]>
Commit | Line | Data |
---|---|---|
ccb1352e | 1 | /* |
971427f3 | 2 | * Copyright (c) 2007-2014 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> | |
a175a723 | 25 | #include <linux/sctp.h> |
ccb1352e JG |
26 | #include <linux/tcp.h> |
27 | #include <linux/udp.h> | |
28 | #include <linux/in6.h> | |
29 | #include <linux/if_arp.h> | |
30 | #include <linux/if_vlan.h> | |
25cd9ba0 | 31 | |
ccb1352e | 32 | #include <net/ip.h> |
3fdbd1ce | 33 | #include <net/ipv6.h> |
ccb1352e JG |
34 | #include <net/checksum.h> |
35 | #include <net/dsfield.h> | |
25cd9ba0 | 36 | #include <net/mpls.h> |
a175a723 | 37 | #include <net/sctp/checksum.h> |
ccb1352e JG |
38 | |
39 | #include "datapath.h" | |
971427f3 | 40 | #include "flow.h" |
ccb1352e JG |
41 | #include "vport.h" |
42 | ||
43 | static int do_execute_actions(struct datapath *dp, struct sk_buff *skb, | |
2ff3e4e4 | 44 | struct sw_flow_key *key, |
651887b0 | 45 | const struct nlattr *attr, int len); |
ccb1352e | 46 | |
971427f3 AZ |
47 | struct deferred_action { |
48 | struct sk_buff *skb; | |
49 | const struct nlattr *actions; | |
50 | ||
51 | /* Store pkt_key clone when creating deferred action. */ | |
52 | struct sw_flow_key pkt_key; | |
53 | }; | |
54 | ||
55 | #define DEFERRED_ACTION_FIFO_SIZE 10 | |
56 | struct action_fifo { | |
57 | int head; | |
58 | int tail; | |
59 | /* Deferred action fifo queue storage. */ | |
60 | struct deferred_action fifo[DEFERRED_ACTION_FIFO_SIZE]; | |
61 | }; | |
62 | ||
63 | static struct action_fifo __percpu *action_fifos; | |
64 | static DEFINE_PER_CPU(int, exec_actions_level); | |
65 | ||
66 | static void action_fifo_init(struct action_fifo *fifo) | |
67 | { | |
68 | fifo->head = 0; | |
69 | fifo->tail = 0; | |
70 | } | |
71 | ||
72 | static bool action_fifo_is_empty(struct action_fifo *fifo) | |
73 | { | |
74 | return (fifo->head == fifo->tail); | |
75 | } | |
76 | ||
77 | static struct deferred_action *action_fifo_get(struct action_fifo *fifo) | |
78 | { | |
79 | if (action_fifo_is_empty(fifo)) | |
80 | return NULL; | |
81 | ||
82 | return &fifo->fifo[fifo->tail++]; | |
83 | } | |
84 | ||
85 | static struct deferred_action *action_fifo_put(struct action_fifo *fifo) | |
86 | { | |
87 | if (fifo->head >= DEFERRED_ACTION_FIFO_SIZE - 1) | |
88 | return NULL; | |
89 | ||
90 | return &fifo->fifo[fifo->head++]; | |
91 | } | |
92 | ||
93 | /* Return true if fifo is not full */ | |
94 | static struct deferred_action *add_deferred_actions(struct sk_buff *skb, | |
95 | struct sw_flow_key *key, | |
96 | const struct nlattr *attr) | |
97 | { | |
98 | struct action_fifo *fifo; | |
99 | struct deferred_action *da; | |
100 | ||
101 | fifo = this_cpu_ptr(action_fifos); | |
102 | da = action_fifo_put(fifo); | |
103 | if (da) { | |
104 | da->skb = skb; | |
105 | da->actions = attr; | |
106 | da->pkt_key = *key; | |
107 | } | |
108 | ||
109 | return da; | |
110 | } | |
111 | ||
ccb1352e JG |
112 | static int make_writable(struct sk_buff *skb, int write_len) |
113 | { | |
2ba5af42 JB |
114 | if (!pskb_may_pull(skb, write_len)) |
115 | return -ENOMEM; | |
116 | ||
ccb1352e JG |
117 | if (!skb_cloned(skb) || skb_clone_writable(skb, write_len)) |
118 | return 0; | |
119 | ||
120 | return pskb_expand_head(skb, 0, 0, GFP_ATOMIC); | |
121 | } | |
122 | ||
25cd9ba0 SH |
123 | static int push_mpls(struct sk_buff *skb, |
124 | const struct ovs_action_push_mpls *mpls) | |
125 | { | |
126 | __be32 *new_mpls_lse; | |
127 | struct ethhdr *hdr; | |
128 | ||
129 | /* Networking stack do not allow simultaneous Tunnel and MPLS GSO. */ | |
130 | if (skb->encapsulation) | |
131 | return -ENOTSUPP; | |
132 | ||
133 | if (skb_cow_head(skb, MPLS_HLEN) < 0) | |
134 | return -ENOMEM; | |
135 | ||
136 | skb_push(skb, MPLS_HLEN); | |
