]>
Commit | Line | Data |
---|---|---|
064af421 | 1 | /* |
e23775f2 | 2 | * Copyright (c) 2007-2015 Nicira, Inc. |
a14bc59f | 3 | * |
a9a29d22 JG |
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 | |
064af421 BP |
17 | */ |
18 | ||
e9141eec PS |
19 | #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt |
20 | ||
064af421 BP |
21 | #include <linux/skbuff.h> |
22 | #include <linux/in.h> | |
23 | #include <linux/ip.h> | |
077257b8 | 24 | #include <linux/openvswitch.h> |
a94ebc39 | 25 | #include <linux/netfilter_ipv6.h> |
10f72e3d | 26 | #include <linux/sctp.h> |
064af421 BP |
27 | #include <linux/tcp.h> |
28 | #include <linux/udp.h> | |
29 | #include <linux/in6.h> | |
401eeb92 | 30 | #include <linux/if_arp.h> |
064af421 | 31 | #include <linux/if_vlan.h> |
a0fb56c1 | 32 | |
a94ebc39 | 33 | #include <net/dst.h> |
064af421 | 34 | #include <net/ip.h> |
bc7a5acd | 35 | #include <net/ipv6.h> |
064af421 | 36 | #include <net/checksum.h> |
530180fd | 37 | #include <net/dsfield.h> |
2baf0e0c | 38 | #include <net/mpls.h> |
10f72e3d | 39 | #include <net/sctp/checksum.h> |
f2459fe7 | 40 | |
f2459fe7 | 41 | #include "datapath.h" |
a94ebc39 | 42 | #include "conntrack.h" |
ccf43786 | 43 | #include "gso.h" |
f2459fe7 | 44 | #include "vport.h" |
064af421 | 45 | |
e74d4817 PS |
46 | static int do_execute_actions(struct datapath *dp, struct sk_buff *skb, |
47 | struct sw_flow_key *key, | |
48 | const struct nlattr *attr, int len); | |
49 | ||
2c8c4fb7 AZ |
50 | struct deferred_action { |
51 | struct sk_buff *skb; | |
52 | const struct nlattr *actions; | |
53 | ||
54 | /* Store pkt_key clone when creating deferred action. */ | |
55 | struct sw_flow_key pkt_key; | |
56 | }; | |
57 | ||
a94ebc39 JS |
58 | #define MAX_L2_LEN (VLAN_ETH_HLEN + 3 * MPLS_HLEN) |
59 | struct ovs_frag_data { | |
60 | unsigned long dst; | |
61 | struct vport *vport; | |
86c2eb45 | 62 | struct ovs_gso_cb cb; |
a94ebc39 JS |
63 | __be16 inner_protocol; |
64 | __u16 vlan_tci; | |
65 | __be16 vlan_proto; | |
66 | unsigned int l2_len; | |
67 | u8 l2_data[MAX_L2_LEN]; | |
68 | }; | |
69 | ||
70 | static DEFINE_PER_CPU(struct ovs_frag_data, ovs_frag_data_storage); | |
71 | ||
2c8c4fb7 AZ |
72 | #define DEFERRED_ACTION_FIFO_SIZE 10 |
73 | struct action_fifo { | |
74 | int head; | |
75 | int tail; | |
76 | /* Deferred action fifo queue storage. */ | |
77 | struct deferred_action fifo[DEFERRED_ACTION_FIFO_SIZE]; | |
78 | }; | |
79 | ||
80 | static struct action_fifo __percpu *action_fifos; | |
81 | #define EXEC_ACTIONS_LEVEL_LIMIT 4 /* limit used to detect packet | |
af465b67 PS |
82 | * looping by the network stack |
83 | */ | |
2c8c4fb7 AZ |
84 | static DEFINE_PER_CPU(int, exec_actions_level); |
85 | ||
86 | static void action_fifo_init(struct action_fifo *fifo) | |
87 | { | |
88 | fifo->head = 0; | |
89 | fifo->tail = 0; | |
90 | } | |
91 | ||
f1f60b85 | 92 | static bool action_fifo_is_empty(const struct action_fifo *fifo) |
2c8c4fb7 AZ |
93 | { |
94 | return (fifo->head == fifo->tail); | |
95 | } | |
96 | ||
e74d4817 | 97 | static struct deferred_action *action_fifo_get(struct action_fifo *fifo) |
