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 | ||
12eb18f7 | 72 | static bool action_fifo_is_empty(const struct action_fifo *fifo) |
971427f3 AZ |
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, | |
12eb18f7 | 95 | const struct sw_flow_key *key, |
971427f3 AZ |
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 | ||
fff06c36 PS |
112 | static void invalidate_flow_key(struct sw_flow_key *key) |
113 | { | |
114 | key->eth.type = htons(0); | |
115 | } | |
116 | ||
117 | static bool is_flow_key_valid(const struct sw_flow_key *key) | |
118 | { | |
119 | return !!key->eth.type; | |
120 | } | |
121 | ||
fff06c36 | 122 | static int push_mpls(struct sk_buff *skb, struct sw_flow_key *key, |
25cd9ba0 SH |
123 | const struct ovs_action_push_mpls *mpls) |
124 | { | |
125 | __be32 *new_mpls_lse; | |
126 | struct ethhdr *hdr; | |
127 | ||
128 | /* Networking stack do not allow simultaneous Tunnel and MPLS GSO. */ | |
129 | if (skb->encapsulation) | |
130 | return -ENOTSUPP; | |
131 | ||
132 | if (skb_cow_head(skb, MPLS_HLEN) < 0) | |
133 | return -ENOMEM; | |
134 | ||
135 | skb_push(skb, MPLS_HLEN); | |
136 | memmove(skb_mac_header(skb) - MPLS_HLEN, skb_mac_header(skb), | |
137 | skb->mac_len); | |
138 | skb_reset_mac_header(skb); | |
139 | ||
140 | new_mpls_lse = (__be32 *)skb_mpls_header(skb); | |
141 | *new_mpls_lse = mpls->mpls_lse; | |
142 | ||
143 | if (skb->ip_summed == CHECKSUM_COMPLETE) | |
144 | skb->csum = csum_add(skb->csum, csum_partial(new_mpls_lse, | |
145 | MPLS_HLEN, 0)); | |
146 | ||
147 | hdr = eth_hdr(skb); | |
148 | hdr->h_proto = mpls->mpls_ethertype; | |
149 | ||
cbe7e76d PS |
150 | if (!skb->inner_protocol) |
151 | skb_set_inner_protocol(skb, skb->protocol); | |
25cd9ba0 SH |
152 | skb->protocol = mpls->mpls_ethertype; |
153 | ||
fff06c36 | 154 | invalidate_flow_key(key); |
25cd9ba0 SH |
155 | return 0; |
156 | } | |
157 | ||
fff06c36 PS |
158 | static int pop_mpls(struct sk_buff *skb, struct sw_flow_key *key, |
159 | const __be16 ethertype) | |
25cd9ba0 SH |
160 | { |
161 | struct ethhdr *hdr; | |
162 | int err; | |
163 | ||
e2195121 | 164 | err = skb_ensure_writable(skb, skb->mac_len + MPLS_HLEN); |
25cd9ba0 SH |
165 | if (unlikely(err)) |
166 | return err; | |
167 | ||
1abcd82c | 168 | skb_postpull_rcsum(skb, skb_mpls_header(skb), MPLS_HLEN); |
25cd9ba0 SH |
169 | |
170 | memmove(skb_mac_header(skb) + MPLS_HLEN, skb_mac_header(skb), | |
171 | skb->mac_len); | |
172 | ||
173 | __skb_pull(skb, MPLS_HLEN); | |
174 | skb_reset_mac_header(skb); | |
175 | ||
176 | /* skb_mpls_header() is used to locate the ethertype | |
177 | * field correctly in the presence of VLAN tags. | |
178 | */ | |
179 | hdr = (struct ethhdr *)(skb_mpls_header(skb) - ETH_HLEN); | |
180 | hdr->h_proto = ethertype; | |
181 | if (eth_p_mpls(skb->protocol)) | |
182 | skb->protocol = ethertype; | |
fff06c36 PS |
183 | |
184 | invalidate_flow_key(key); | |
25cd9ba0 SH |
185 | return 0; |
186 | } | |
187 | ||
83d2b9ba JR |
188 | /* 'KEY' must not have any bits set outside of the 'MASK' */ |
189 | #define MASKED(OLD, KEY, MASK) ((KEY) | ((OLD) & ~(MASK))) | |
190 | #define SET_MASKED(OLD, KEY, MASK) ((OLD) = MASKED(OLD, KEY, MASK)) | |
191 | ||
192 | static int set_mpls(struct sk_buff *skb, struct sw_flow_key *flow_key, | |
193 | const __be32 *mpls_lse, const __be32 *mask) | |
25cd9ba0 SH |
194 | { |
195 | __be32 *stack; | |
83d2b9ba | 196 | __be32 lse; |
25cd9ba0 SH |
197 | int err; |
198 | ||
e2195121 | 199 | err = skb_ensure_writable(skb, skb->mac_len + MPLS_HLEN); |
25cd9ba0 SH |
200 | if (unlikely(err)) |
201 | return err; | |
202 | ||
203 | stack = (__be32 *)skb_mpls_header(skb); | |
83d2b9ba | 204 | lse = MASKED(*stack, *mpls_lse, *mask); |
25cd9ba0 | 205 | if (skb->ip_summed == CHECKSUM_COMPLETE) { |
83d2b9ba JR |
206 | __be32 diff[] = { ~(*stack), lse }; |
207 | ||
25cd9ba0 SH |
208 | skb->csum = ~csum_partial((char *)diff, sizeof(diff), |
209 | ~skb->csum); | |
210 | } | |
211 | ||
83d2b9ba JR |
212 | *stack = lse; |
213 | flow_key->mpls.top_lse = lse; | |
25cd9ba0 SH |
214 | return 0; |