137 | memmove(skb_mac_header(skb) - MPLS_HLEN, skb_mac_header(skb), | |
138 | skb->mac_len); | |
139 | skb_reset_mac_header(skb); | |
140 | ||
141 | new_mpls_lse = (__be32 *)skb_mpls_header(skb); | |
142 | *new_mpls_lse = mpls->mpls_lse; | |
143 | ||
144 | if (skb->ip_summed == CHECKSUM_COMPLETE) | |
145 | skb->csum = csum_add(skb->csum, csum_partial(new_mpls_lse, | |
146 | MPLS_HLEN, 0)); | |
147 | ||
148 | hdr = eth_hdr(skb); | |
149 | hdr->h_proto = mpls->mpls_ethertype; | |
150 | ||
151 | skb_set_inner_protocol(skb, skb->protocol); | |
152 | skb->protocol = mpls->mpls_ethertype; | |
153 | ||
154 | return 0; | |
155 | } | |
156 | ||
157 | static int pop_mpls(struct sk_buff *skb, const __be16 ethertype) | |
158 | { | |
159 | struct ethhdr *hdr; | |
160 | int err; | |
161 | ||
162 | err = make_writable(skb, skb->mac_len + MPLS_HLEN); | |
163 | if (unlikely(err)) | |
164 | return err; | |
165 | ||
166 | if (skb->ip_summed == CHECKSUM_COMPLETE) | |
167 | skb->csum = csum_sub(skb->csum, | |
168 | csum_partial(skb_mpls_header(skb), | |
169 | MPLS_HLEN, 0)); | |
170 | ||
171 | memmove(skb_mac_header(skb) + MPLS_HLEN, skb_mac_header(skb), | |
172 | skb->mac_len); | |
173 | ||
174 | __skb_pull(skb, MPLS_HLEN); | |
175 | skb_reset_mac_header(skb); | |
176 | ||
177 | /* skb_mpls_header() is used to locate the ethertype | |
178 | * field correctly in the presence of VLAN tags. | |
179 | */ | |
180 | hdr = (struct ethhdr *)(skb_mpls_header(skb) - ETH_HLEN); | |
181 | hdr->h_proto = ethertype; | |
182 | if (eth_p_mpls(skb->protocol)) | |
183 | skb->protocol = ethertype; | |
184 | return 0; | |
185 | } | |
186 | ||
187 | static int set_mpls(struct sk_buff *skb, const __be32 *mpls_lse) | |
188 | { | |
189 | __be32 *stack; | |
190 | int err; | |
191 | ||
192 | err = make_writable(skb, skb->mac_len + MPLS_HLEN); | |
193 | if (unlikely(err)) | |
194 | return err; | |
195 | ||
196 | stack = (__be32 *)skb_mpls_header(skb); | |
197 | if (skb->ip_summed == CHECKSUM_COMPLETE) { | |
198 | __be32 diff[] = { ~(*stack), *mpls_lse }; | |
199 | ||
200 | skb->csum = ~csum_partial((char *)diff, sizeof(diff), | |
201 | ~skb->csum); | |
202 | } | |
203 | ||
204 | *stack = *mpls_lse; | |
205 | ||
206 | return 0; | |
207 | } | |
208 | ||
39855b5b | 209 | /* remove VLAN header from packet and update csum accordingly. */ |
ccb1352e JG |
210 | static int __pop_vlan_tci(struct sk_buff *skb, __be16 *current_tci) |
211 | { | |
212 | struct vlan_hdr *vhdr; | |
213 | int err; | |
214 | ||
215 | err = make_writable(skb, VLAN_ETH_HLEN); | |
216 | if (unlikely(err)) | |
217 | return err; | |
218 | ||
219 | if (skb->ip_summed == CHECKSUM_COMPLETE) | |
220 | skb->csum = csum_sub(skb->csum, csum_partial(skb->data | |
7b024082 | 221 | + (2 * ETH_ALEN), VLAN_HLEN, 0)); |
ccb1352e JG |
222 | |
223 | vhdr = (struct vlan_hdr *)(skb->data + ETH_HLEN); | |
224 | *current_tci = vhdr->h_vlan_TCI; | |
225 | ||
226 | memmove(skb->data + VLAN_HLEN, skb->data, 2 * ETH_ALEN); | |
227 | __skb_pull(skb, VLAN_HLEN); | |
228 | ||
229 | vlan_set_encap_proto(skb, vhdr); | |
230 | skb->mac_header += VLAN_HLEN; | |
25cd9ba0 | 231 | |
2ba5af42 JB |
232 | if (skb_network_offset(skb) < ETH_HLEN) |
233 | skb_set_network_header(skb, ETH_HLEN); | |
ccb1352e | 234 | |
25cd9ba0 SH |
235 | /* Update mac_len for subsequent MPLS actions */ |
236 | skb_reset_mac_len(skb); | |
ccb1352e JG |
237 | return 0; |
238 | } | |
239 | ||
240 | static int pop_vlan(struct sk_buff *skb) | |
241 | { | |
242 | __be16 tci; | |
243 | int err; | |
244 | ||
245 | if (likely(vlan_tx_tag_present(skb))) { | |
246 | skb->vlan_tci = 0; | |
247 | } else { | |
248 | if (unlikely(skb->protocol != htons(ETH_P_8021Q) || | |
249 | skb->len < VLAN_ETH_HLEN)) | |
250 | return 0; | |
251 | ||
