba0 | 32 | |
7f8a436e | 33 | #include <net/dst.h> |
ccb1352e | 34 | #include <net/ip.h> |
3fdbd1ce | 35 | #include <net/ipv6.h> |
7b85b4df | 36 | #include <net/ip6_fib.h> |
ccb1352e JG |
37 | #include <net/checksum.h> |
38 | #include <net/dsfield.h> | |
25cd9ba0 | 39 | #include <net/mpls.h> |
a175a723 | 40 | #include <net/sctp/checksum.h> |
ccb1352e JG |
41 | |
42 | #include "datapath.h" | |
971427f3 | 43 | #include "flow.h" |
7f8a436e | 44 | #include "conntrack.h" |
ccb1352e JG |
45 | #include "vport.h" |
46 | ||
971427f3 AZ |
47 | struct deferred_action { |
48 | struct sk_buff *skb; | |
49 | const struct nlattr *actions; | |
47c697aa | 50 | int actions_len; |
971427f3 AZ |
51 | |
52 | /* Store pkt_key clone when creating deferred action. */ | |
53 | struct sw_flow_key pkt_key; | |
54 | }; | |
55 | ||
7f8a436e JS |
56 | #define MAX_L2_LEN (VLAN_ETH_HLEN + 3 * MPLS_HLEN) |
57 | struct ovs_frag_data { | |
58 | unsigned long dst; | |
59 | struct vport *vport; | |
60 | struct ovs_skb_cb cb; | |
61 | __be16 inner_protocol; | |
c66549ff JB |
62 | u16 network_offset; /* valid only for MPLS */ |
63 | u16 vlan_tci; | |
7f8a436e JS |
64 | __be16 vlan_proto; |
65 | unsigned int l2_len; | |
e2d9d835 | 66 | u8 mac_proto; |
7f8a436e JS |
67 | u8 l2_data[MAX_L2_LEN]; |
68 | }; | |
69 | ||
70 | static DEFINE_PER_CPU(struct ovs_frag_data, ovs_frag_data_storage); | |
71 | ||
971427f3 | 72 | #define DEFERRED_ACTION_FIFO_SIZE 10 |
2679d040 LR |
73 | #define OVS_RECURSION_LIMIT 5 |
74 | #define OVS_DEFERRED_ACTION_THRESHOLD (OVS_RECURSION_LIMIT - 2) | |
971427f3 AZ |
75 | struct action_fifo { |
76 | int head; | |
77 | int tail; | |
78 | /* Deferred action fifo queue storage. */ | |
79 | struct deferred_action fifo[DEFERRED_ACTION_FIFO_SIZE]; | |
80 | }; | |
81 | ||
4572ef52 | 82 | struct action_flow_keys { |
2679d040 LR |
83 | struct sw_flow_key key[OVS_DEFERRED_ACTION_THRESHOLD]; |
84 | }; | |
85 | ||
971427f3 | 86 | static struct action_fifo __percpu *action_fifos; |
4572ef52 | 87 | static struct action_flow_keys __percpu *flow_keys; |
971427f3 AZ |
88 | static DEFINE_PER_CPU(int, exec_actions_level); |
89 | ||
4572ef52 | 90 | /* Make a clone of the 'key', using the pre-allocated percpu 'flow_keys' |
91 | * space. Return NULL if out of key spaces. | |
92 | */ | |
93 | static struct sw_flow_key *clone_key(const struct sw_flow_key *key_) | |
94 | { | |
95 | struct action_flow_keys *keys = this_cpu_ptr(flow_keys); | |
96 | int level = this_cpu_read(exec_actions_level); | |
97 | struct sw_flow_key *key = NULL; | |
98 | ||
99 | if (level <= OVS_DEFERRED_ACTION_THRESHOLD) { | |
100 | key = &keys->key[level - 1]; | |
101 | *key = *key_; | |
102 | } | |
103 | ||
104 | return key; | |
105 | } | |
106 | ||
971427f3 AZ |
107 | static void action_fifo_init(struct action_fifo *fifo) |
108 | { | |
109 | fifo->head = 0; | |
110 | fifo->tail = 0; | |
111 | } | |
112 | ||
12eb18f7 | 113 | static bool action_fifo_is_empty(const struct action_fifo *fifo) |
971427f3 AZ |
114 | { |
115 | return (fifo->head == fifo->tail); | |
116 | } | |
117 | ||
118 | static struct deferred_action *action_fifo_get(struct action_fifo *fifo) | |
119 | { | |
120 | if (action_fifo_is_empty(fifo)) | |
121 | return NULL; | |
122 | ||
123 | return &fifo->fifo[fifo->tail++]; | |
124 | } | |
125 | ||
126 | static struct deferred_action *action_fifo_put(struct action_fifo *fifo) | |
127 | { | |
128 | if (fifo->head >= DEFERRED_ACTION_FIFO_SIZE - 1) | |
129 | return NULL; | |
130 | ||
131 | return &fifo->fifo[fifo->head++]; | |
132 | } | |
133 | ||
134 | /* Return true if fifo is not full */ | |
135 | static struct deferred_action *add_deferred_actions(struct sk_buff *skb, | |
47c697aa | 136 | const struct sw_flow_key *key, |
137 | const struct nlattr *actions, | |
138 | const int actions_len) | |
971427f3 AZ |
139 | { |
140 | struct action_fifo *fifo; | |
141 | struct deferred_action *da; | |
142 | ||
143 | fifo = this_cpu_ptr(action_fifos); | |
144 | da = action_fifo_put(fifo); | |
145 | if (da) { | |
146 | da->skb = skb; | |
47c697aa | 147 | da->actions = actions; |
148 | da->actions_len = actions_len; | |
971427f3 AZ |
149 | da->pkt_key = *key; |
150 | } | |
151 | ||
152 | return da; | |
153 | } | |
154 | ||
fff06c36 PS |
155 | static void invalidate_flow_key(struct sw_flow_key *key) |
156 | { | |
329f45bc | 157 | key->mac_proto |= SW_FLOW_KEY_INVALID; |
fff06c36 PS |
158 | } |
159 | ||
160 | static bool is_flow_key_valid(const struct sw_flow_key *key) | |
161 | { | |
329f45bc | 162 | return !(key->mac_proto & SW_FLOW_KEY_INVALID); |
fff06c36 PS |
163 | } |
164 | ||
bef7f756 | 165 | static int clone_execute(struct datapath *dp, struct sk_buff *skb, |
166 | struct sw_flow_key *key, | |
167 | u32 recirc_id, | |
168 | const struct nlattr *actions, int len, | |
169 | bool last, bool clone_flow_key); | |
170 | ||
bc7cc599 SH |
171 | static void update_ethertype(struct sk_buff *skb, struct ethhdr *hdr, |
172 | __be16 ethertype) | |
173 | { | |
174 | if (skb->ip_summed == CHECKSUM_COMPLETE) { | |
175 | __be16 diff[] = { ~(hdr->h_proto), ethertype }; | |
176 | ||
177 | skb->csum = ~csum_partial((char *)diff, sizeof(diff), | |
178 | ~skb->csum); | |
179 | } | |
180 | ||
181 | hdr->h_proto = ethertype; | |
182 | } | |
183 | ||
fff06c36 | 184 | static int push_mpls(struct sk_buff *skb, struct sw_flow_key *key, |
25cd9ba0 SH |
185 | const struct ovs_action_push_mpls *mpls) |
186 | { | |
85de4a21 | 187 | struct mpls_shim_hdr *new_mpls_lse; |
25cd9ba0 SH |
188 | |
189 | /* Networking stack do not allow simultaneous Tunnel and MPLS GSO. */ | |
190 | if (skb->encapsulation) | |
191 | return -ENOTSUPP; | |
192 | ||
193 | if (skb_cow_head(skb, MPLS_HLEN) < 0) | |
194 | return -ENOMEM; | |
195 | ||
48d2ab60 DA |
196 | if (!skb->inner_protocol) { |
197 | skb_set_inner_network_header(skb, skb->mac_len); | |
198 | skb_set_inner_protocol(skb, skb->protocol); | |
199 | } | |
200 | ||
25cd9ba0 SH |
201 | skb_push(skb, MPLS_HLEN); |
202 | memmove(skb_mac_header(skb) - MPLS_HLEN, skb_mac_header(skb), | |
203 | skb->mac_len); | |
204 | skb_reset_mac_header(skb); | |
48d2ab60 | 205 | skb_set_network_header(skb, skb->mac_len); |
25cd9ba0 | 206 | |
85de4a21 JB |
207 | new_mpls_lse = mpls_hdr(skb); |
208 | new_mpls_lse->label_stack_entry = mpls->mpls_lse; | |