215 | } | |
216 | ||
fff06c36 | 217 | static int pop_vlan(struct sk_buff *skb, struct sw_flow_key *key) |
ccb1352e | 218 | { |
ccb1352e JG |
219 | int err; |
220 | ||
93515d53 | 221 | err = skb_vlan_pop(skb); |
df8a39de | 222 | if (skb_vlan_tag_present(skb)) |
93515d53 JP |
223 | invalidate_flow_key(key); |
224 | else | |
fff06c36 | 225 | key->eth.tci = 0; |
93515d53 | 226 | return err; |
ccb1352e JG |
227 | } |
228 | ||
fff06c36 PS |
229 | static int push_vlan(struct sk_buff *skb, struct sw_flow_key *key, |
230 | const struct ovs_action_push_vlan *vlan) | |
ccb1352e | 231 | { |
df8a39de | 232 | if (skb_vlan_tag_present(skb)) |
fff06c36 | 233 | invalidate_flow_key(key); |
93515d53 | 234 | else |
fff06c36 | 235 | key->eth.tci = vlan->vlan_tci; |
93515d53 JP |
236 | return skb_vlan_push(skb, vlan->vlan_tpid, |
237 | ntohs(vlan->vlan_tci) & ~VLAN_TAG_PRESENT); | |
ccb1352e JG |
238 | } |
239 | ||
83d2b9ba JR |
240 | /* 'src' is already properly masked. */ |
241 | static void ether_addr_copy_masked(u8 *dst_, const u8 *src_, const u8 *mask_) | |
242 | { | |
243 | u16 *dst = (u16 *)dst_; | |
244 | const u16 *src = (const u16 *)src_; | |
245 | const u16 *mask = (const u16 *)mask_; | |
246 | ||
247 | SET_MASKED(dst[0], src[0], mask[0]); | |
248 | SET_MASKED(dst[1], src[1], mask[1]); | |
249 | SET_MASKED(dst[2], src[2], mask[2]); | |
250 | } | |
251 | ||
252 | static int set_eth_addr(struct sk_buff *skb, struct sw_flow_key *flow_key, | |
253 | const struct ovs_key_ethernet *key, | |
254 | const struct ovs_key_ethernet *mask) | |
ccb1352e JG |
255 | { |
256 | int err; | |
83d2b9ba | 257 | |
e2195121 | 258 | err = skb_ensure_writable(skb, ETH_HLEN); |
ccb1352e JG |
259 | if (unlikely(err)) |
260 | return err; | |
261 | ||
b34df5e8 PS |
262 | skb_postpull_rcsum(skb, eth_hdr(skb), ETH_ALEN * 2); |
263 | ||
83d2b9ba JR |
264 | ether_addr_copy_masked(eth_hdr(skb)->h_source, key->eth_src, |
265 | mask->eth_src); | |
266 | ether_addr_copy_masked(eth_hdr(skb)->h_dest, key->eth_dst, | |
267 | mask->eth_dst); | |
ccb1352e | 268 | |
b34df5e8 PS |
269 | ovs_skb_postpush_rcsum(skb, eth_hdr(skb), ETH_ALEN * 2); |
270 | ||
83d2b9ba JR |
271 | ether_addr_copy(flow_key->eth.src, eth_hdr(skb)->h_source); |
272 | ether_addr_copy(flow_key->eth.dst, eth_hdr(skb)->h_dest); | |
ccb1352e JG |
273 | return 0; |
274 | } | |
275 | ||
276 | static void set_ip_addr(struct sk_buff *skb, struct iphdr *nh, | |
fff06c36 | 277 | __be32 *addr, __be32 new_addr) |
ccb1352e JG |
278 | { |
279 | int transport_len = skb->len - skb_transport_offset(skb); | |
280 | ||
281 | if (nh->protocol == IPPROTO_TCP) { | |
282 | if (likely(transport_len >= sizeof(struct tcphdr))) | |
283 | inet_proto_csum_replace4(&tcp_hdr(skb)->check, skb, | |
284 | *addr, new_addr, 1); | |
285 | } else if (nh->protocol == IPPROTO_UDP) { | |
81e5d41d JG |
286 | if (likely(transport_len >= sizeof(struct udphdr))) { |
287 | struct udphdr *uh = udp_hdr(skb); | |
288 | ||
289 | if (uh->check || skb->ip_summed == CHECKSUM_PARTIAL) { | |
290 | inet_proto_csum_replace4(&uh->check, skb, | |
291 | *addr, new_addr, 1); | |
292 | if (!uh->check) | |
293 | uh->check = CSUM_MANGLED_0; | |
294 | } | |
295 | } | |
ccb1352e JG |
296 | } |
297 | ||
298 | csum_replace4(&nh->check, *addr, new_addr); | |
7539fadc | 299 | skb_clear_hash(skb); |
ccb1352e JG |
300 | *addr = new_addr; |
301 | } | |
302 | ||
3fdbd1ce AA |
303 | static void update_ipv6_checksum(struct sk_buff *skb, u8 l4_proto, |
304 | __be32 addr[4], const __be32 new_addr[4]) | |
305 | { | |
306 | int transport_len = skb->len - skb_transport_offset(skb); | |
307 | ||
856447d0 | 308 | if (l4_proto == NEXTHDR_TCP) { |
3fdbd1ce AA |
309 | if (likely(transport_len >= sizeof(struct tcphdr))) |
310 | inet_proto_csum_replace16(&tcp_hdr(skb)->check, skb, | |
311 | addr, new_addr, 1); | |
856447d0 | 312 | } else if (l4_proto == NEXTHDR_UDP) { |
3fdbd1ce AA |
313 | if (likely(transport_len >= sizeof(struct udphdr))) { |
314 | struct udphdr *uh = udp_hdr(skb); | |
315 | ||
316 | if (uh->check || skb->ip_summed == CHECKSUM_PARTIAL) { | |