252 | err = __pop_vlan_tci(skb, &tci); | |
253 | if (err) | |
254 | return err; | |
255 | } | |
256 | /* move next vlan tag to hw accel tag */ | |
257 | if (likely(skb->protocol != htons(ETH_P_8021Q) || | |
258 | skb->len < VLAN_ETH_HLEN)) | |
259 | return 0; | |
260 | ||
261 | err = __pop_vlan_tci(skb, &tci); | |
262 | if (unlikely(err)) | |
263 | return err; | |
264 | ||
86a9bad3 | 265 | __vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), ntohs(tci)); |
ccb1352e JG |
266 | return 0; |
267 | } | |
268 | ||
269 | static int push_vlan(struct sk_buff *skb, const struct ovs_action_push_vlan *vlan) | |
270 | { | |
271 | if (unlikely(vlan_tx_tag_present(skb))) { | |
272 | u16 current_tag; | |
273 | ||
274 | /* push down current VLAN tag */ | |
275 | current_tag = vlan_tx_tag_get(skb); | |
276 | ||
86a9bad3 | 277 | if (!__vlan_put_tag(skb, skb->vlan_proto, current_tag)) |
ccb1352e | 278 | return -ENOMEM; |
25cd9ba0 SH |
279 | /* Update mac_len for subsequent MPLS actions */ |
280 | skb->mac_len += VLAN_HLEN; | |
ccb1352e JG |
281 | |
282 | if (skb->ip_summed == CHECKSUM_COMPLETE) | |
283 | skb->csum = csum_add(skb->csum, csum_partial(skb->data | |
7b024082 | 284 | + (2 * ETH_ALEN), VLAN_HLEN, 0)); |
ccb1352e JG |
285 | |
286 | } | |
86a9bad3 | 287 | __vlan_hwaccel_put_tag(skb, vlan->vlan_tpid, ntohs(vlan->vlan_tci) & ~VLAN_TAG_PRESENT); |
ccb1352e JG |
288 | return 0; |
289 | } | |
290 | ||
291 | static int set_eth_addr(struct sk_buff *skb, | |
292 | const struct ovs_key_ethernet *eth_key) | |
293 | { | |
294 | int err; | |
295 | err = make_writable(skb, ETH_HLEN); | |
296 | if (unlikely(err)) | |
297 | return err; | |
298 | ||
b34df5e8 PS |
299 | skb_postpull_rcsum(skb, eth_hdr(skb), ETH_ALEN * 2); |
300 | ||
8c63ff09 JP |
301 | ether_addr_copy(eth_hdr(skb)->h_source, eth_key->eth_src); |
302 | ether_addr_copy(eth_hdr(skb)->h_dest, eth_key->eth_dst); | |
ccb1352e | 303 | |
b34df5e8 PS |
304 | ovs_skb_postpush_rcsum(skb, eth_hdr(skb), ETH_ALEN * 2); |
305 | ||
ccb1352e JG |
306 | return 0; |
307 | } | |
308 | ||
309 | static void set_ip_addr(struct sk_buff *skb, struct iphdr *nh, | |
310 | __be32 *addr, __be32 new_addr) | |
311 | { | |
312 | int transport_len = skb->len - skb_transport_offset(skb); | |
313 | ||
314 | if (nh->protocol == IPPROTO_TCP) { | |
315 | if (likely(transport_len >= sizeof(struct tcphdr))) | |
316 | inet_proto_csum_replace4(&tcp_hdr(skb)->check, skb, | |
317 | *addr, new_addr, 1); | |
318 | } else if (nh->protocol == IPPROTO_UDP) { | |
81e5d41d JG |
319 | if (likely(transport_len >= sizeof(struct udphdr))) { |
320 | struct udphdr *uh = udp_hdr(skb); | |
321 | ||
322 | if (uh->check || skb->ip_summed == CHECKSUM_PARTIAL) { | |
323 | inet_proto_csum_replace4(&uh->check, skb, | |
324 | *addr, new_addr, 1); | |
325 | if (!uh->check) | |
326 | uh->check = CSUM_MANGLED_0; | |
327 | } | |
328 | } | |
ccb1352e JG |
329 | } |
330 | ||
331 | csum_replace4(&nh->check, *addr, new_addr); | |
7539fadc | 332 | skb_clear_hash(skb); |
ccb1352e JG |
333 | *addr = new_addr; |
334 | } | |
335 | ||
3fdbd1ce AA |
336 | static void update_ipv6_checksum(struct sk_buff *skb, u8 l4_proto, |
337 | __be32 addr[4], const __be32 new_addr[4]) | |
338 | { | |
339 | int transport_len = skb->len - skb_transport_offset(skb); | |
340 | ||
341 | if (l4_proto == IPPROTO_TCP) { | |
342 | if (likely(transport_len >= sizeof(struct tcphdr))) | |
343 | inet_proto_csum_replace16(&tcp_hdr(skb)->check, skb, | |
344 | addr, new_addr, 1); | |
345 | } else if (l4_proto == IPPROTO_UDP) { | |
346 | if (likely(transport_len >= sizeof(struct udphdr))) { | |
347 | struct udphdr *uh = udp_hdr(skb); | |
348 | ||
349 | if (uh->check || skb->ip_summed == CHECKSUM_PARTIAL) { | |
350 | inet_proto_csum_replace16(&uh->check, skb, | |