25cd9ba0 | 209 | |
6b83d28a | 210 | skb_postpush_rcsum(skb, new_mpls_lse, MPLS_HLEN); |
25cd9ba0 | 211 | |
1560a074 JB |
212 | if (ovs_key_mac_proto(key) == MAC_PROTO_ETHERNET) |
213 | update_ethertype(skb, eth_hdr(skb), mpls->mpls_ethertype); | |
25cd9ba0 SH |
214 | skb->protocol = mpls->mpls_ethertype; |
215 | ||
fff06c36 | 216 | invalidate_flow_key(key); |
25cd9ba0 SH |
217 | return 0; |
218 | } | |
219 | ||
fff06c36 PS |
220 | static int pop_mpls(struct sk_buff *skb, struct sw_flow_key *key, |
221 | const __be16 ethertype) | |
25cd9ba0 | 222 | { |
25cd9ba0 SH |
223 | int err; |
224 | ||
e2195121 | 225 | err = skb_ensure_writable(skb, skb->mac_len + MPLS_HLEN); |
25cd9ba0 SH |
226 | if (unlikely(err)) |
227 | return err; | |
228 | ||
85de4a21 | 229 | skb_postpull_rcsum(skb, mpls_hdr(skb), MPLS_HLEN); |
25cd9ba0 SH |
230 | |
231 | memmove(skb_mac_header(skb) + MPLS_HLEN, skb_mac_header(skb), | |
232 | skb->mac_len); | |
233 | ||
234 | __skb_pull(skb, MPLS_HLEN); | |
235 | skb_reset_mac_header(skb); | |
48d2ab60 | 236 | skb_set_network_header(skb, skb->mac_len); |
25cd9ba0 | 237 | |
1560a074 JB |
238 | if (ovs_key_mac_proto(key) == MAC_PROTO_ETHERNET) { |
239 | struct ethhdr *hdr; | |
240 | ||
241 | /* mpls_hdr() is used to locate the ethertype field correctly in the | |
242 | * presence of VLAN tags. | |
243 | */ | |
244 | hdr = (struct ethhdr *)((void *)mpls_hdr(skb) - ETH_HLEN); | |
245 | update_ethertype(skb, hdr, ethertype); | |
246 | } | |
25cd9ba0 SH |
247 | if (eth_p_mpls(skb->protocol)) |
248 | skb->protocol = ethertype; | |
fff06c36 PS |
249 | |
250 | invalidate_flow_key(key); | |
25cd9ba0 SH |
251 | return 0; |
252 | } | |
253 | ||
83d2b9ba JR |
254 | static int set_mpls(struct sk_buff *skb, struct sw_flow_key *flow_key, |
255 | const __be32 *mpls_lse, const __be32 *mask) | |
25cd9ba0 | 256 | { |
85de4a21 | 257 | struct mpls_shim_hdr *stack; |
83d2b9ba | 258 | __be32 lse; |
25cd9ba0 SH |
259 | int err; |
260 | ||
e2195121 | 261 | err = skb_ensure_writable(skb, skb->mac_len + MPLS_HLEN); |
25cd9ba0 SH |
262 | if (unlikely(err)) |
263 | return err; | |
264 | ||
85de4a21 JB |
265 | stack = mpls_hdr(skb); |
266 | lse = OVS_MASKED(stack->label_stack_entry, *mpls_lse, *mask); | |
25cd9ba0 | 267 | if (skb->ip_summed == CHECKSUM_COMPLETE) { |
85de4a21 | 268 | __be32 diff[] = { ~(stack->label_stack_entry), lse }; |
83d2b9ba | 269 | |
25cd9ba0 SH |
270 | skb->csum = ~csum_partial((char *)diff, sizeof(diff), |
271 | ~skb->csum); | |
272 | } | |
273 | ||
85de4a21 | 274 | stack->label_stack_entry = lse; |
83d2b9ba | 275 | flow_key->mpls.top_lse = lse; |
25cd9ba0 SH |
276 | return 0; |
277 | } | |
278 | ||
fff06c36 | 279 | static int pop_vlan(struct sk_buff *skb, struct sw_flow_key *key) |
ccb1352e | 280 | { |
ccb1352e JG |
281 | int err; |
282 | ||
93515d53 | 283 | err = skb_vlan_pop(skb); |
018c1dda | 284 | if (skb_vlan_tag_present(skb)) { |
93515d53 | 285 | invalidate_flow_key(key); |
018c1dda EG |
286 | } else { |
287 | key->eth.vlan.tci = 0; | |
288 | key->eth.vlan.tpid = 0; | |
289 | } | |
93515d53 | 290 | return err; |
ccb1352e JG |
291 | } |
292 | ||
fff06c36 PS |
293 | static int push_vlan(struct sk_buff *skb, struct sw_flow_key *key, |
294 | const struct ovs_action_push_vlan *vlan) | |
ccb1352e | 295 | { |
018c1dda | 296 | if (skb_vlan_tag_present(skb)) { |
fff06c36 | 297 | invalidate_flow_key(key); |
018c1dda EG |
298 | } else { |
299 | key->eth.vlan.tci = vlan->vlan_tci; | |
300 | key->eth.vlan.tpid = vlan->vlan_tpid; | |
301 | } | |
93515d53 JP |
302 | return skb_vlan_push(skb, vlan->vlan_tpid, |
303 | ntohs(vlan->vlan_tci) & ~VLAN_TAG_PRESENT); | |
ccb1352e JG |
304 | } |
305 | ||
83d2b9ba JR |
306 | /* 'src' is already properly masked. */ |
307 | static void ether_addr_copy_masked(u8 *dst_, const u8 *src_, const u8 *mask_) | |
308 | { | |
309 | u16 *dst = (u16 *)dst_; | |
310 | const u16 *src = (const u16 *)src_; | |
311 | const u16 *mask = (const u16 *)mask_; | |
312 | ||
be26b9a8 JS |
313 | OVS_SET_MASKED(dst[0], src[0], mask[0]); |
314 | OVS_SET_MASKED(dst[1], src[1], mask[1]); | |
315 | OVS_SET_MASKED(dst[2], src[2], mask[2]); | |
83d2b9ba JR |
316 | } |
317 | ||
318 | static int set_eth_addr(struct sk_buff *skb, struct sw_flow_key *flow_key, | |
319 | const struct ovs_key_ethernet *key, | |
320 | const struct ovs_key_ethernet *mask) | |
ccb1352e JG |
321 | { |
322 | int err; | |
83d2b9ba | 323 | |
e2195121 | 324 | err = skb_ensure_writable(skb, ETH_HLEN); |
ccb1352e JG |
325 | if (unlikely(err)) |
326 | return err; | |
327 | ||
b34df5e8 PS |
328 | skb_postpull_rcsum(skb, eth_hdr(skb), ETH_ALEN * 2); |
329 | ||
83d2b9ba JR |
330 | ether_addr_copy_masked(eth_hdr(skb)->h_source, key->eth_src, |
331 | mask->eth_src); | |
332 | ether_addr_copy_masked(eth_hdr(skb)->h_dest, key->eth_dst, | |
333 | mask->eth_dst); | |
ccb1352e | 334 | |
6b83d28a | 335 | skb_postpush_rcsum(skb, eth_hdr(skb), ETH_ALEN * 2); |
b34df5e8 | 336 | |
83d2b9ba JR |
337 | ether_addr_copy(flow_key->eth.src, eth_hdr(skb)->h_source); |
338 | ether_addr_copy(flow_key->eth.dst, eth_hdr(skb)->h_dest); | |
ccb1352e JG |
339 | return 0; |
340 | } | |
341 | ||
91820da6 JB |
342 | /* pop_eth does not support VLAN packets as this action is never called |
343 | * for them. | |
344 | */ | |
345 | static int pop_eth(struct sk_buff *skb, struct sw_flow_key *key) | |
346 | { | |
347 | skb_pull_rcsum(skb, ETH_HLEN); | |
348 | skb_reset_mac_header(skb); | |
349 | skb_reset_mac_len(skb); | |
350 | ||
351 | /* safe right before invalidate_flow_key */ | |
352 | key->mac_proto = MAC_PROTO_NONE; | |
353 | invalidate_flow_key(key); | |
354 | return 0; | |
355 | } | |
356 | ||
357 | static int push_eth(struct sk_buff *skb, struct sw_flow_key *key, | |
358 | const struct ovs_action_push_eth *ethh) | |
359 | { | |
360 | struct ethhdr *hdr; | |
361 | ||
362 | /* Add the new Ethernet header */ | |
363 | if (skb_cow_head(skb, ETH_HLEN) < 0) | |
364 | return -ENOMEM; | |
365 | ||
366 | skb_push(skb, ETH_HLEN); | |
367 | skb_reset_mac_header(skb); | |
368 | skb_reset_mac_len(skb); | |
369 | ||
370 | hdr = eth_hdr(skb); | |
371 | ether_addr_copy(hdr->h_source, ethh->addresses.eth_src); | |
372 | ether_addr_copy(hdr->h_dest, ethh->addresses.eth_dst); | |