2c8c4fb7 AZ |
98 | { |
99 | if (action_fifo_is_empty(fifo)) | |
100 | return NULL; | |
101 | ||
102 | return &fifo->fifo[fifo->tail++]; | |
103 | } | |
104 | ||
e74d4817 | 105 | static struct deferred_action *action_fifo_put(struct action_fifo *fifo) |
2c8c4fb7 AZ |
106 | { |
107 | if (fifo->head >= DEFERRED_ACTION_FIFO_SIZE - 1) | |
108 | return NULL; | |
109 | ||
110 | return &fifo->fifo[fifo->head++]; | |
111 | } | |
112 | ||
e74d4817 PS |
113 | /* Return queue entry if fifo is not full */ |
114 | static struct deferred_action *add_deferred_actions(struct sk_buff *skb, | |
f1f60b85 | 115 | const struct sw_flow_key *key, |
e74d4817 | 116 | const struct nlattr *attr) |
2c8c4fb7 AZ |
117 | { |
118 | struct action_fifo *fifo; | |
119 | struct deferred_action *da; | |
120 | ||
121 | fifo = this_cpu_ptr(action_fifos); | |
122 | da = action_fifo_put(fifo); | |
123 | if (da) { | |
124 | da->skb = skb; | |
125 | da->actions = attr; | |
e74d4817 | 126 | da->pkt_key = *key; |
2c8c4fb7 AZ |
127 | } |
128 | ||
e74d4817 | 129 | return da; |
e16138e2 PS |
130 | } |
131 | ||
e74d4817 | 132 | static void invalidate_flow_key(struct sw_flow_key *key) |
e16138e2 | 133 | { |
e74d4817 | 134 | key->eth.type = htons(0); |
e16138e2 PS |
135 | } |
136 | ||
f1f60b85 | 137 | static bool is_flow_key_valid(const struct sw_flow_key *key) |
e16138e2 | 138 | { |
e74d4817 | 139 | return !!key->eth.type; |
e16138e2 PS |
140 | } |
141 | ||
e74d4817 | 142 | static int push_mpls(struct sk_buff *skb, struct sw_flow_key *key, |
ccf43786 SH |
143 | const struct ovs_action_push_mpls *mpls) |
144 | { | |
145 | __be32 *new_mpls_lse; | |
146 | struct ethhdr *hdr; | |
147 | ||
2baf0e0c | 148 | /* Networking stack do not allow simultaneous Tunnel and MPLS GSO. */ |
8063e095 | 149 | if (skb->encapsulation) |
2baf0e0c PS |
150 | return -ENOTSUPP; |
151 | ||
ccf43786 SH |
152 | if (skb_cow_head(skb, MPLS_HLEN) < 0) |
153 | return -ENOMEM; | |
154 | ||
155 | skb_push(skb, MPLS_HLEN); | |
156 | memmove(skb_mac_header(skb) - MPLS_HLEN, skb_mac_header(skb), | |
157 | skb->mac_len); | |
158 | skb_reset_mac_header(skb); | |
159 | ||
2baf0e0c | 160 | new_mpls_lse = (__be32 *)skb_mpls_header(skb); |
ccf43786 SH |
161 | *new_mpls_lse = mpls->mpls_lse; |
162 | ||
163 | if (skb->ip_summed == CHECKSUM_COMPLETE) | |
164 | skb->csum = csum_add(skb->csum, csum_partial(new_mpls_lse, | |
165 | MPLS_HLEN, 0)); | |
166 | ||
167 | hdr = eth_hdr(skb); | |
168 | hdr->h_proto = mpls->mpls_ethertype; | |
169 | if (!ovs_skb_get_inner_protocol(skb)) | |
170 | ovs_skb_set_inner_protocol(skb, skb->protocol); | |
171 | skb->protocol = mpls->mpls_ethertype; | |
2baf0e0c | 172 | |
e74d4817 | 173 | invalidate_flow_key(key); |
ccf43786 SH |
174 | return 0; |
175 | } | |
176 | ||
e74d4817 PS |
177 | static int pop_mpls(struct sk_buff *skb, struct sw_flow_key *key, |
178 | const __be16 ethertype) | |
ccf43786 SH |
179 | { |
180 | struct ethhdr *hdr; | |
181 | int err; | |
182 | ||
5cce04b6 | 183 | err = skb_ensure_writable(skb, skb->mac_len + MPLS_HLEN); |
ccf43786 SH |
184 | if (unlikely(err)) |