351 | addr, new_addr, 1); | |
352 | if (!uh->check) | |
353 | uh->check = CSUM_MANGLED_0; | |
354 | } | |
355 | } | |
356 | } | |
357 | } | |
358 | ||
359 | static void set_ipv6_addr(struct sk_buff *skb, u8 l4_proto, | |
360 | __be32 addr[4], const __be32 new_addr[4], | |
361 | bool recalculate_csum) | |
362 | { | |
363 | if (recalculate_csum) | |
364 | update_ipv6_checksum(skb, l4_proto, addr, new_addr); | |
365 | ||
7539fadc | 366 | skb_clear_hash(skb); |
3fdbd1ce AA |
367 | memcpy(addr, new_addr, sizeof(__be32[4])); |
368 | } | |
369 | ||
370 | static void set_ipv6_tc(struct ipv6hdr *nh, u8 tc) | |
371 | { | |
372 | nh->priority = tc >> 4; | |
373 | nh->flow_lbl[0] = (nh->flow_lbl[0] & 0x0F) | ((tc & 0x0F) << 4); | |
374 | } | |
375 | ||
376 | static void set_ipv6_fl(struct ipv6hdr *nh, u32 fl) | |
377 | { | |
378 | nh->flow_lbl[0] = (nh->flow_lbl[0] & 0xF0) | (fl & 0x000F0000) >> 16; | |
379 | nh->flow_lbl[1] = (fl & 0x0000FF00) >> 8; | |
380 | nh->flow_lbl[2] = fl & 0x000000FF; | |
381 | } | |
382 | ||
ccb1352e JG |
383 | static void set_ip_ttl(struct sk_buff *skb, struct iphdr *nh, u8 new_ttl) |
384 | { | |
385 | csum_replace2(&nh->check, htons(nh->ttl << 8), htons(new_ttl << 8)); | |
386 | nh->ttl = new_ttl; | |
387 | } | |
388 | ||
389 | static int set_ipv4(struct sk_buff *skb, const struct ovs_key_ipv4 *ipv4_key) | |
390 | { | |
391 | struct iphdr *nh; | |
392 | int err; | |
393 | ||
394 | err = make_writable(skb, skb_network_offset(skb) + | |
395 | sizeof(struct iphdr)); | |
396 | if (unlikely(err)) | |
397 | return err; | |
398 | ||
399 | nh = ip_hdr(skb); | |
400 | ||
401 | if (ipv4_key->ipv4_src != nh->saddr) | |
402 | set_ip_addr(skb, nh, &nh->saddr, ipv4_key->ipv4_src); | |
403 | ||
404 | if (ipv4_key->ipv4_dst != nh->daddr) | |
405 | set_ip_addr(skb, nh, &nh->daddr, ipv4_key->ipv4_dst); | |
406 | ||
407 | if (ipv4_key->ipv4_tos != nh->tos) | |
408 | ipv4_change_dsfield(nh, 0, ipv4_key->ipv4_tos); | |
409 | ||
410 | if (ipv4_key->ipv4_ttl != nh->ttl) | |
411 | set_ip_ttl(skb, nh, ipv4_key->ipv4_ttl); | |
412 | ||
413 | return 0; | |
414 | } | |
415 | ||
3fdbd1ce AA |
416 | static int set_ipv6(struct sk_buff *skb, const struct ovs_key_ipv6 *ipv6_key) |
417 | { | |
418 | struct ipv6hdr *nh; | |
419 | int err; | |
420 | __be32 *saddr; | |
421 | __be32 *daddr; | |
422 | ||
423 | err = make_writable(skb, skb_network_offset(skb) + | |
424 | sizeof(struct ipv6hdr)); | |
425 | if (unlikely(err)) | |
426 | return err; | |
427 | ||
428 | nh = ipv6_hdr(skb); | |
429 | saddr = (__be32 *)&nh->saddr; | |
430 | daddr = (__be32 *)&nh->daddr; | |
431 | ||
432 | if (memcmp(ipv6_key->ipv6_src, saddr, sizeof(ipv6_key->ipv6_src))) | |
433 | set_ipv6_addr(skb, ipv6_key->ipv6_proto, saddr, | |
434 | ipv6_key->ipv6_src, true); | |
435 | ||
436 | if (memcmp(ipv6_key->ipv6_dst, daddr, sizeof(ipv6_key->ipv6_dst))) { | |
437 | unsigned int offset = 0; | |
438 | int flags = IP6_FH_F_SKIP_RH; | |
439 | bool recalc_csum = true; | |
440 | ||
441 | if (ipv6_ext_hdr(nh->nexthdr)) | |
442 | recalc_csum = ipv6_find_hdr(skb, &offset, | |
443 | NEXTHDR_ROUTING, NULL, | |
444 | &flags) != NEXTHDR_ROUTING; | |
445 | ||
446 | set_ipv6_addr(skb, ipv6_key->ipv6_proto, daddr, | |
447 | ipv6_key->ipv6_dst, recalc_csum); | |
448 | } | |
449 | ||
450 | set_ipv6_tc(nh, ipv6_key->ipv6_tclass); | |
451 | set_ipv6_fl(nh, ntohl(ipv6_key->ipv6_label)); | |
452 | nh->hop_limit = ipv6_key->ipv6_hlimit; | |
453 | ||
454 | return 0; | |
455 | } | |
456 | ||
ccb1352e JG |
457 | /* Must follow make_writable() since that can move the skb data. */ |
458 | static void set_tp_port(struct sk_buff *skb, __be16 *port, | |
459 | __be16 new_port, __sum16 *check) | |
460 | { | |
461 | inet_proto_csum_replace2(check, skb, *port, new_port, 0); | |