317 | inet_proto_csum_replace16(&uh->check, skb, | |
318 | addr, new_addr, 1); | |
319 | if (!uh->check) | |
320 | uh->check = CSUM_MANGLED_0; | |
321 | } | |
322 | } | |
856447d0 JG |
323 | } else if (l4_proto == NEXTHDR_ICMP) { |
324 | if (likely(transport_len >= sizeof(struct icmp6hdr))) | |
325 | inet_proto_csum_replace16(&icmp6_hdr(skb)->icmp6_cksum, | |
326 | skb, addr, new_addr, 1); | |
3fdbd1ce AA |
327 | } |
328 | } | |
329 | ||
83d2b9ba JR |
330 | static void mask_ipv6_addr(const __be32 old[4], const __be32 addr[4], |
331 | const __be32 mask[4], __be32 masked[4]) | |
332 | { | |
333 | masked[0] = MASKED(old[0], addr[0], mask[0]); | |
334 | masked[1] = MASKED(old[1], addr[1], mask[1]); | |
335 | masked[2] = MASKED(old[2], addr[2], mask[2]); | |
336 | masked[3] = MASKED(old[3], addr[3], mask[3]); | |
337 | } | |
338 | ||
3fdbd1ce AA |
339 | static void set_ipv6_addr(struct sk_buff *skb, u8 l4_proto, |
340 | __be32 addr[4], const __be32 new_addr[4], | |
341 | bool recalculate_csum) | |
342 | { | |
343 | if (recalculate_csum) | |
344 | update_ipv6_checksum(skb, l4_proto, addr, new_addr); | |
345 | ||
7539fadc | 346 | skb_clear_hash(skb); |
3fdbd1ce AA |
347 | memcpy(addr, new_addr, sizeof(__be32[4])); |
348 | } | |
349 | ||
83d2b9ba | 350 | static void set_ipv6_fl(struct ipv6hdr *nh, u32 fl, u32 mask) |
3fdbd1ce | 351 | { |
83d2b9ba JR |
352 | /* Bits 21-24 are always unmasked, so this retains their values. */ |
353 | SET_MASKED(nh->flow_lbl[0], (u8)(fl >> 16), (u8)(mask >> 16)); | |
354 | SET_MASKED(nh->flow_lbl[1], (u8)(fl >> 8), (u8)(mask >> 8)); | |
355 | SET_MASKED(nh->flow_lbl[2], (u8)fl, (u8)mask); | |
3fdbd1ce AA |
356 | } |
357 | ||
83d2b9ba JR |
358 | static void set_ip_ttl(struct sk_buff *skb, struct iphdr *nh, u8 new_ttl, |
359 | u8 mask) | |
3fdbd1ce | 360 | { |
83d2b9ba | 361 | new_ttl = MASKED(nh->ttl, new_ttl, mask); |
3fdbd1ce | 362 | |
ccb1352e JG |
363 | csum_replace2(&nh->check, htons(nh->ttl << 8), htons(new_ttl << 8)); |
364 | nh->ttl = new_ttl; | |
365 | } | |
366 | ||
83d2b9ba JR |
367 | static int set_ipv4(struct sk_buff *skb, struct sw_flow_key *flow_key, |
368 | const struct ovs_key_ipv4 *key, | |
369 | const struct ovs_key_ipv4 *mask) | |
ccb1352e JG |
370 | { |
371 | struct iphdr *nh; | |
83d2b9ba | 372 | __be32 new_addr; |
ccb1352e JG |
373 | int err; |
374 | ||
e2195121 JP |
375 | err = skb_ensure_writable(skb, skb_network_offset(skb) + |
376 | sizeof(struct iphdr)); | |
ccb1352e JG |
377 | if (unlikely(err)) |
378 | return err; | |
379 | ||
380 | nh = ip_hdr(skb); | |
381 | ||
83d2b9ba JR |
382 | /* Setting an IP addresses is typically only a side effect of |
383 | * matching on them in the current userspace implementation, so it | |
384 | * makes sense to check if the value actually changed. | |
385 | */ | |
386 | if (mask->ipv4_src) { | |
387 | new_addr = MASKED(nh->saddr, key->ipv4_src, mask->ipv4_src); | |
ccb1352e | 388 | |
83d2b9ba JR |
389 | if (unlikely(new_addr != nh->saddr)) { |
390 | set_ip_addr(skb, nh, &nh->saddr, new_addr); | |
391 | flow_key->ipv4.addr.src = new_addr; | |
392 | } | |
fff06c36 | 393 | } |
83d2b9ba JR |
394 | if (mask->ipv4_dst) { |
395 | new_addr = MASKED(nh->daddr, key->ipv4_dst, mask->ipv4_dst); | |
ccb1352e | 396 | |
83d2b9ba JR |
397 | if (unlikely(new_addr != nh->daddr)) { |
398 | set_ip_addr(skb, nh, &nh->daddr, new_addr); | |
399 | flow_key->ipv4.addr.dst = new_addr; | |
400 | } | |
fff06c36 | 401 | } |
83d2b9ba JR |
402 | if (mask->ipv4_tos) { |
403 | ipv4_change_dsfield(nh, ~mask->ipv4_tos, key->ipv4_tos); | |
404 | flow_key->ip.tos = nh->tos; | |
405 | } | |
406 | if (mask->ipv4_ttl) { | |
407 | set_ip_ttl(skb, nh, key->ipv4_ttl, mask->ipv4_ttl); | |
408 | flow_key->ip.ttl = nh->ttl; | |
fff06c36 | 409 | } |
ccb1352e JG |
410 | |
411 | return 0; | |
412 | } | |
413 | ||
83d2b9ba JR |
414 | static bool is_ipv6_mask_nonzero(const __be32 addr[4]) |
415 | { | |
416 | return !!(addr[0] | addr[1] | addr[2] | addr[3]); | |
417 | } | |
418 | ||
419 | static int set_ipv6(struct sk_buff *skb, struct sw_flow_key *flow_key, | |