373 | hdr->h_proto = skb->protocol; | |
374 | ||
375 | skb_postpush_rcsum(skb, hdr, ETH_HLEN); | |
376 | ||
377 | /* safe right before invalidate_flow_key */ | |
378 | key->mac_proto = MAC_PROTO_ETHERNET; | |
379 | invalidate_flow_key(key); | |
380 | return 0; | |
381 | } | |
382 | ||
3576fd79 GG |
383 | static void update_ip_l4_checksum(struct sk_buff *skb, struct iphdr *nh, |
384 | __be32 addr, __be32 new_addr) | |
ccb1352e JG |
385 | { |
386 | int transport_len = skb->len - skb_transport_offset(skb); | |
387 | ||
3576fd79 GG |
388 | if (nh->frag_off & htons(IP_OFFSET)) |
389 | return; | |
390 | ||
ccb1352e JG |
391 | if (nh->protocol == IPPROTO_TCP) { |
392 | if (likely(transport_len >= sizeof(struct tcphdr))) | |
393 | inet_proto_csum_replace4(&tcp_hdr(skb)->check, skb, | |
4b048d6d | 394 | addr, new_addr, true); |
ccb1352e | 395 | } else if (nh->protocol == IPPROTO_UDP) { |
81e5d41d JG |
396 | if (likely(transport_len >= sizeof(struct udphdr))) { |
397 | struct udphdr *uh = udp_hdr(skb); | |
398 | ||
399 | if (uh->check || skb->ip_summed == CHECKSUM_PARTIAL) { | |
400 | inet_proto_csum_replace4(&uh->check, skb, | |
4b048d6d | 401 | addr, new_addr, true); |
81e5d41d JG |
402 | if (!uh->check) |
403 | uh->check = CSUM_MANGLED_0; | |
404 | } | |
405 | } | |
ccb1352e | 406 | } |
3576fd79 | 407 | } |
ccb1352e | 408 | |
3576fd79 GG |
409 | static void set_ip_addr(struct sk_buff *skb, struct iphdr *nh, |
410 | __be32 *addr, __be32 new_addr) | |
411 | { | |
412 | update_ip_l4_checksum(skb, nh, *addr, new_addr); | |
ccb1352e | 413 | csum_replace4(&nh->check, *addr, new_addr); |
7539fadc | 414 | skb_clear_hash(skb); |
ccb1352e JG |
415 | *addr = new_addr; |
416 | } | |
417 | ||
3fdbd1ce AA |
418 | static void update_ipv6_checksum(struct sk_buff *skb, u8 l4_proto, |
419 | __be32 addr[4], const __be32 new_addr[4]) | |
420 | { | |
421 | int transport_len = skb->len - skb_transport_offset(skb); | |
422 | ||
856447d0 | 423 | if (l4_proto == NEXTHDR_TCP) { |
3fdbd1ce AA |
424 | if (likely(transport_len >= sizeof(struct tcphdr))) |
425 | inet_proto_csum_replace16(&tcp_hdr(skb)->check, skb, | |
4b048d6d | 426 | addr, new_addr, true); |
856447d0 | 427 | } else if (l4_proto == NEXTHDR_UDP) { |
3fdbd1ce AA |
428 | if (likely(transport_len >= sizeof(struct udphdr))) { |
429 | struct udphdr *uh = udp_hdr(skb); | |
430 | ||
431 | if (uh->check || skb->ip_summed == CHECKSUM_PARTIAL) { | |
432 | inet_proto_csum_replace16(&uh->check, skb, | |
4b048d6d | 433 | addr, new_addr, true); |
3fdbd1ce AA |
434 | if (!uh->check) |
435 | uh->check = CSUM_MANGLED_0; | |
436 | } | |
437 | } | |
856447d0 JG |
438 | } else if (l4_proto == NEXTHDR_ICMP) { |
439 | if (likely(transport_len >= sizeof(struct icmp6hdr))) | |
440 | inet_proto_csum_replace16(&icmp6_hdr(skb)->icmp6_cksum, | |
4b048d6d | 441 | skb, addr, new_addr, true); |
3fdbd1ce AA |
442 | } |
443 | } | |
444 | ||
83d2b9ba JR |
445 | static void mask_ipv6_addr(const __be32 old[4], const __be32 addr[4], |
446 | const __be32 mask[4], __be32 masked[4]) | |
447 | { | |
be26b9a8 JS |
448 | masked[0] = OVS_MASKED(old[0], addr[0], mask[0]); |
449 | masked[1] = OVS_MASKED(old[1], addr[1], mask[1]); | |
450 | masked[2] = OVS_MASKED(old[2], addr[2], mask[2]); | |
451 | masked[3] = OVS_MASKED(old[3], addr[3], mask[3]); | |
83d2b9ba JR |
452 | } |
453 | ||
3fdbd1ce AA |
454 | static void set_ipv6_addr(struct sk_buff *skb, u8 l4_proto, |
455 | __be32 addr[4], const __be32 new_addr[4], | |
456 | bool recalculate_csum) | |
457 | { | |
458 | if (recalculate_csum) | |
459 | update_ipv6_checksum(skb, l4_proto, addr, new_addr); | |
460 | ||
7539fadc | 461 | skb_clear_hash(skb); |
3fdbd1ce AA |
462 | memcpy(addr, new_addr, sizeof(__be32[4])); |
463 | } | |
464 | ||
83d2b9ba | 465 | static void set_ipv6_fl(struct ipv6hdr *nh, u32 fl, u32 mask) |
3fdbd1ce | 466 | { |
83d2b9ba | 467 | /* Bits 21-24 are always unmasked, so this retains their values. */ |
be26b9a8 JS |
468 | OVS_SET_MASKED(nh->flow_lbl[0], (u8)(fl >> 16), (u8)(mask >> 16)); |
469 | OVS_SET_MASKED(nh->flow_lbl[1], (u8)(fl >> 8), (u8)(mask >> 8)); | |
470 | OVS_SET_MASKED(nh->flow_lbl[2], (u8)fl, (u8)mask); | |
3fdbd1ce AA |
471 | } |
472 | ||
83d2b9ba JR |
473 | static void set_ip_ttl(struct sk_buff *skb, struct iphdr *nh, u8 new_ttl, |
474 | u8 mask) | |
3fdbd1ce | 475 | { |
be26b9a8 | 476 | new_ttl = OVS_MASKED(nh->ttl, new_ttl, mask); |
3fdbd1ce | 477 | |
ccb1352e JG |
478 | csum_replace2(&nh->check, htons(nh->ttl << 8), htons(new_ttl << 8)); |
479 | nh->ttl = new_ttl; | |
480 | } | |
481 | ||
83d2b9ba JR |
482 | static int set_ipv4(struct sk_buff *skb, struct sw_flow_key *flow_key, |
483 | const struct ovs_key_ipv4 *key, | |
484 | const struct ovs_key_ipv4 *mask) | |
ccb1352e JG |
485 | { |
486 | struct iphdr *nh; | |
83d2b9ba | 487 | __be32 new_addr; |
ccb1352e JG |
488 | int err; |
489 | ||
e2195121 JP |
490 | err = skb_ensure_writable(skb, skb_network_offset(skb) + |
491 | sizeof(struct iphdr)); | |
ccb1352e JG |
492 | if (unlikely(err)) |
493 | return err; | |
494 | ||
495 | nh = ip_hdr(skb); | |
496 | ||
83d2b9ba JR |
497 | /* Setting an IP addresses is typically only a side effect of |
498 | * matching on them in the current userspace implementation, so it | |
499 | * makes sense to check if the value actually changed. | |
500 | */ | |
501 | if (mask->ipv4_src) { | |
be26b9a8 | 502 | new_addr = OVS_MASKED(nh->saddr, key->ipv4_src, mask->ipv4_src); |
ccb1352e | 503 | |
83d2b9ba JR |
504 | if (unlikely(new_addr != nh->saddr)) { |
505 | set_ip_addr(skb, nh, &nh->saddr, new_addr); | |
506 | flow_key->ipv4.addr.src = new_addr; | |
507 | } | |
fff06c36 | 508 | } |
83d2b9ba | 509 | if (mask->ipv4_dst) { |
be26b9a8 | 510 | new_addr = OVS_MASKED(nh->daddr, key->ipv4_dst, mask->ipv4_dst); |
ccb1352e | 511 | |
83d2b9ba JR |
512 | if (unlikely(new_addr != nh->daddr)) { |
513 | set_ip_addr(skb, nh, &nh->daddr, new_addr); | |
514 | flow_key->ipv4.addr.dst = new_addr; | |
515 | } | |
fff06c36 | 516 | } |
83d2b9ba JR |
517 | if (mask->ipv4_tos) { |
518 | ipv4_change_dsfield(nh, ~mask->ipv4_tos, key->ipv4_tos); | |
519 | flow_key->ip.tos = nh->tos; | |
520 | } | |
521 | if (mask->ipv4_ttl) { | |
522 | set_ip_ttl(skb, nh, key->ipv4_ttl, mask->ipv4_ttl); | |
523 | flow_key->ip.ttl = nh->ttl; | |