462 | *port = new_port; | |
7539fadc | 463 | skb_clear_hash(skb); |
ccb1352e JG |
464 | } |
465 | ||
81e5d41d JG |
466 | static void set_udp_port(struct sk_buff *skb, __be16 *port, __be16 new_port) |
467 | { | |
468 | struct udphdr *uh = udp_hdr(skb); | |
469 | ||
470 | if (uh->check && skb->ip_summed != CHECKSUM_PARTIAL) { | |
471 | set_tp_port(skb, port, new_port, &uh->check); | |
472 | ||
473 | if (!uh->check) | |
474 | uh->check = CSUM_MANGLED_0; | |
475 | } else { | |
476 | *port = new_port; | |
7539fadc | 477 | skb_clear_hash(skb); |
81e5d41d JG |
478 | } |
479 | } | |
480 | ||
481 | static int set_udp(struct sk_buff *skb, const struct ovs_key_udp *udp_port_key) | |
ccb1352e JG |
482 | { |
483 | struct udphdr *uh; | |
484 | int err; | |
485 | ||
486 | err = make_writable(skb, skb_transport_offset(skb) + | |
487 | sizeof(struct udphdr)); | |
488 | if (unlikely(err)) | |
489 | return err; | |
490 | ||
491 | uh = udp_hdr(skb); | |
492 | if (udp_port_key->udp_src != uh->source) | |
81e5d41d | 493 | set_udp_port(skb, &uh->source, udp_port_key->udp_src); |
ccb1352e JG |
494 | |
495 | if (udp_port_key->udp_dst != uh->dest) | |
81e5d41d | 496 | set_udp_port(skb, &uh->dest, udp_port_key->udp_dst); |
ccb1352e JG |
497 | |
498 | return 0; | |
499 | } | |
500 | ||
81e5d41d | 501 | static int set_tcp(struct sk_buff *skb, const struct ovs_key_tcp *tcp_port_key) |
ccb1352e JG |
502 | { |
503 | struct tcphdr *th; | |
504 | int err; | |
505 | ||
506 | err = make_writable(skb, skb_transport_offset(skb) + | |
507 | sizeof(struct tcphdr)); | |
508 | if (unlikely(err)) | |
509 | return err; | |
510 | ||
511 | th = tcp_hdr(skb); | |
512 | if (tcp_port_key->tcp_src != th->source) | |
513 | set_tp_port(skb, &th->source, tcp_port_key->tcp_src, &th->check); | |
514 | ||
515 | if (tcp_port_key->tcp_dst != th->dest) | |
516 | set_tp_port(skb, &th->dest, tcp_port_key->tcp_dst, &th->check); | |
517 | ||
518 | return 0; | |
519 | } | |
520 | ||
a175a723 JS |
521 | static int set_sctp(struct sk_buff *skb, |
522 | const struct ovs_key_sctp *sctp_port_key) | |
523 | { | |
524 | struct sctphdr *sh; | |
525 | int err; | |
526 | unsigned int sctphoff = skb_transport_offset(skb); | |
527 | ||
528 | err = make_writable(skb, sctphoff + sizeof(struct sctphdr)); | |
529 | if (unlikely(err)) | |
530 | return err; | |
531 | ||
532 | sh = sctp_hdr(skb); | |
533 | if (sctp_port_key->sctp_src != sh->source || | |
534 | sctp_port_key->sctp_dst != sh->dest) { | |
535 | __le32 old_correct_csum, new_csum, old_csum; | |
536 | ||
537 | old_csum = sh->checksum; | |
538 | old_correct_csum = sctp_compute_cksum(skb, sctphoff); | |
539 | ||
540 | sh->source = sctp_port_key->sctp_src; | |
541 | sh->dest = sctp_port_key->sctp_dst; | |
542 | ||
543 | new_csum = sctp_compute_cksum(skb, sctphoff); | |
544 | ||
545 | /* Carry any checksum errors through. */ | |
546 | sh->checksum = old_csum ^ old_correct_csum ^ new_csum; | |
547 | ||
7539fadc | 548 | skb_clear_hash(skb); |
a175a723 JS |
549 | } |
550 | ||
551 | return 0; | |
552 | } | |
553 | ||
738967b8 | 554 | static void do_output(struct datapath *dp, struct sk_buff *skb, int out_port) |
ccb1352e | 555 | { |
738967b8 | 556 | struct vport *vport = ovs_vport_rcu(dp, out_port); |
ccb1352e | 557 | |
738967b8 AZ |
558 | if (likely(vport)) |
559 | ovs_vport_send(vport, skb); | |
560 | else | |
ccb1352e | 561 | kfree_skb(skb); |
ccb1352e JG |
562 | } |
563 | ||
564 | static int output_userspace(struct datapath *dp, struct sk_buff *skb, | |
2ff3e4e4 | 565 | struct sw_flow_key *key, const struct nlattr *attr) |
ccb1352e | 566 | { |
8f0aad6f | 567 | struct ovs_tunnel_info info; |
ccb1352e JG |
568 | struct dp_upcall_info upcall; |
569 | const struct nlattr *a; | |