420 | const struct ovs_key_ipv6 *key, | |
421 | const struct ovs_key_ipv6 *mask) | |
3fdbd1ce AA |
422 | { |
423 | struct ipv6hdr *nh; | |
424 | int err; | |
3fdbd1ce | 425 | |
e2195121 JP |
426 | err = skb_ensure_writable(skb, skb_network_offset(skb) + |
427 | sizeof(struct ipv6hdr)); | |
3fdbd1ce AA |
428 | if (unlikely(err)) |
429 | return err; | |
430 | ||
431 | nh = ipv6_hdr(skb); | |
3fdbd1ce | 432 | |
83d2b9ba JR |
433 | /* Setting an IP addresses is typically only a side effect of |
434 | * matching on them in the current userspace implementation, so it | |
435 | * makes sense to check if the value actually changed. | |
436 | */ | |
437 | if (is_ipv6_mask_nonzero(mask->ipv6_src)) { | |
438 | __be32 *saddr = (__be32 *)&nh->saddr; | |
439 | __be32 masked[4]; | |
440 | ||
441 | mask_ipv6_addr(saddr, key->ipv6_src, mask->ipv6_src, masked); | |
442 | ||
443 | if (unlikely(memcmp(saddr, masked, sizeof(masked)))) { | |
444 | set_ipv6_addr(skb, key->ipv6_proto, saddr, masked, | |
445 | true); | |
446 | memcpy(&flow_key->ipv6.addr.src, masked, | |
447 | sizeof(flow_key->ipv6.addr.src)); | |
448 | } | |
449 | } | |
450 | if (is_ipv6_mask_nonzero(mask->ipv6_dst)) { | |
3fdbd1ce AA |
451 | unsigned int offset = 0; |
452 | int flags = IP6_FH_F_SKIP_RH; | |
453 | bool recalc_csum = true; | |
83d2b9ba JR |
454 | __be32 *daddr = (__be32 *)&nh->daddr; |
455 | __be32 masked[4]; | |
456 | ||
457 | mask_ipv6_addr(daddr, key->ipv6_dst, mask->ipv6_dst, masked); | |
458 | ||
459 | if (unlikely(memcmp(daddr, masked, sizeof(masked)))) { | |
460 | if (ipv6_ext_hdr(nh->nexthdr)) | |
461 | recalc_csum = (ipv6_find_hdr(skb, &offset, | |
462 | NEXTHDR_ROUTING, | |
463 | NULL, &flags) | |
464 | != NEXTHDR_ROUTING); | |
465 | ||
466 | set_ipv6_addr(skb, key->ipv6_proto, daddr, masked, | |
467 | recalc_csum); | |
468 | memcpy(&flow_key->ipv6.addr.dst, masked, | |
469 | sizeof(flow_key->ipv6.addr.dst)); | |
470 | } | |
471 | } | |
472 | if (mask->ipv6_tclass) { | |
473 | ipv6_change_dsfield(nh, ~mask->ipv6_tclass, key->ipv6_tclass); | |
474 | flow_key->ip.tos = ipv6_get_dsfield(nh); | |
475 | } | |
476 | if (mask->ipv6_label) { | |
477 | set_ipv6_fl(nh, ntohl(key->ipv6_label), | |
478 | ntohl(mask->ipv6_label)); | |
479 | flow_key->ipv6.label = | |
480 | *(__be32 *)nh & htonl(IPV6_FLOWINFO_FLOWLABEL); | |
481 | } | |
482 | if (mask->ipv6_hlimit) { | |
483 | SET_MASKED(nh->hop_limit, key->ipv6_hlimit, mask->ipv6_hlimit); | |
484 | flow_key->ip.ttl = nh->hop_limit; | |
3fdbd1ce | 485 | } |
3fdbd1ce AA |
486 | return 0; |
487 | } | |
488 | ||
e2195121 | 489 | /* Must follow skb_ensure_writable() since that can move the skb data. */ |
ccb1352e | 490 | static void set_tp_port(struct sk_buff *skb, __be16 *port, |
83d2b9ba | 491 | __be16 new_port, __sum16 *check) |
ccb1352e JG |
492 | { |
493 | inet_proto_csum_replace2(check, skb, *port, new_port, 0); | |
494 | *port = new_port; | |
81e5d41d JG |
495 | } |
496 | ||
83d2b9ba JR |
497 | static int set_udp(struct sk_buff *skb, struct sw_flow_key *flow_key, |
498 | const struct ovs_key_udp *key, | |
499 | const struct ovs_key_udp *mask) | |
ccb1352e JG |
500 | { |
501 | struct udphdr *uh; | |
83d2b9ba | 502 | __be16 src, dst; |
ccb1352e JG |
503 | int err; |
504 | ||
e2195121 JP |
505 | err = skb_ensure_writable(skb, skb_transport_offset(skb) + |
506 | sizeof(struct udphdr)); | |
ccb1352e JG |
507 | if (unlikely(err)) |
508 | return err; | |
509 | ||
510 | uh = udp_hdr(skb); | |
83d2b9ba JR |
511 | /* Either of the masks is non-zero, so do not bother checking them. */ |
512 | src = MASKED(uh->source, key->udp_src, mask->udp_src); | |
513 | dst = MASKED(uh->dest, key->udp_dst, mask->udp_dst); | |
ccb1352e | 514 | |
83d2b9ba JR |
515 | if (uh->check && skb->ip_summed != CHECKSUM_PARTIAL) { |
516 | if (likely(src != uh->source)) { | |
517 | set_tp_port(skb, &uh->source, src, &uh->check); | |
518 | flow_key->tp.src = src; | |
519 | } | |
520 | if (likely(dst != uh->dest)) { | |
521 | set_tp_port(skb, &uh->dest, dst, &uh->check); | |
522 | flow_key->tp.dst = dst; | |
523 | } | |
524 | ||