fff06c36 | 524 | } |
ccb1352e JG |
525 | |
526 | return 0; | |
527 | } | |
528 | ||
83d2b9ba JR |
529 | static bool is_ipv6_mask_nonzero(const __be32 addr[4]) |
530 | { | |
531 | return !!(addr[0] | addr[1] | addr[2] | addr[3]); | |
532 | } | |
533 | ||
534 | static int set_ipv6(struct sk_buff *skb, struct sw_flow_key *flow_key, | |
535 | const struct ovs_key_ipv6 *key, | |
536 | const struct ovs_key_ipv6 *mask) | |
3fdbd1ce AA |
537 | { |
538 | struct ipv6hdr *nh; | |
539 | int err; | |
3fdbd1ce | 540 | |
e2195121 JP |
541 | err = skb_ensure_writable(skb, skb_network_offset(skb) + |
542 | sizeof(struct ipv6hdr)); | |
3fdbd1ce AA |
543 | if (unlikely(err)) |
544 | return err; | |
545 | ||
546 | nh = ipv6_hdr(skb); | |
3fdbd1ce | 547 | |
83d2b9ba JR |
548 | /* Setting an IP addresses is typically only a side effect of |
549 | * matching on them in the current userspace implementation, so it | |
550 | * makes sense to check if the value actually changed. | |
551 | */ | |
552 | if (is_ipv6_mask_nonzero(mask->ipv6_src)) { | |
553 | __be32 *saddr = (__be32 *)&nh->saddr; | |
554 | __be32 masked[4]; | |
555 | ||
556 | mask_ipv6_addr(saddr, key->ipv6_src, mask->ipv6_src, masked); | |
557 | ||
558 | if (unlikely(memcmp(saddr, masked, sizeof(masked)))) { | |
b4f70527 | 559 | set_ipv6_addr(skb, flow_key->ip.proto, saddr, masked, |
83d2b9ba JR |
560 | true); |
561 | memcpy(&flow_key->ipv6.addr.src, masked, | |
562 | sizeof(flow_key->ipv6.addr.src)); | |
563 | } | |
564 | } | |
565 | if (is_ipv6_mask_nonzero(mask->ipv6_dst)) { | |
3fdbd1ce AA |
566 | unsigned int offset = 0; |
567 | int flags = IP6_FH_F_SKIP_RH; | |
568 | bool recalc_csum = true; | |
83d2b9ba JR |
569 | __be32 *daddr = (__be32 *)&nh->daddr; |
570 | __be32 masked[4]; | |
571 | ||
572 | mask_ipv6_addr(daddr, key->ipv6_dst, mask->ipv6_dst, masked); | |
573 | ||
574 | if (unlikely(memcmp(daddr, masked, sizeof(masked)))) { | |
575 | if (ipv6_ext_hdr(nh->nexthdr)) | |
576 | recalc_csum = (ipv6_find_hdr(skb, &offset, | |
577 | NEXTHDR_ROUTING, | |
578 | NULL, &flags) | |
579 | != NEXTHDR_ROUTING); | |
580 | ||
b4f70527 | 581 | set_ipv6_addr(skb, flow_key->ip.proto, daddr, masked, |
83d2b9ba JR |
582 | recalc_csum); |
583 | memcpy(&flow_key->ipv6.addr.dst, masked, | |
584 | sizeof(flow_key->ipv6.addr.dst)); | |
585 | } | |
586 | } | |
587 | if (mask->ipv6_tclass) { | |
588 | ipv6_change_dsfield(nh, ~mask->ipv6_tclass, key->ipv6_tclass); | |
589 | flow_key->ip.tos = ipv6_get_dsfield(nh); | |
590 | } | |
591 | if (mask->ipv6_label) { | |
592 | set_ipv6_fl(nh, ntohl(key->ipv6_label), | |
593 | ntohl(mask->ipv6_label)); | |
594 | flow_key->ipv6.label = | |
595 | *(__be32 *)nh & htonl(IPV6_FLOWINFO_FLOWLABEL); | |
596 | } | |
597 | if (mask->ipv6_hlimit) { | |
be26b9a8 JS |
598 | OVS_SET_MASKED(nh->hop_limit, key->ipv6_hlimit, |
599 | mask->ipv6_hlimit); | |
83d2b9ba | 600 | flow_key->ip.ttl = nh->hop_limit; |
3fdbd1ce | 601 | } |
3fdbd1ce AA |
602 | return 0; |
603 | } | |
604 | ||
e2195121 | 605 | /* Must follow skb_ensure_writable() since that can move the skb data. */ |
ccb1352e | 606 | static void set_tp_port(struct sk_buff *skb, __be16 *port, |
83d2b9ba | 607 | __be16 new_port, __sum16 *check) |
ccb1352e | 608 | { |
4b048d6d | 609 | inet_proto_csum_replace2(check, skb, *port, new_port, false); |
ccb1352e | 610 | *port = new_port; |
81e5d41d JG |
611 | } |
612 | ||
83d2b9ba JR |
613 | static int set_udp(struct sk_buff *skb, struct sw_flow_key *flow_key, |
614 | const struct ovs_key_udp *key, | |
615 | const struct ovs_key_udp *mask) | |
ccb1352e JG |
616 | { |
617 | struct udphdr *uh; | |
83d2b9ba | 618 | __be16 src, dst; |
ccb1352e JG |
619 | int err; |
620 | ||
e2195121 JP |
621 | err = skb_ensure_writable(skb, skb_transport_offset(skb) + |
622 | sizeof(struct udphdr)); | |
ccb1352e JG |
623 | if (unlikely(err)) |
624 | return err; | |
625 | ||
626 | uh = udp_hdr(skb); | |
83d2b9ba | 627 | /* Either of the masks is non-zero, so do not bother checking them. */ |
be26b9a8 JS |
628 | src = OVS_MASKED(uh->source, key->udp_src, mask->udp_src); |
629 | dst = OVS_MASKED(uh->dest, key->udp_dst, mask->udp_dst); | |
ccb1352e | 630 | |
83d2b9ba JR |
631 | if (uh->check && skb->ip_summed != CHECKSUM_PARTIAL) { |
632 | if (likely(src != uh->source)) { | |
633 | set_tp_port(skb, &uh->source, src, &uh->check); | |
634 | flow_key->tp.src = src; | |
635 | } | |
636 | if (likely(dst != uh->dest)) { | |
637 | set_tp_port(skb, &uh->dest, dst, &uh->check); | |
638 | flow_key->tp.dst = dst; | |
639 | } | |
640 | ||
641 | if (unlikely(!uh->check)) | |
642 | uh->check = CSUM_MANGLED_0; | |
643 | } else { | |
644 | uh->source = src; | |
645 | uh->dest = dst; | |
646 | flow_key->tp.src = src; | |
647 | flow_key->tp.dst = dst; | |
fff06c36 | 648 | } |
ccb1352e | 649 | |
83d2b9ba JR |
650 | skb_clear_hash(skb); |
651 | ||
ccb1352e JG |
652 | return 0; |
653 | } | |
654 | ||
83d2b9ba JR |
655 | static int set_tcp(struct sk_buff *skb, struct sw_flow_key *flow_key, |
656 | const struct ovs_key_tcp *key, | |
657 | const struct ovs_key_tcp *mask) | |
ccb1352e JG |
658 | { |
659 | struct tcphdr *th; | |
83d2b9ba | 660 | __be16 src, dst; |
ccb1352e JG |
661 | int err; |
662 | ||
e2195121 JP |
663 | err = skb_ensure_writable(skb, skb_transport_offset(skb) + |
664 | sizeof(struct tcphdr)); | |
ccb1352e JG |
665 | if (unlikely(err)) |
666 | return err; | |
667 | ||
668 | th = tcp_hdr(skb); | |
be26b9a8 | 669 | src = OVS_MASKED(th->source, key->tcp_src, mask->tcp_src); |
83d2b9ba JR |
670 | if (likely(src != th->source)) { |
671 | set_tp_port(skb, &th->source, src, &th->check); | |
672 | flow_key->tp.src = src; | |
fff06c36 | 673 | } |
be26b9a8 | 674 | dst = OVS_MASKED(th->dest, key->tcp_dst, mask->tcp_dst); |
83d2b9ba JR |
675 | if (likely(dst != th->dest)) { |
676 | set_tp_port(skb, &th->dest, dst, &th->check); | |
677 | flow_key->tp.dst = dst; | |
fff06c36 | 678 | } |
83d2b9ba | 679 | skb_clear_hash(skb); |
ccb1352e JG |
680 | |
681 | return 0; | |
682 | } | |
683 | ||
83d2b9ba JR |
684 | static int set_sctp(struct sk_buff *skb, struct sw_flow_key *flow_key, |
685 | const struct ovs_key_sctp *key, | |
686 | const struct ovs_key_sctp *mask) | |
a175a723 | 687 | { |
83d2b9ba | 688 | unsigned int sctphoff = skb_transport_offset(skb); |