570 | int rem; | |
571 | ||
572 | upcall.cmd = OVS_PACKET_CMD_ACTION; | |
2ff3e4e4 | 573 | upcall.key = key; |
ccb1352e | 574 | upcall.userdata = NULL; |
15e47304 | 575 | upcall.portid = 0; |
8f0aad6f | 576 | upcall.egress_tun_info = NULL; |
ccb1352e JG |
577 | |
578 | for (a = nla_data(attr), rem = nla_len(attr); rem > 0; | |
579 | a = nla_next(a, &rem)) { | |
580 | switch (nla_type(a)) { | |
581 | case OVS_USERSPACE_ATTR_USERDATA: | |
582 | upcall.userdata = a; | |
583 | break; | |
584 | ||
585 | case OVS_USERSPACE_ATTR_PID: | |
15e47304 | 586 | upcall.portid = nla_get_u32(a); |
ccb1352e | 587 | break; |
8f0aad6f WZ |
588 | |
589 | case OVS_USERSPACE_ATTR_EGRESS_TUN_PORT: { | |
590 | /* Get out tunnel info. */ | |
591 | struct vport *vport; | |
592 | ||
593 | vport = ovs_vport_rcu(dp, nla_get_u32(a)); | |
594 | if (vport) { | |
595 | int err; | |
596 | ||
597 | err = ovs_vport_get_egress_tun_info(vport, skb, | |
598 | &info); | |
599 | if (!err) | |
600 | upcall.egress_tun_info = &info; | |
601 | } | |
602 | break; | |
ccb1352e | 603 | } |
8f0aad6f WZ |
604 | |
605 | } /* End of switch. */ | |
ccb1352e JG |
606 | } |
607 | ||
608 | return ovs_dp_upcall(dp, skb, &upcall); | |
609 | } | |
610 | ||
611 | static int sample(struct datapath *dp, struct sk_buff *skb, | |
2ff3e4e4 | 612 | struct sw_flow_key *key, const struct nlattr *attr) |
ccb1352e JG |
613 | { |
614 | const struct nlattr *acts_list = NULL; | |
615 | const struct nlattr *a; | |
616 | int rem; | |
617 | ||
618 | for (a = nla_data(attr), rem = nla_len(attr); rem > 0; | |
619 | a = nla_next(a, &rem)) { | |
620 | switch (nla_type(a)) { | |
621 | case OVS_SAMPLE_ATTR_PROBABILITY: | |
63862b5b | 622 | if (prandom_u32() >= nla_get_u32(a)) |
ccb1352e JG |
623 | return 0; |
624 | break; | |
625 | ||
626 | case OVS_SAMPLE_ATTR_ACTIONS: | |
627 | acts_list = a; | |
628 | break; | |
629 | } | |
630 | } | |
631 | ||
651887b0 SH |
632 | rem = nla_len(acts_list); |
633 | a = nla_data(acts_list); | |
634 | ||
32ae87ff AZ |
635 | /* Actions list is empty, do nothing */ |
636 | if (unlikely(!rem)) | |
637 | return 0; | |
651887b0 | 638 | |
32ae87ff AZ |
639 | /* The only known usage of sample action is having a single user-space |
640 | * action. Treat this usage as a special case. | |
641 | * The output_userspace() should clone the skb to be sent to the | |
642 | * user space. This skb will be consumed by its caller. | |
651887b0 | 643 | */ |
32ae87ff | 644 | if (likely(nla_type(a) == OVS_ACTION_ATTR_USERSPACE && |
941d8ebc | 645 | nla_is_last(a, rem))) |
32ae87ff AZ |
646 | return output_userspace(dp, skb, key, a); |
647 | ||
648 | skb = skb_clone(skb, GFP_ATOMIC); | |
649 | if (!skb) | |
650 | /* Skip the sample action when out of memory. */ | |
651 | return 0; | |
652 | ||
971427f3 AZ |
653 | if (!add_deferred_actions(skb, key, a)) { |
654 | if (net_ratelimit()) | |
655 | pr_warn("%s: deferred actions limit reached, dropping sample action\n", | |
656 | ovs_dp_name(dp)); | |
657 | ||
658 | kfree_skb(skb); | |
659 | } | |
660 | return 0; | |
661 | } | |
662 | ||
663 | static void execute_hash(struct sk_buff *skb, struct sw_flow_key *key, | |
664 | const struct nlattr *attr) | |
665 | { | |
666 | struct ovs_action_hash *hash_act = nla_data(attr); | |
667 | u32 hash = 0; | |
668 | ||
669 | /* OVS_HASH_ALG_L4 is the only possible hash algorithm. */ | |
670 | hash = skb_get_hash(skb); | |
671 | hash = jhash_1word(hash, hash_act->hash_basis); | |
672 | if (!hash) | |
673 | hash = 0x1; | |
674 | ||
675 | key->ovs_flow_hash = hash; | |
ccb1352e JG |
676 | } |
677 | ||
678 | static int execute_set_action(struct sk_buff *skb, | |
679 | const struct nlattr *nested_attr) | |
680 | { | |
681 | int err = 0; | |
682 | ||
683 | switch (nla_type(nested_attr)) { | |