525 | if (unlikely(!uh->check)) | |
526 | uh->check = CSUM_MANGLED_0; | |
527 | } else { | |
528 | uh->source = src; | |
529 | uh->dest = dst; | |
530 | flow_key->tp.src = src; | |
531 | flow_key->tp.dst = dst; | |
fff06c36 | 532 | } |
ccb1352e | 533 | |
83d2b9ba JR |
534 | skb_clear_hash(skb); |
535 | ||
ccb1352e JG |
536 | return 0; |
537 | } | |
538 | ||
83d2b9ba JR |
539 | static int set_tcp(struct sk_buff *skb, struct sw_flow_key *flow_key, |
540 | const struct ovs_key_tcp *key, | |
541 | const struct ovs_key_tcp *mask) | |
ccb1352e JG |
542 | { |
543 | struct tcphdr *th; | |
83d2b9ba | 544 | __be16 src, dst; |
ccb1352e JG |
545 | int err; |
546 | ||
e2195121 JP |
547 | err = skb_ensure_writable(skb, skb_transport_offset(skb) + |
548 | sizeof(struct tcphdr)); | |
ccb1352e JG |
549 | if (unlikely(err)) |
550 | return err; | |
551 | ||
552 | th = tcp_hdr(skb); | |
83d2b9ba JR |
553 | src = MASKED(th->source, key->tcp_src, mask->tcp_src); |
554 | if (likely(src != th->source)) { | |
555 | set_tp_port(skb, &th->source, src, &th->check); | |
556 | flow_key->tp.src = src; | |
fff06c36 | 557 | } |
83d2b9ba JR |
558 | dst = MASKED(th->dest, key->tcp_dst, mask->tcp_dst); |
559 | if (likely(dst != th->dest)) { | |
560 | set_tp_port(skb, &th->dest, dst, &th->check); | |
561 | flow_key->tp.dst = dst; | |
fff06c36 | 562 | } |
83d2b9ba | 563 | skb_clear_hash(skb); |
ccb1352e JG |
564 | |
565 | return 0; | |
566 | } | |
567 | ||
83d2b9ba JR |
568 | static int set_sctp(struct sk_buff *skb, struct sw_flow_key *flow_key, |
569 | const struct ovs_key_sctp *key, | |
570 | const struct ovs_key_sctp *mask) | |
a175a723 | 571 | { |
83d2b9ba | 572 | unsigned int sctphoff = skb_transport_offset(skb); |
a175a723 | 573 | struct sctphdr *sh; |
83d2b9ba | 574 | __le32 old_correct_csum, new_csum, old_csum; |
a175a723 | 575 | int err; |
a175a723 | 576 | |
e2195121 | 577 | err = skb_ensure_writable(skb, sctphoff + sizeof(struct sctphdr)); |
a175a723 JS |
578 | if (unlikely(err)) |
579 | return err; | |
580 | ||
581 | sh = sctp_hdr(skb); | |
83d2b9ba JR |
582 | old_csum = sh->checksum; |
583 | old_correct_csum = sctp_compute_cksum(skb, sctphoff); | |
a175a723 | 584 | |
83d2b9ba JR |
585 | sh->source = MASKED(sh->source, key->sctp_src, mask->sctp_src); |
586 | sh->dest = MASKED(sh->dest, key->sctp_dst, mask->sctp_dst); | |
a175a723 | 587 | |
83d2b9ba | 588 | new_csum = sctp_compute_cksum(skb, sctphoff); |
a175a723 | 589 | |
83d2b9ba JR |
590 | /* Carry any checksum errors through. */ |
591 | sh->checksum = old_csum ^ old_correct_csum ^ new_csum; | |
a175a723 | 592 | |
83d2b9ba JR |
593 | skb_clear_hash(skb); |
594 | flow_key->tp.src = sh->source; | |
595 | flow_key->tp.dst = sh->dest; | |
a175a723 JS |
596 | |
597 | return 0; | |
598 | } | |
599 | ||
738967b8 | 600 | static void do_output(struct datapath *dp, struct sk_buff *skb, int out_port) |
ccb1352e | 601 | { |
738967b8 | 602 | struct vport *vport = ovs_vport_rcu(dp, out_port); |
ccb1352e | 603 | |
738967b8 AZ |
604 | if (likely(vport)) |
605 | ovs_vport_send(vport, skb); | |
606 | else | |
ccb1352e | 607 | kfree_skb(skb); |
ccb1352e JG |
608 | } |
609 | ||
610 | static int output_userspace(struct datapath *dp, struct sk_buff *skb, | |
2ff3e4e4 | 611 | struct sw_flow_key *key, const struct nlattr *attr) |
ccb1352e | 612 | { |
8f0aad6f | 613 | struct ovs_tunnel_info info; |
ccb1352e JG |
614 | struct dp_upcall_info upcall; |
615 | const struct nlattr *a; | |
616 | int rem; | |
617 | ||
618 | upcall.cmd = OVS_PACKET_CMD_ACTION; | |
ccb1352e | 619 | upcall.userdata = NULL; |
15e47304 | 620 | upcall.portid = 0; |
8f0aad6f | 621 | upcall.egress_tun_info = NULL; |
ccb1352e JG |
622 | |
623 | for (a = nla_data(attr), rem = nla_len(attr); rem > 0; | |
624 | a = nla_next(a, &rem)) { | |
625 | switch (nla_type(a)) { | |
626 | case OVS_USERSPACE_ATTR_USERDATA: | |
627 | upcall.userdata = a; | |
628 | break; | |
629 | ||
630 | case OVS_USERSPACE_ATTR_PID: | |
15e47304 | 631 | upcall.portid = nla_get_u32(a); |
ccb1352e | 632 | break; |
8f0aad6f WZ |
633 | |
634 | case OVS_USERSPACE_ATTR_EGRESS_TUN_PORT: { | |