a175a723 | 689 | struct sctphdr *sh; |
83d2b9ba | 690 | __le32 old_correct_csum, new_csum, old_csum; |
a175a723 | 691 | int err; |
a175a723 | 692 | |
e2195121 | 693 | err = skb_ensure_writable(skb, sctphoff + sizeof(struct sctphdr)); |
a175a723 JS |
694 | if (unlikely(err)) |
695 | return err; | |
696 | ||
697 | sh = sctp_hdr(skb); | |
83d2b9ba JR |
698 | old_csum = sh->checksum; |
699 | old_correct_csum = sctp_compute_cksum(skb, sctphoff); | |
a175a723 | 700 | |
be26b9a8 JS |
701 | sh->source = OVS_MASKED(sh->source, key->sctp_src, mask->sctp_src); |
702 | sh->dest = OVS_MASKED(sh->dest, key->sctp_dst, mask->sctp_dst); | |
a175a723 | 703 | |
83d2b9ba | 704 | new_csum = sctp_compute_cksum(skb, sctphoff); |
a175a723 | 705 | |
83d2b9ba JR |
706 | /* Carry any checksum errors through. */ |
707 | sh->checksum = old_csum ^ old_correct_csum ^ new_csum; | |
a175a723 | 708 | |
83d2b9ba JR |
709 | skb_clear_hash(skb); |
710 | flow_key->tp.src = sh->source; | |
711 | flow_key->tp.dst = sh->dest; | |
a175a723 JS |
712 | |
713 | return 0; | |
714 | } | |
715 | ||
188515fb | 716 | static int ovs_vport_output(struct net *net, struct sock *sk, struct sk_buff *skb) |
7f8a436e JS |
717 | { |
718 | struct ovs_frag_data *data = this_cpu_ptr(&ovs_frag_data_storage); | |
719 | struct vport *vport = data->vport; | |
720 | ||
721 | if (skb_cow_head(skb, data->l2_len) < 0) { | |
722 | kfree_skb(skb); | |
723 | return -ENOMEM; | |
724 | } | |
725 | ||
726 | __skb_dst_copy(skb, data->dst); | |
727 | *OVS_CB(skb) = data->cb; | |
728 | skb->inner_protocol = data->inner_protocol; | |
729 | skb->vlan_tci = data->vlan_tci; | |
730 | skb->vlan_proto = data->vlan_proto; | |
731 | ||
732 | /* Reconstruct the MAC header. */ | |
733 | skb_push(skb, data->l2_len); | |
734 | memcpy(skb->data, &data->l2_data, data->l2_len); | |
6b83d28a | 735 | skb_postpush_rcsum(skb, skb->data, data->l2_len); |
7f8a436e JS |
736 | skb_reset_mac_header(skb); |
737 | ||
c66549ff JB |
738 | if (eth_p_mpls(skb->protocol)) { |
739 | skb->inner_network_header = skb->network_header; | |
740 | skb_set_network_header(skb, data->network_offset); | |
741 | skb_reset_mac_len(skb); | |
742 | } | |
743 | ||
e2d9d835 | 744 | ovs_vport_send(vport, skb, data->mac_proto); |
7f8a436e JS |
745 | return 0; |
746 | } | |
747 | ||
748 | static unsigned int | |
749 | ovs_dst_get_mtu(const struct dst_entry *dst) | |
750 | { | |
751 | return dst->dev->mtu; | |
752 | } | |
753 | ||
754 | static struct dst_ops ovs_dst_ops = { | |
755 | .family = AF_UNSPEC, | |
756 | .mtu = ovs_dst_get_mtu, | |
757 | }; | |
758 | ||
759 | /* prepare_frag() is called once per (larger-than-MTU) frame; its inverse is | |
760 | * ovs_vport_output(), which is called once per fragmented packet. | |
761 | */ | |
c66549ff | 762 | static void prepare_frag(struct vport *vport, struct sk_buff *skb, |
e2d9d835 | 763 | u16 orig_network_offset, u8 mac_proto) |
7f8a436e JS |
764 | { |
765 | unsigned int hlen = skb_network_offset(skb); | |
766 | struct ovs_frag_data *data; | |
767 | ||
768 | data = this_cpu_ptr(&ovs_frag_data_storage); | |
769 | data->dst = skb->_skb_refdst; | |
770 | data->vport = vport; | |
771 | data->cb = *OVS_CB(skb); | |
772 | data->inner_protocol = skb->inner_protocol; | |
c66549ff | 773 | data->network_offset = orig_network_offset; |
7f8a436e JS |
774 | data->vlan_tci = skb->vlan_tci; |
775 | data->vlan_proto = skb->vlan_proto; | |
e2d9d835 | 776 | data->mac_proto = mac_proto; |
7f8a436e JS |
777 | data->l2_len = hlen; |
778 | memcpy(&data->l2_data, skb->data, hlen); | |
779 | ||
780 | memset(IPCB(skb), 0, sizeof(struct inet_skb_parm)); | |
781 | skb_pull(skb, hlen); | |
782 | } | |
783 | ||
c559cd3a | 784 | static void ovs_fragment(struct net *net, struct vport *vport, |
e2d9d835 JB |
785 | struct sk_buff *skb, u16 mru, |
786 | struct sw_flow_key *key) | |
7f8a436e | 787 | { |
c66549ff JB |
788 | u16 orig_network_offset = 0; |
789 | ||
790 | if (eth_p_mpls(skb->protocol)) { | |
791 | orig_network_offset = skb_network_offset(skb); | |
792 | skb->network_header = skb->inner_network_header; | |
793 | } | |
794 | ||
7f8a436e JS |
795 | if (skb_network_offset(skb) > MAX_L2_LEN) { |
796 | OVS_NLERR(1, "L2 header too long to fragment"); | |
b8f22570 | 797 | goto err; |
7f8a436e JS |
798 | } |
799 | ||
e2d9d835 | 800 | if (key->eth.type == htons(ETH_P_IP)) { |
7f8a436e JS |
801 | struct dst_entry ovs_dst; |
802 | unsigned long orig_dst; | |
803 | ||
e2d9d835 JB |
804 | prepare_frag(vport, skb, orig_network_offset, |
805 | ovs_key_mac_proto(key)); | |
7f8a436e JS |
806 | dst_init(&ovs_dst, &ovs_dst_ops, NULL, 1, |
807 | DST_OBSOLETE_NONE, DST_NOCOUNT); | |
808 | ovs_dst.dev = vport->dev; | |
809 | ||
810 | orig_dst = skb->_skb_refdst; | |
811 | skb_dst_set_noref(skb, &ovs_dst); | |
812 | IPCB(skb)->frag_max_size = mru; | |
813 | ||
694869b3 | 814 | ip_do_fragment(net, skb->sk, skb, ovs_vport_output); |
7f8a436e | 815 | refdst_drop(orig_dst); |
e2d9d835 | 816 | } else if (key->eth.type == htons(ETH_P_IPV6)) { |
7f8a436e JS |
817 | const struct nf_ipv6_ops *v6ops = nf_get_ipv6_ops(); |
818 | unsigned long orig_dst; | |
819 | struct rt6_info ovs_rt; | |
820 | ||
f1304f7b | 821 | if (!v6ops) |
b8f22570 | 822 | goto err; |
7f8a436e | 823 | |
e2d9d835 JB |
824 | prepare_frag(vport, skb, orig_network_offset, |
825 | ovs_key_mac_proto(key)); | |
7f8a436e JS |
826 | memset(&ovs_rt, 0, sizeof(ovs_rt)); |
827 | dst_init(&ovs_rt.dst, &ovs_dst_ops, NULL, 1, | |
828 | DST_OBSOLETE_NONE, DST_NOCOUNT); | |
829 | ovs_rt.dst.dev = vport->dev; | |
830 | ||
831 | orig_dst = skb->_skb_refdst; | |
832 | skb_dst_set_noref(skb, &ovs_rt.dst); | |
833 | IP6CB(skb)->frag_max_size = mru; | |
834 | ||
7d8c6e39 | 835 | v6ops->fragment(net, skb->sk, skb, ovs_vport_output); |
7f8a436e JS |
836 | refdst_drop(orig_dst); |
837 | } else { | |
838 | WARN_ONCE(1, "Failed fragment ->%s: eth=%04x, MRU=%d, MTU=%d.", | |
e2d9d835 | 839 | ovs_vport_name(vport), ntohs(key->eth.type), mru, |
7f8a436e | 840 | vport->dev->mtu); |
b8f22570 | 841 | goto err; |
7f8a436e | 842 | } |
b8f22570 JS |
843 | |
844 | return; | |
845 | err: | |
846 | kfree_skb(skb); | |
7f8a436e JS |
847 | } |
848 | ||
849 | static void do_output(struct datapath *dp, struct sk_buff *skb, int out_port, | |