684 | case OVS_KEY_ATTR_PRIORITY: | |
685 | skb->priority = nla_get_u32(nested_attr); | |
686 | break; | |
687 | ||
39c7caeb AA |
688 | case OVS_KEY_ATTR_SKB_MARK: |
689 | skb->mark = nla_get_u32(nested_attr); | |
690 | break; | |
691 | ||
f0b128c1 JG |
692 | case OVS_KEY_ATTR_TUNNEL_INFO: |
693 | OVS_CB(skb)->egress_tun_info = nla_data(nested_attr); | |
7d5437c7 PS |
694 | break; |
695 | ||
ccb1352e JG |
696 | case OVS_KEY_ATTR_ETHERNET: |
697 | err = set_eth_addr(skb, nla_data(nested_attr)); | |
698 | break; | |
699 | ||
700 | case OVS_KEY_ATTR_IPV4: | |
701 | err = set_ipv4(skb, nla_data(nested_attr)); | |
702 | break; | |
703 | ||
3fdbd1ce AA |
704 | case OVS_KEY_ATTR_IPV6: |
705 | err = set_ipv6(skb, nla_data(nested_attr)); | |
706 | break; | |
707 | ||
ccb1352e | 708 | case OVS_KEY_ATTR_TCP: |
81e5d41d | 709 | err = set_tcp(skb, nla_data(nested_attr)); |
ccb1352e JG |
710 | break; |
711 | ||
712 | case OVS_KEY_ATTR_UDP: | |
81e5d41d | 713 | err = set_udp(skb, nla_data(nested_attr)); |
ccb1352e | 714 | break; |
a175a723 JS |
715 | |
716 | case OVS_KEY_ATTR_SCTP: | |
717 | err = set_sctp(skb, nla_data(nested_attr)); | |
718 | break; | |
25cd9ba0 SH |
719 | |
720 | case OVS_KEY_ATTR_MPLS: | |
721 | err = set_mpls(skb, nla_data(nested_attr)); | |
722 | break; | |
ccb1352e JG |
723 | } |
724 | ||
725 | return err; | |
726 | } | |
727 | ||
971427f3 AZ |
728 | static int execute_recirc(struct datapath *dp, struct sk_buff *skb, |
729 | struct sw_flow_key *key, | |
730 | const struct nlattr *a, int rem) | |
731 | { | |
732 | struct deferred_action *da; | |
733 | int err; | |
734 | ||
735 | err = ovs_flow_key_update(skb, key); | |
736 | if (err) | |
737 | return err; | |
738 | ||
941d8ebc | 739 | if (!nla_is_last(a, rem)) { |
971427f3 AZ |
740 | /* Recirc action is the not the last action |
741 | * of the action list, need to clone the skb. | |
742 | */ | |
743 | skb = skb_clone(skb, GFP_ATOMIC); | |
744 | ||
745 | /* Skip the recirc action when out of memory, but | |
746 | * continue on with the rest of the action list. | |
747 | */ | |
748 | if (!skb) | |
749 | return 0; | |
750 | } | |
751 | ||
752 | da = add_deferred_actions(skb, key, NULL); | |
753 | if (da) { | |
754 | da->pkt_key.recirc_id = nla_get_u32(a); | |
755 | } else { | |
756 | kfree_skb(skb); | |
757 | ||
758 | if (net_ratelimit()) | |
759 | pr_warn("%s: deferred action limit reached, drop recirc action\n", | |
760 | ovs_dp_name(dp)); | |
761 | } | |
762 | ||
763 | return 0; | |
764 | } | |
765 | ||
ccb1352e JG |
766 | /* Execute a list of actions against 'skb'. */ |
767 | static int do_execute_actions(struct datapath *dp, struct sk_buff *skb, | |
2ff3e4e4 | 768 | struct sw_flow_key *key, |
651887b0 | 769 | const struct nlattr *attr, int len) |
ccb1352e JG |
770 | { |
771 | /* Every output action needs a separate clone of 'skb', but the common | |
772 | * case is just a single output action, so that doing a clone and | |
773 | * then freeing the original skbuff is wasteful. So the following code | |
774 | * is slightly obscure just to avoid that. */ | |
775 | int prev_port = -1; | |
776 | const struct nlattr *a; | |
777 | int rem; | |
778 | ||
779 | for (a = attr, rem = len; rem > 0; | |
780 | a = nla_next(a, &rem)) { | |
781 | int err = 0; | |
782 | ||
738967b8 AZ |
783 | if (unlikely(prev_port != -1)) { |
784 | struct sk_buff *out_skb = skb_clone(skb, GFP_ATOMIC); | |
785 | ||
786 | if (out_skb) | |
787 | do_output(dp, out_skb, prev_port); | |
788 | ||
ccb1352e JG |
789 | prev_port = -1; |
790 | } | |
791 | ||
792 | switch (nla_type(a)) { | |
793 | case OVS_ACTION_ATTR_OUTPUT: | |
794 | prev_port = nla_get_u32(a); | |
795 | break; | |
796 | ||
797 | case OVS_ACTION_ATTR_USERSPACE: | |