635 | /* Get out tunnel info. */ | |
636 | struct vport *vport; | |
637 | ||
638 | vport = ovs_vport_rcu(dp, nla_get_u32(a)); | |
639 | if (vport) { | |
640 | int err; | |
641 | ||
642 | err = ovs_vport_get_egress_tun_info(vport, skb, | |
643 | &info); | |
644 | if (!err) | |
645 | upcall.egress_tun_info = &info; | |
646 | } | |
647 | break; | |
ccb1352e | 648 | } |
8f0aad6f WZ |
649 | |
650 | } /* End of switch. */ | |
ccb1352e JG |
651 | } |
652 | ||
e8eedb85 | 653 | return ovs_dp_upcall(dp, skb, key, &upcall); |
ccb1352e JG |
654 | } |
655 | ||
656 | static int sample(struct datapath *dp, struct sk_buff *skb, | |
2ff3e4e4 | 657 | struct sw_flow_key *key, const struct nlattr *attr) |
ccb1352e JG |
658 | { |
659 | const struct nlattr *acts_list = NULL; | |
660 | const struct nlattr *a; | |
661 | int rem; | |
662 | ||
663 | for (a = nla_data(attr), rem = nla_len(attr); rem > 0; | |
664 | a = nla_next(a, &rem)) { | |
665 | switch (nla_type(a)) { | |
666 | case OVS_SAMPLE_ATTR_PROBABILITY: | |
63862b5b | 667 | if (prandom_u32() >= nla_get_u32(a)) |
ccb1352e JG |
668 | return 0; |
669 | break; | |
670 | ||
671 | case OVS_SAMPLE_ATTR_ACTIONS: | |
672 | acts_list = a; | |
673 | break; | |
674 | } | |
675 | } | |
676 | ||
651887b0 SH |
677 | rem = nla_len(acts_list); |
678 | a = nla_data(acts_list); | |
679 | ||
32ae87ff AZ |
680 | /* Actions list is empty, do nothing */ |
681 | if (unlikely(!rem)) | |
682 | return 0; | |
651887b0 | 683 | |
32ae87ff AZ |
684 | /* The only known usage of sample action is having a single user-space |
685 | * action. Treat this usage as a special case. | |
686 | * The output_userspace() should clone the skb to be sent to the | |
687 | * user space. This skb will be consumed by its caller. | |
651887b0 | 688 | */ |
32ae87ff | 689 | if (likely(nla_type(a) == OVS_ACTION_ATTR_USERSPACE && |
941d8ebc | 690 | nla_is_last(a, rem))) |
32ae87ff AZ |
691 | return output_userspace(dp, skb, key, a); |
692 | ||
693 | skb = skb_clone(skb, GFP_ATOMIC); | |
694 | if (!skb) | |
695 | /* Skip the sample action when out of memory. */ | |
696 | return 0; | |
697 | ||
971427f3 AZ |
698 | if (!add_deferred_actions(skb, key, a)) { |
699 | if (net_ratelimit()) | |
700 | pr_warn("%s: deferred actions limit reached, dropping sample action\n", | |
701 | ovs_dp_name(dp)); | |
702 | ||
703 | kfree_skb(skb); | |
704 | } | |
705 | return 0; | |
706 | } | |
707 | ||
708 | static void execute_hash(struct sk_buff *skb, struct sw_flow_key *key, | |
709 | const struct nlattr *attr) | |
710 | { | |
711 | struct ovs_action_hash *hash_act = nla_data(attr); | |
712 | u32 hash = 0; | |
713 | ||
714 | /* OVS_HASH_ALG_L4 is the only possible hash algorithm. */ | |
715 | hash = skb_get_hash(skb); | |
716 | hash = jhash_1word(hash, hash_act->hash_basis); | |
717 | if (!hash) | |
718 | hash = 0x1; | |
719 | ||
720 | key->ovs_flow_hash = hash; | |
ccb1352e JG |
721 | } |
722 | ||
83d2b9ba JR |
723 | static int execute_set_action(struct sk_buff *skb, |
724 | struct sw_flow_key *flow_key, | |
725 | const struct nlattr *a) | |
726 | { | |
727 | /* Only tunnel set execution is supported without a mask. */ | |
728 | if (nla_type(a) == OVS_KEY_ATTR_TUNNEL_INFO) { | |
729 | OVS_CB(skb)->egress_tun_info = nla_data(a); | |
730 | return 0; | |
731 | } | |
732 | ||
733 | return -EINVAL; | |
734 | } | |
735 | ||
736 | /* Mask is at the midpoint of the data. */ | |
737 | #define get_mask(a, type) ((const type)nla_data(a) + 1) | |
738 | ||
739 | static int execute_masked_set_action(struct sk_buff *skb, | |
740 | struct sw_flow_key *flow_key, | |
741 | const struct nlattr *a) | |
ccb1352e JG |
742 | { |
743 | int err = 0; | |
744 | ||
83d2b9ba | 745 | switch (nla_type(a)) { |
ccb1352e | 746 | case OVS_KEY_ATTR_PRIORITY: |
83d2b9ba JR |
747 | SET_MASKED(skb->priority, nla_get_u32(a), *get_mask(a, u32 *)); |
748 | flow_key->phy.priority = skb->priority; | |
ccb1352e JG |
749 | break; |
750 | ||
39c7caeb | 751 | case OVS_KEY_ATTR_SKB_MARK: |
83d2b9ba JR |
752 | SET_MASKED(skb->mark, nla_get_u32(a), *get_mask(a, u32 *)); |