850 | struct sw_flow_key *key) | |
ccb1352e | 851 | { |
738967b8 | 852 | struct vport *vport = ovs_vport_rcu(dp, out_port); |
ccb1352e | 853 | |
7f8a436e JS |
854 | if (likely(vport)) { |
855 | u16 mru = OVS_CB(skb)->mru; | |
f2a4d086 WT |
856 | u32 cutlen = OVS_CB(skb)->cutlen; |
857 | ||
858 | if (unlikely(cutlen > 0)) { | |
e2d9d835 | 859 | if (skb->len - cutlen > ovs_mac_header_len(key)) |
f2a4d086 WT |
860 | pskb_trim(skb, skb->len - cutlen); |
861 | else | |
e2d9d835 | 862 | pskb_trim(skb, ovs_mac_header_len(key)); |
f2a4d086 | 863 | } |
7f8a436e | 864 | |
738314a0 JB |
865 | if (likely(!mru || |
866 | (skb->len <= mru + vport->dev->hard_header_len))) { | |
e2d9d835 | 867 | ovs_vport_send(vport, skb, ovs_key_mac_proto(key)); |
7f8a436e | 868 | } else if (mru <= vport->dev->mtu) { |
c559cd3a | 869 | struct net *net = read_pnet(&dp->net); |
7f8a436e | 870 | |
e2d9d835 | 871 | ovs_fragment(net, vport, skb, mru, key); |
7f8a436e JS |
872 | } else { |
873 | kfree_skb(skb); | |
874 | } | |
875 | } else { | |
ccb1352e | 876 | kfree_skb(skb); |
7f8a436e | 877 | } |
ccb1352e JG |
878 | } |
879 | ||
880 | static int output_userspace(struct datapath *dp, struct sk_buff *skb, | |
ccea7445 | 881 | struct sw_flow_key *key, const struct nlattr *attr, |
f2a4d086 WT |
882 | const struct nlattr *actions, int actions_len, |
883 | uint32_t cutlen) | |
ccb1352e JG |
884 | { |
885 | struct dp_upcall_info upcall; | |
886 | const struct nlattr *a; | |
887 | int rem; | |
888 | ||
ccea7445 | 889 | memset(&upcall, 0, sizeof(upcall)); |
ccb1352e | 890 | upcall.cmd = OVS_PACKET_CMD_ACTION; |
7f8a436e | 891 | upcall.mru = OVS_CB(skb)->mru; |
ccb1352e JG |
892 | |
893 | for (a = nla_data(attr), rem = nla_len(attr); rem > 0; | |
894 | a = nla_next(a, &rem)) { | |
895 | switch (nla_type(a)) { | |
896 | case OVS_USERSPACE_ATTR_USERDATA: | |
897 | upcall.userdata = a; | |
898 | break; | |
899 | ||
900 | case OVS_USERSPACE_ATTR_PID: | |
15e47304 | 901 | upcall.portid = nla_get_u32(a); |
ccb1352e | 902 | break; |
8f0aad6f WZ |
903 | |
904 | case OVS_USERSPACE_ATTR_EGRESS_TUN_PORT: { | |
905 | /* Get out tunnel info. */ | |
906 | struct vport *vport; | |
907 | ||
908 | vport = ovs_vport_rcu(dp, nla_get_u32(a)); | |
909 | if (vport) { | |
910 | int err; | |
911 | ||
fc4099f1 PS |
912 | err = dev_fill_metadata_dst(vport->dev, skb); |
913 | if (!err) | |
914 | upcall.egress_tun_info = skb_tunnel_info(skb); | |
8f0aad6f | 915 | } |
4c222798 | 916 | |
8f0aad6f | 917 | break; |
ccb1352e | 918 | } |
8f0aad6f | 919 | |
ccea7445 NM |
920 | case OVS_USERSPACE_ATTR_ACTIONS: { |
921 | /* Include actions. */ | |
922 | upcall.actions = actions; | |
923 | upcall.actions_len = actions_len; | |
924 | break; | |
925 | } | |
926 | ||
8f0aad6f | 927 | } /* End of switch. */ |
ccb1352e JG |
928 | } |
929 | ||
f2a4d086 | 930 | return ovs_dp_upcall(dp, skb, key, &upcall, cutlen); |
ccb1352e JG |
931 | } |
932 | ||
798c1661 | 933 | /* When 'last' is true, sample() should always consume the 'skb'. |
934 | * Otherwise, sample() should keep 'skb' intact regardless what | |
935 | * actions are executed within sample(). | |
936 | */ | |
ccb1352e | 937 | static int sample(struct datapath *dp, struct sk_buff *skb, |
ccea7445 | 938 | struct sw_flow_key *key, const struct nlattr *attr, |
798c1661 | 939 | bool last) |
ccb1352e | 940 | { |
798c1661 | 941 | struct nlattr *actions; |
942 | struct nlattr *sample_arg; | |
798c1661 | 943 | int rem = nla_len(attr); |
798c1661 | 944 | const struct sample_arg *arg; |
bef7f756 | 945 | bool clone_flow_key; |
ccb1352e | 946 | |
798c1661 | 947 | /* The first action is always 'OVS_SAMPLE_ATTR_ARG'. */ |
948 | sample_arg = nla_data(attr); | |
949 | arg = nla_data(sample_arg); | |
950 | actions = nla_next(sample_arg, &rem); | |
e05176a3 | 951 | |
798c1661 | 952 | if ((arg->probability != U32_MAX) && |
953 | (!arg->probability || prandom_u32() > arg->probability)) { | |
954 | if (last) | |
955 | consume_skb(skb); | |
956 | return 0; | |
ccb1352e JG |
957 | } |
958 | ||
bef7f756 | 959 | clone_flow_key = !arg->exec; |
960 | return clone_execute(dp, skb, key, 0, actions, rem, last, | |
961 | clone_flow_key); | |
971427f3 AZ |
962 | } |
963 | ||
964 | static void execute_hash(struct sk_buff *skb, struct sw_flow_key *key, | |
965 | const struct nlattr *attr) | |
966 | { | |
967 | struct ovs_action_hash *hash_act = nla_data(attr); | |
968 | u32 hash = 0; | |
969 | ||
970 | /* OVS_HASH_ALG_L4 is the only possible hash algorithm. */ | |
971 | hash = skb_get_hash(skb); | |
972 | hash = jhash_1word(hash, hash_act->hash_basis); | |
973 | if (!hash) | |
974 | hash = 0x1; | |
975 | ||
976 | key->ovs_flow_hash = hash; | |
ccb1352e JG |
977 | } |
978 | ||
83d2b9ba JR |
979 | static int execute_set_action(struct sk_buff *skb, |
980 | struct sw_flow_key *flow_key, | |
981 | const struct nlattr *a) | |
982 | { | |
983 | /* Only tunnel set execution is supported without a mask. */ | |
984 | if (nla_type(a) == OVS_KEY_ATTR_TUNNEL_INFO) { | |
34ae932a TG |
985 | struct ovs_tunnel_info *tun = nla_data(a); |
986 | ||
987 | skb_dst_drop(skb); | |
988 | dst_hold((struct dst_entry *)tun->tun_dst); | |
989 | skb_dst_set(skb, (struct dst_entry *)tun->tun_dst); | |
83d2b9ba JR |
990 | return 0; |
991 | } | |
992 | ||
993 | return -EINVAL; | |
994 | } | |
995 | ||
996 | /* Mask is at the midpoint of the data. */ | |
997 | #define get_mask(a, type) ((const type)nla_data(a) + 1) | |
998 | ||
999 | static int execute_masked_set_action(struct sk_buff *skb, | |
1000 | struct sw_flow_key *flow_key, | |
1001 | const struct nlattr *a) | |
ccb1352e JG |
1002 | { |
1003 | int err = 0; | |
1004 | ||
83d2b9ba | 1005 | switch (nla_type(a)) { |
ccb1352e | 1006 | case OVS_KEY_ATTR_PRIORITY: |
be26b9a8 JS |
1007 | OVS_SET_MASKED(skb->priority, nla_get_u32(a), |
1008 | *get_mask(a, u32 *)); | |
83d2b9ba | 1009 | flow_key->phy.priority = skb->priority; |
ccb1352e JG |
1010 | break; |
1011 | ||
39c7caeb | 1012 | case OVS_KEY_ATTR_SKB_MARK: |
be26b9a8 | 1013 | OVS_SET_MASKED(skb->mark, nla_get_u32(a), *get_mask(a, u32 *)); |
83d2b9ba | 1014 | flow_key->phy.skb_mark = skb->mark; |
39c7caeb AA |
1015 | break; |
1016 | ||
f0b128c1 | 1017 | case OVS_KEY_ATTR_TUNNEL_INFO: |
83d2b9ba JR |
1018 | /* Masked data not supported for tunnel. */ |