2ff3e4e4 | 798 | output_userspace(dp, skb, key, a); |
ccb1352e JG |
799 | break; |
800 | ||
971427f3 AZ |
801 | case OVS_ACTION_ATTR_HASH: |
802 | execute_hash(skb, key, a); | |
803 | break; | |
804 | ||
25cd9ba0 SH |
805 | case OVS_ACTION_ATTR_PUSH_MPLS: |
806 | err = push_mpls(skb, nla_data(a)); | |
807 | break; | |
808 | ||
809 | case OVS_ACTION_ATTR_POP_MPLS: | |
810 | err = pop_mpls(skb, nla_get_be16(a)); | |
811 | break; | |
812 | ||
ccb1352e JG |
813 | case OVS_ACTION_ATTR_PUSH_VLAN: |
814 | err = push_vlan(skb, nla_data(a)); | |
815 | if (unlikely(err)) /* skb already freed. */ | |
816 | return err; | |
817 | break; | |
818 | ||
819 | case OVS_ACTION_ATTR_POP_VLAN: | |
820 | err = pop_vlan(skb); | |
821 | break; | |
822 | ||
971427f3 AZ |
823 | case OVS_ACTION_ATTR_RECIRC: |
824 | err = execute_recirc(dp, skb, key, a, rem); | |
941d8ebc | 825 | if (nla_is_last(a, rem)) { |
971427f3 AZ |
826 | /* If this is the last action, the skb has |
827 | * been consumed or freed. | |
828 | * Return immediately. | |
829 | */ | |
830 | return err; | |
831 | } | |
832 | break; | |
833 | ||
ccb1352e JG |
834 | case OVS_ACTION_ATTR_SET: |
835 | err = execute_set_action(skb, nla_data(a)); | |
836 | break; | |
837 | ||
838 | case OVS_ACTION_ATTR_SAMPLE: | |
2ff3e4e4 | 839 | err = sample(dp, skb, key, a); |
fe984c08 AZ |
840 | if (unlikely(err)) /* skb already freed. */ |
841 | return err; | |
ccb1352e JG |
842 | break; |
843 | } | |
844 | ||
845 | if (unlikely(err)) { | |
846 | kfree_skb(skb); | |
847 | return err; | |
848 | } | |
849 | } | |
850 | ||
651887b0 | 851 | if (prev_port != -1) |
ccb1352e | 852 | do_output(dp, skb, prev_port); |
651887b0 | 853 | else |
ccb1352e JG |
854 | consume_skb(skb); |
855 | ||
856 | return 0; | |
857 | } | |
858 | ||
971427f3 AZ |
859 | static void process_deferred_actions(struct datapath *dp) |
860 | { | |
861 | struct action_fifo *fifo = this_cpu_ptr(action_fifos); | |
862 | ||
863 | /* Do not touch the FIFO in case there is no deferred actions. */ | |
864 | if (action_fifo_is_empty(fifo)) | |
865 | return; | |
866 | ||
867 | /* Finishing executing all deferred actions. */ | |
868 | do { | |
869 | struct deferred_action *da = action_fifo_get(fifo); | |
870 | struct sk_buff *skb = da->skb; | |
871 | struct sw_flow_key *key = &da->pkt_key; | |
872 | const struct nlattr *actions = da->actions; | |
873 | ||
874 | if (actions) | |
875 | do_execute_actions(dp, skb, key, actions, | |
876 | nla_len(actions)); | |
877 | else | |
878 | ovs_dp_process_packet(skb, key); | |
879 | } while (!action_fifo_is_empty(fifo)); | |
880 | ||
881 | /* Reset FIFO for the next packet. */ | |
882 | action_fifo_init(fifo); | |
883 | } | |
884 | ||
ccb1352e | 885 | /* Execute a list of actions against 'skb'. */ |
2ff3e4e4 | 886 | int ovs_execute_actions(struct datapath *dp, struct sk_buff *skb, |
d98612b8 | 887 | struct sw_flow_actions *acts, struct sw_flow_key *key) |
ccb1352e | 888 | { |
971427f3 | 889 | int level = this_cpu_read(exec_actions_level); |
971427f3 AZ |
890 | int err; |
891 | ||
971427f3 | 892 | this_cpu_inc(exec_actions_level); |
f0b128c1 | 893 | OVS_CB(skb)->egress_tun_info = NULL; |
971427f3 AZ |
894 | err = do_execute_actions(dp, skb, key, |
895 | acts->actions, acts->actions_len); | |
896 | ||
897 | if (!level) | |
898 | process_deferred_actions(dp); | |
899 | ||
900 | this_cpu_dec(exec_actions_level); | |
901 | return err; | |
902 | } | |
903 | ||
904 | int action_fifos_init(void) | |
905 | { | |
906 | action_fifos = alloc_percpu(struct action_fifo); | |
907 | if (!action_fifos) | |
908 | return -ENOMEM; | |
ccb1352e | 909 | |
971427f3 AZ |
910 | return 0; |
911 | } | |
912 | ||
913 | void action_fifos_exit(void) | |
914 | { | |
915 | free_percpu(action_fifos); | |
ccb1352e | 916 | } |