753 | flow_key->phy.skb_mark = skb->mark; | |
39c7caeb AA |
754 | break; |
755 | ||
f0b128c1 | 756 | case OVS_KEY_ATTR_TUNNEL_INFO: |
83d2b9ba JR |
757 | /* Masked data not supported for tunnel. */ |
758 | err = -EINVAL; | |
7d5437c7 PS |
759 | break; |
760 | ||
ccb1352e | 761 | case OVS_KEY_ATTR_ETHERNET: |
83d2b9ba JR |
762 | err = set_eth_addr(skb, flow_key, nla_data(a), |
763 | get_mask(a, struct ovs_key_ethernet *)); | |
ccb1352e JG |
764 | break; |
765 | ||
766 | case OVS_KEY_ATTR_IPV4: | |
83d2b9ba JR |
767 | err = set_ipv4(skb, flow_key, nla_data(a), |
768 | get_mask(a, struct ovs_key_ipv4 *)); | |
ccb1352e JG |
769 | break; |
770 | ||
3fdbd1ce | 771 | case OVS_KEY_ATTR_IPV6: |
83d2b9ba JR |
772 | err = set_ipv6(skb, flow_key, nla_data(a), |
773 | get_mask(a, struct ovs_key_ipv6 *)); | |
3fdbd1ce AA |
774 | break; |
775 | ||
ccb1352e | 776 | case OVS_KEY_ATTR_TCP: |
83d2b9ba JR |
777 | err = set_tcp(skb, flow_key, nla_data(a), |
778 | get_mask(a, struct ovs_key_tcp *)); | |
ccb1352e JG |
779 | break; |
780 | ||
781 | case OVS_KEY_ATTR_UDP: | |
83d2b9ba JR |
782 | err = set_udp(skb, flow_key, nla_data(a), |
783 | get_mask(a, struct ovs_key_udp *)); | |
ccb1352e | 784 | break; |
a175a723 JS |
785 | |
786 | case OVS_KEY_ATTR_SCTP: | |
83d2b9ba JR |
787 | err = set_sctp(skb, flow_key, nla_data(a), |
788 | get_mask(a, struct ovs_key_sctp *)); | |
a175a723 | 789 | break; |
25cd9ba0 SH |
790 | |
791 | case OVS_KEY_ATTR_MPLS: | |
83d2b9ba JR |
792 | err = set_mpls(skb, flow_key, nla_data(a), get_mask(a, |
793 | __be32 *)); | |
25cd9ba0 | 794 | break; |
ccb1352e JG |
795 | } |
796 | ||
797 | return err; | |
798 | } | |
799 | ||
971427f3 AZ |
800 | static int execute_recirc(struct datapath *dp, struct sk_buff *skb, |
801 | struct sw_flow_key *key, | |
802 | const struct nlattr *a, int rem) | |
803 | { | |
804 | struct deferred_action *da; | |
971427f3 | 805 | |
fff06c36 PS |
806 | if (!is_flow_key_valid(key)) { |
807 | int err; | |
808 | ||
809 | err = ovs_flow_key_update(skb, key); | |
810 | if (err) | |
811 | return err; | |
812 | } | |
813 | BUG_ON(!is_flow_key_valid(key)); | |
971427f3 | 814 | |
941d8ebc | 815 | if (!nla_is_last(a, rem)) { |
971427f3 AZ |
816 | /* Recirc action is the not the last action |
817 | * of the action list, need to clone the skb. | |
818 | */ | |
819 | skb = skb_clone(skb, GFP_ATOMIC); | |
820 | ||
821 | /* Skip the recirc action when out of memory, but | |
822 | * continue on with the rest of the action list. | |
823 | */ | |
824 | if (!skb) | |
825 | return 0; | |
826 | } | |
827 | ||
828 | da = add_deferred_actions(skb, key, NULL); | |
829 | if (da) { | |
830 | da->pkt_key.recirc_id = nla_get_u32(a); | |
831 | } else { | |
832 | kfree_skb(skb); | |
833 | ||
834 | if (net_ratelimit()) | |
835 | pr_warn("%s: deferred action limit reached, drop recirc action\n", | |
836 | ovs_dp_name(dp)); | |
837 | } | |
838 | ||
839 | return 0; | |
840 | } | |
841 | ||
ccb1352e JG |
842 | /* Execute a list of actions against 'skb'. */ |
843 | static int do_execute_actions(struct datapath *dp, struct sk_buff *skb, | |
2ff3e4e4 | 844 | struct sw_flow_key *key, |
651887b0 | 845 | const struct nlattr *attr, int len) |
ccb1352e JG |
846 | { |
847 | /* Every output action needs a separate clone of 'skb', but the common | |
848 | * case is just a single output action, so that doing a clone and | |
849 | * then freeing the original skbuff is wasteful. So the following code | |
fff06c36 PS |
850 | * is slightly obscure just to avoid that. |
851 | */ | |
ccb1352e JG |
852 | int prev_port = -1; |
853 | const struct nlattr *a; | |
854 | int rem; | |
855 | ||
856 | for (a = attr, rem = len; rem > 0; | |
857 | a = nla_next(a, &rem)) { | |
858 | int err = 0; | |
859 | ||
738967b8 AZ |
860 | if (unlikely(prev_port != -1)) { |
861 | struct sk_buff *out_skb = skb_clone(skb, GFP_ATOMIC); | |
862 | ||
863 | if (out_skb) | |
864 | do_output(dp, out_skb, prev_port); | |
865 | ||
ccb1352e JG |
866 | prev_port = -1; |
867 | } | |
868 | ||
869 | switch (nla_type(a)) { | |
870 | case OVS_ACTION_ATTR_OUTPUT: | |