1019 | err = -EINVAL; | |
7d5437c7 PS |
1020 | break; |
1021 | ||
ccb1352e | 1022 | case OVS_KEY_ATTR_ETHERNET: |
83d2b9ba JR |
1023 | err = set_eth_addr(skb, flow_key, nla_data(a), |
1024 | get_mask(a, struct ovs_key_ethernet *)); | |
ccb1352e JG |
1025 | break; |
1026 | ||
1027 | case OVS_KEY_ATTR_IPV4: | |
83d2b9ba JR |
1028 | err = set_ipv4(skb, flow_key, nla_data(a), |
1029 | get_mask(a, struct ovs_key_ipv4 *)); | |
ccb1352e JG |
1030 | break; |
1031 | ||
3fdbd1ce | 1032 | case OVS_KEY_ATTR_IPV6: |
83d2b9ba JR |
1033 | err = set_ipv6(skb, flow_key, nla_data(a), |
1034 | get_mask(a, struct ovs_key_ipv6 *)); | |
3fdbd1ce AA |
1035 | break; |
1036 | ||
ccb1352e | 1037 | case OVS_KEY_ATTR_TCP: |
83d2b9ba JR |
1038 | err = set_tcp(skb, flow_key, nla_data(a), |
1039 | get_mask(a, struct ovs_key_tcp *)); | |
ccb1352e JG |
1040 | break; |
1041 | ||
1042 | case OVS_KEY_ATTR_UDP: | |
83d2b9ba JR |
1043 | err = set_udp(skb, flow_key, nla_data(a), |
1044 | get_mask(a, struct ovs_key_udp *)); | |
ccb1352e | 1045 | break; |
a175a723 JS |
1046 | |
1047 | case OVS_KEY_ATTR_SCTP: | |
83d2b9ba JR |
1048 | err = set_sctp(skb, flow_key, nla_data(a), |
1049 | get_mask(a, struct ovs_key_sctp *)); | |
a175a723 | 1050 | break; |
25cd9ba0 SH |
1051 | |
1052 | case OVS_KEY_ATTR_MPLS: | |
83d2b9ba JR |
1053 | err = set_mpls(skb, flow_key, nla_data(a), get_mask(a, |
1054 | __be32 *)); | |
25cd9ba0 | 1055 | break; |
7f8a436e JS |
1056 | |
1057 | case OVS_KEY_ATTR_CT_STATE: | |
1058 | case OVS_KEY_ATTR_CT_ZONE: | |
182e3042 | 1059 | case OVS_KEY_ATTR_CT_MARK: |
33db4125 | 1060 | case OVS_KEY_ATTR_CT_LABELS: |
9dd7f890 JR |
1061 | case OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV4: |
1062 | case OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6: | |
7f8a436e JS |
1063 | err = -EINVAL; |
1064 | break; | |
ccb1352e JG |
1065 | } |
1066 | ||
1067 | return err; | |
1068 | } | |
1069 | ||
971427f3 AZ |
1070 | static int execute_recirc(struct datapath *dp, struct sk_buff *skb, |
1071 | struct sw_flow_key *key, | |
bef7f756 | 1072 | const struct nlattr *a, bool last) |
971427f3 | 1073 | { |
bef7f756 | 1074 | u32 recirc_id; |
971427f3 | 1075 | |
fff06c36 PS |
1076 | if (!is_flow_key_valid(key)) { |
1077 | int err; | |
1078 | ||
1079 | err = ovs_flow_key_update(skb, key); | |
1080 | if (err) | |
1081 | return err; | |
1082 | } | |
1083 | BUG_ON(!is_flow_key_valid(key)); | |
971427f3 | 1084 | |
bef7f756 | 1085 | recirc_id = nla_get_u32(a); |
1086 | return clone_execute(dp, skb, key, recirc_id, NULL, 0, last, true); | |
971427f3 AZ |
1087 | } |
1088 | ||
ccb1352e JG |
1089 | /* Execute a list of actions against 'skb'. */ |
1090 | static int do_execute_actions(struct datapath *dp, struct sk_buff *skb, | |
2ff3e4e4 | 1091 | struct sw_flow_key *key, |
651887b0 | 1092 | const struct nlattr *attr, int len) |
ccb1352e | 1093 | { |
ccb1352e JG |
1094 | const struct nlattr *a; |
1095 | int rem; | |
1096 | ||
1097 | for (a = attr, rem = len; rem > 0; | |
1098 | a = nla_next(a, &rem)) { | |
1099 | int err = 0; | |
1100 | ||
5b8784aa | 1101 | switch (nla_type(a)) { |
1102 | case OVS_ACTION_ATTR_OUTPUT: { | |
1103 | int port = nla_get_u32(a); | |
1104 | struct sk_buff *clone; | |
1105 | ||
1106 | /* Every output action needs a separate clone | |
1107 | * of 'skb', In case the output action is the | |
1108 | * last action, cloning can be avoided. | |
1109 | */ | |
1110 | if (nla_is_last(a, rem)) { | |
1111 | do_output(dp, skb, port, key); | |
1112 | /* 'skb' has been used for output. | |
1113 | */ | |
1114 | return 0; | |
1115 | } | |
738967b8 | 1116 | |
5b8784aa | 1117 | clone = skb_clone(skb, GFP_ATOMIC); |
1118 | if (clone) | |
1119 | do_output(dp, clone, port, key); | |
f2a4d086 | 1120 | OVS_CB(skb)->cutlen = 0; |
ccb1352e | 1121 | break; |
5b8784aa | 1122 | } |
ccb1352e | 1123 | |
f2a4d086 WT |
1124 | case OVS_ACTION_ATTR_TRUNC: { |
1125 | struct ovs_action_trunc *trunc = nla_data(a); | |
1126 | ||
1127 | if (skb->len > trunc->max_len) | |
1128 | OVS_CB(skb)->cutlen = skb->len - trunc->max_len; | |
1129 | break; | |
1130 | } | |
1131 | ||
ccb1352e | 1132 | case OVS_ACTION_ATTR_USERSPACE: |
f2a4d086 WT |
1133 | output_userspace(dp, skb, key, a, attr, |
1134 | len, OVS_CB(skb)->cutlen); | |
1135 | OVS_CB(skb)->cutlen = 0; | |
ccb1352e JG |
1136 | break; |
1137 | ||
971427f3 AZ |
1138 | case OVS_ACTION_ATTR_HASH: |
1139 | execute_hash(skb, key, a); | |
1140 | break; | |
1141 | ||
25cd9ba0 | 1142 | case OVS_ACTION_ATTR_PUSH_MPLS: |
fff06c36 | 1143 | err = push_mpls(skb, key, nla_data(a)); |
25cd9ba0 SH |
1144 | break; |
1145 | ||
1146 | case OVS_ACTION_ATTR_POP_MPLS: | |
fff06c36 | 1147 | err = pop_mpls(skb, key, nla_get_be16(a)); |
25cd9ba0 SH |
1148 | break; |
1149 | ||
ccb1352e | 1150 | case OVS_ACTION_ATTR_PUSH_VLAN: |
fff06c36 | 1151 | err = push_vlan(skb, key, nla_data(a)); |
ccb1352e JG |
1152 | break; |
1153 | ||
1154 | case OVS_ACTION_ATTR_POP_VLAN: | |
fff06c36 | 1155 | err = pop_vlan(skb, key); |
ccb1352e JG |
1156 | break; |
1157 | ||
bef7f756 | 1158 | case OVS_ACTION_ATTR_RECIRC: { |
1159 | bool last = nla_is_last(a, rem); | |
1160 | ||
1161 | err = execute_recirc(dp, skb, key, a, last); | |
1162 | if (last) { | |
971427f3 AZ |
1163 | /* If this is the last action, the skb has |
1164 | * been consumed or freed. | |
1165 | * Return immediately. | |
1166 | */ | |
1167 | return err; | |
1168 | } | |
1169 | break; | |
bef7f756 | 1170 | } |
971427f3 | 1171 | |
ccb1352e | 1172 | case OVS_ACTION_ATTR_SET: |
fff06c36 | 1173 | err = execute_set_action(skb, key, nla_data(a)); |
ccb1352e JG |
1174 | break; |
1175 | ||
83d2b9ba JR |
1176 | case OVS_ACTION_ATTR_SET_MASKED: |
1177 | case OVS_ACTION_ATTR_SET_TO_MASKED: | |
1178 | err = execute_masked_set_action(skb, key, nla_data(a)); | |
1179 | break; | |
1180 | ||
798c1661 | 1181 | case OVS_ACTION_ATTR_SAMPLE: { |
1182 | bool last = nla_is_last(a, rem); | |
1183 | ||
1184 | err = sample(dp, skb, key, a, last); | |
1185 | if (last) | |
1186 | return err; | |
1187 | ||
ccb1352e | 1188 | break; |
798c1661 | 1189 | } |
7f8a436e JS |
1190 | |
1191 | case OVS_ACTION_ATTR_CT: | |
ec0d043d JS |
1192 | if (!is_flow_key_valid(key)) { |
1193 | err = ovs_flow_key_update(skb, key); | |
1194 | if (err) | |
1195 | return err; | |
1196 | } | |
1197 | ||
7f8a436e JS |
1198 | err = ovs_ct_execute(ovs_dp_get_net(dp), skb, key, |
1199 | nla_data(a)); | |
1200 | ||
1201 | /* Hide stolen IP fragments from user space. */ | |
74c16618 JS |
1202 | if (err) |
1203 | return err == -EINPROGRESS ? 0 : err; | |
7f8a436e | 1204 | break; |
91820da6 | 1205 | |
b8226962 EG |
1206 | case OVS_ACTION_ATTR_CT_CLEAR: |
1207 | err = ovs_ct_clear(skb, key); | |
1208 | break; | |
1209 | ||
91820da6 JB |
1210 | case OVS_ACTION_ATTR_PUSH_ETH: |
1211 | err = push_eth(skb, key, nla_data(a)); | |
1212 | break; | |
1213 | ||
1214 | case OVS_ACTION_ATTR_POP_ETH: | |
1215 | err = pop_eth(skb, key); | |
1216 | break; | |
ccb1352e JG |
1217 | } |
1218 | ||
1219 | if (unlikely(err)) { | |
1220 | kfree_skb(skb); | |
1221 | return err; | |
1222 | } | |
1223 | } | |
1224 | ||
5b8784aa | 1225 | consume_skb(skb); |
ccb1352e JG |
1226 | return 0; |
1227 | } | |
1228 | ||
bef7f756 | 1229 | /* Execute the actions on the clone of the packet. The effect of the |
1230 | * execution does not affect the original 'skb' nor the original 'key'. | |
1231 | * | |
1232 | * The execution may be deferred in case the actions can not be executed | |
1233 | * immediately. | |
1234 | */ | |
1235 | static int clone_execute(struct datapath *dp, struct sk_buff *skb, | |
1236 | struct sw_flow_key *key, u32 recirc_id, | |
1237 | const struct nlattr *actions, int len, | |
1238 | bool last, bool clone_flow_key) | |
1239 | { | |
1240 | struct deferred_action *da; | |
1241 | struct sw_flow_key *clone; | |
1242 | ||
1243 | skb = last ? skb : skb_clone(skb, GFP_ATOMIC); | |
1244 | if (!skb) { | |
1245 | /* Out of memory, skip this action. | |
1246 | */ | |
1247 | return 0; | |
1248 | } | |
1249 | ||
1250 | /* When clone_flow_key is false, the 'key' will not be change | |
1251 | * by the actions, then the 'key' can be used directly. | |
1252 | * Otherwise, try to clone key from the next recursion level of | |
1253 | * 'flow_keys'. If clone is successful, execute the actions | |
1254 | * without deferring. | |
1255 | */ | |
1256 | clone = clone_flow_key ? clone_key(key) : key; | |
1257 | if (clone) { | |
1258 | int err = 0; | |
1259 | ||
1260 | if (actions) { /* Sample action */ | |
1261 | if (clone_flow_key) | |
1262 | __this_cpu_inc(exec_actions_level); | |
1263 | ||
1264 | err = do_execute_actions(dp, skb, clone, | |
1265 | actions, len); | |
1266 | ||
1267 | if (clone_flow_key) | |
1268 | __this_cpu_dec(exec_actions_level); | |
1269 | } else { /* Recirc action */ | |
1270 | clone->recirc_id = recirc_id; | |
1271 | ovs_dp_process_packet(skb, clone); | |
1272 | } | |
1273 | return err; | |
1274 | } | |
1275 | ||
1276 | /* Out of 'flow_keys' space. Defer actions */ | |
1277 | da = add_deferred_actions(skb, key, actions, len); | |
1278 | if (da) { | |
1279 | if (!actions) { /* Recirc action */ | |
1280 | key = &da->pkt_key; | |
1281 | key->recirc_id = recirc_id; | |
1282 | } | |
1283 | } else { | |
1284 | /* Out of per CPU action FIFO space. Drop the 'skb' and | |
1285 | * log an error. | |
1286 | */ | |
1287 | kfree_skb(skb); | |
1288 | ||
1289 | if (net_ratelimit()) { | |
1290 | if (actions) { /* Sample action */ | |
1291 | pr_warn("%s: deferred action limit reached, drop sample action\n", | |
1292 | ovs_dp_name(dp)); | |
1293 | } else { /* Recirc action */ | |
1294 | pr_warn("%s: deferred action limit reached, drop recirc action\n", | |
1295 | ovs_dp_name(dp)); | |
1296 | } | |
1297 | } | |
1298 | } | |
1299 | return 0; | |
1300 | } | |
1301 | ||
971427f3 AZ |
1302 | static void process_deferred_actions(struct datapath *dp) |
1303 | { | |
1304 | struct action_fifo *fifo = this_cpu_ptr(action_fifos); | |
1305 | ||
1306 | /* Do not touch the FIFO in case there is no deferred actions. */ | |
1307 | if (action_fifo_is_empty(fifo)) | |
1308 | return; | |
1309 | ||
1310 | /* Finishing executing all deferred actions. */ | |
1311 | do { | |
1312 | struct deferred_action *da = action_fifo_get(fifo); | |
1313 | struct sk_buff *skb = da->skb; | |
1314 | struct sw_flow_key *key = &da->pkt_key; | |
1315 | const struct nlattr *actions = da->actions; | |
47c697aa | 1316 | int actions_len = da->actions_len; |
971427f3 AZ |
1317 | |
1318 | if (actions) | |
47c697aa | 1319 | do_execute_actions(dp, skb, key, actions, actions_len); |
971427f3 AZ |
1320 | else |
1321 | ovs_dp_process_packet(skb, key); | |
1322 | } while (!action_fifo_is_empty(fifo)); | |
1323 | ||
1324 | /* Reset FIFO for the next packet. */ | |
1325 | action_fifo_init(fifo); | |
1326 | } | |
1327 | ||
ccb1352e | 1328 | /* Execute a list of actions against 'skb'. */ |
2ff3e4e4 | 1329 | int ovs_execute_actions(struct datapath *dp, struct sk_buff *skb, |
12eb18f7 TG |
1330 | const struct sw_flow_actions *acts, |
1331 | struct sw_flow_key *key) | |
ccb1352e | 1332 | { |
b064d0d8 HFS |
1333 | int err, level; |
1334 | ||
1335 | level = __this_cpu_inc_return(exec_actions_level); | |
2679d040 | 1336 | if (unlikely(level > OVS_RECURSION_LIMIT)) { |
b064d0d8 HFS |
1337 | net_crit_ratelimited("ovs: recursion limit reached on datapath %s, probable configuration error\n", |
1338 | ovs_dp_name(dp)); | |
1339 | kfree_skb(skb); | |
1340 | err = -ENETDOWN; | |
1341 | goto out; | |
1342 | } | |
971427f3 | 1343 | |
494bea39 | 1344 | OVS_CB(skb)->acts_origlen = acts->orig_len; |
971427f3 AZ |
1345 | err = do_execute_actions(dp, skb, key, |
1346 | acts->actions, acts->actions_len); | |
1347 | ||
b064d0d8 | 1348 | if (level == 1) |
971427f3 AZ |
1349 | process_deferred_actions(dp); |
1350 | ||
b064d0d8 HFS |
1351 | out: |
1352 | __this_cpu_dec(exec_actions_level); | |
971427f3 AZ |
1353 | return err; |
1354 | } | |
1355 | ||
1356 | int action_fifos_init(void) | |
1357 | { | |
1358 | action_fifos = alloc_percpu(struct action_fifo); | |
1359 | if (!action_fifos) | |
1360 | return -ENOMEM; | |
ccb1352e | 1361 | |
4572ef52 | 1362 | flow_keys = alloc_percpu(struct action_flow_keys); |
1363 | if (!flow_keys) { | |
2679d040 LR |
1364 | free_percpu(action_fifos); |
1365 | return -ENOMEM; | |
1366 | } | |
1367 | ||
971427f3 AZ |
1368 | return 0; |
1369 | } | |
1370 | ||
1371 | void action_fifos_exit(void) | |
1372 | { | |
1373 | free_percpu(action_fifos); | |
4572ef52 | 1374 | free_percpu(flow_keys); |
ccb1352e | 1375 | } |