871 | prev_port = nla_get_u32(a); | |
872 | break; | |
873 | ||
874 | case OVS_ACTION_ATTR_USERSPACE: | |
2ff3e4e4 | 875 | output_userspace(dp, skb, key, a); |
ccb1352e JG |
876 | break; |
877 | ||
971427f3 AZ |
878 | case OVS_ACTION_ATTR_HASH: |
879 | execute_hash(skb, key, a); | |
880 | break; | |
881 | ||
25cd9ba0 | 882 | case OVS_ACTION_ATTR_PUSH_MPLS: |
fff06c36 | 883 | err = push_mpls(skb, key, nla_data(a)); |
25cd9ba0 SH |
884 | break; |
885 | ||
886 | case OVS_ACTION_ATTR_POP_MPLS: | |
fff06c36 | 887 | err = pop_mpls(skb, key, nla_get_be16(a)); |
25cd9ba0 SH |
888 | break; |
889 | ||
ccb1352e | 890 | case OVS_ACTION_ATTR_PUSH_VLAN: |
fff06c36 | 891 | err = push_vlan(skb, key, nla_data(a)); |
ccb1352e JG |
892 | break; |
893 | ||
894 | case OVS_ACTION_ATTR_POP_VLAN: | |
fff06c36 | 895 | err = pop_vlan(skb, key); |
ccb1352e JG |
896 | break; |
897 | ||
971427f3 AZ |
898 | case OVS_ACTION_ATTR_RECIRC: |
899 | err = execute_recirc(dp, skb, key, a, rem); | |
941d8ebc | 900 | if (nla_is_last(a, rem)) { |
971427f3 AZ |
901 | /* If this is the last action, the skb has |
902 | * been consumed or freed. | |
903 | * Return immediately. | |
904 | */ | |
905 | return err; | |
906 | } | |
907 | break; | |
908 | ||
ccb1352e | 909 | case OVS_ACTION_ATTR_SET: |
fff06c36 | 910 | err = execute_set_action(skb, key, nla_data(a)); |
ccb1352e JG |
911 | break; |
912 | ||
83d2b9ba JR |
913 | case OVS_ACTION_ATTR_SET_MASKED: |
914 | case OVS_ACTION_ATTR_SET_TO_MASKED: | |
915 | err = execute_masked_set_action(skb, key, nla_data(a)); | |
916 | break; | |
917 | ||
ccb1352e | 918 | case OVS_ACTION_ATTR_SAMPLE: |
2ff3e4e4 | 919 | err = sample(dp, skb, key, a); |
ccb1352e JG |
920 | break; |
921 | } | |
922 | ||
923 | if (unlikely(err)) { | |
924 | kfree_skb(skb); | |
925 | return err; | |
926 | } | |
927 | } | |
928 | ||
651887b0 | 929 | if (prev_port != -1) |
ccb1352e | 930 | do_output(dp, skb, prev_port); |
651887b0 | 931 | else |
ccb1352e JG |
932 | consume_skb(skb); |
933 | ||
934 | return 0; | |
935 | } | |
936 | ||
971427f3 AZ |
937 | static void process_deferred_actions(struct datapath *dp) |
938 | { | |
939 | struct action_fifo *fifo = this_cpu_ptr(action_fifos); | |
940 | ||
941 | /* Do not touch the FIFO in case there is no deferred actions. */ | |
942 | if (action_fifo_is_empty(fifo)) | |
943 | return; | |
944 | ||
945 | /* Finishing executing all deferred actions. */ | |
946 | do { | |
947 | struct deferred_action *da = action_fifo_get(fifo); | |
948 | struct sk_buff *skb = da->skb; | |
949 | struct sw_flow_key *key = &da->pkt_key; | |
950 | const struct nlattr *actions = da->actions; | |
951 | ||
952 | if (actions) | |
953 | do_execute_actions(dp, skb, key, actions, | |
954 | nla_len(actions)); | |
955 | else | |
956 | ovs_dp_process_packet(skb, key); | |
957 | } while (!action_fifo_is_empty(fifo)); | |
958 | ||
959 | /* Reset FIFO for the next packet. */ | |
960 | action_fifo_init(fifo); | |
961 | } | |
962 | ||
ccb1352e | 963 | /* Execute a list of actions against 'skb'. */ |
2ff3e4e4 | 964 | int ovs_execute_actions(struct datapath *dp, struct sk_buff *skb, |
12eb18f7 TG |
965 | const struct sw_flow_actions *acts, |
966 | struct sw_flow_key *key) | |
ccb1352e | 967 | { |
971427f3 | 968 | int level = this_cpu_read(exec_actions_level); |
971427f3 AZ |
969 | int err; |
970 | ||
971427f3 | 971 | this_cpu_inc(exec_actions_level); |
f0b128c1 | 972 | OVS_CB(skb)->egress_tun_info = NULL; |
971427f3 AZ |
973 | err = do_execute_actions(dp, skb, key, |
974 | acts->actions, acts->actions_len); | |
975 | ||
976 | if (!level) | |
977 | process_deferred_actions(dp); | |
978 | ||
979 | this_cpu_dec(exec_actions_level); | |
980 | return err; | |
981 | } | |
982 | ||
983 | int action_fifos_init(void) | |
984 | { | |
985 | action_fifos = alloc_percpu(struct action_fifo); | |
986 | if (!action_fifos) | |
987 | return -ENOMEM; | |
ccb1352e | 988 | |
971427f3 AZ |
989 | return 0; |
990 | } | |
991 | ||
992 | void action_fifos_exit(void) | |
993 | { | |
994 | free_percpu(action_fifos); | |
ccb1352e | 995 | } |