185 | return err; | |
186 | ||
f021a62f | 187 | skb_postpull_rcsum(skb, skb_mpls_header(skb), MPLS_HLEN); |
ccf43786 SH |
188 | |
189 | memmove(skb_mac_header(skb) + MPLS_HLEN, skb_mac_header(skb), | |
190 | skb->mac_len); | |
191 | ||
192 | __skb_pull(skb, MPLS_HLEN); | |
193 | skb_reset_mac_header(skb); | |
194 | ||
2baf0e0c | 195 | /* skb_mpls_header() is used to locate the ethertype |
ccf43786 SH |
196 | * field correctly in the presence of VLAN tags. |
197 | */ | |
2baf0e0c | 198 | hdr = (struct ethhdr *)(skb_mpls_header(skb) - ETH_HLEN); |
ccf43786 SH |
199 | hdr->h_proto = ethertype; |
200 | if (eth_p_mpls(skb->protocol)) | |
201 | skb->protocol = ethertype; | |
2baf0e0c | 202 | |
e74d4817 | 203 | invalidate_flow_key(key); |
ccf43786 SH |
204 | return 0; |
205 | } | |
206 | ||
b940b3d7 JR |
207 | static int set_mpls(struct sk_buff *skb, struct sw_flow_key *flow_key, |
208 | const __be32 *mpls_lse, const __be32 *mask) | |
ccf43786 | 209 | { |
2baf0e0c | 210 | __be32 *stack; |
b940b3d7 | 211 | __be32 lse; |
ccf43786 SH |
212 | int err; |
213 | ||
5cce04b6 | 214 | err = skb_ensure_writable(skb, skb->mac_len + MPLS_HLEN); |
ccf43786 SH |
215 | if (unlikely(err)) |
216 | return err; | |
217 | ||
2baf0e0c | 218 | stack = (__be32 *)skb_mpls_header(skb); |
e281bb23 | 219 | lse = OVS_MASKED(*stack, *mpls_lse, *mask); |
ccf43786 | 220 | if (skb->ip_summed == CHECKSUM_COMPLETE) { |
b940b3d7 JR |
221 | __be32 diff[] = { ~(*stack), lse }; |
222 | ||
ccf43786 SH |
223 | skb->csum = ~csum_partial((char *)diff, sizeof(diff), |
224 | ~skb->csum); | |
225 | } | |
226 | ||
b940b3d7 JR |
227 | *stack = lse; |
228 | flow_key->mpls.top_lse = lse; | |
ccf43786 SH |
229 | return 0; |
230 | } | |
231 | ||
e74d4817 | 232 | static int pop_vlan(struct sk_buff *skb, struct sw_flow_key *key) |
064af421 | 233 | { |
d9065a90 | 234 | int err; |
10db8b20 | 235 | |
97894370 | 236 | err = skb_vlan_pop(skb); |
efd8a18e | 237 | if (skb_vlan_tag_present(skb)) |
97894370 TG |
238 | invalidate_flow_key(key); |
239 | else | |
e74d4817 | 240 | key->eth.tci = 0; |
97894370 | 241 | return err; |
d9065a90 PS |
242 | } |
243 | ||
e74d4817 PS |
244 | static int push_vlan(struct sk_buff *skb, struct sw_flow_key *key, |
245 | const struct ovs_action_push_vlan *vlan) | |
d9065a90 | 246 | { |
efd8a18e | 247 | if (skb_vlan_tag_present(skb)) |
e74d4817 | 248 | invalidate_flow_key(key); |
97894370 | 249 | else |
e74d4817 | 250 | key->eth.tci = vlan->vlan_tci; |
97894370 TG |
251 | return skb_vlan_push(skb, vlan->vlan_tpid, |
252 | ntohs(vlan->vlan_tci) & ~VLAN_TAG_PRESENT); | |
064af421 BP |
253 | } |
254 | ||
b940b3d7 JR |
255 | /* 'src' is already properly masked. */ |
256 | static void ether_addr_copy_masked(u8 *dst_, const u8 *src_, const u8 *mask_) | |
257 | { | |
258 | u16 *dst = (u16 *)dst_; | |
259 | const u16 *src = (const u16 *)src_; | |
260 | const u16 *mask = (const u16 *)mask_; | |
261 | ||
e281bb23 JS |
262 | OVS_SET_MASKED(dst[0], src[0], mask[0]); |
263 | OVS_SET_MASKED(dst[1], src[1], mask[1]); | |
264 | OVS_SET_MASKED(dst[2], src[2], mask[2]); | |
b940b3d7 JR |
265 | } |
266 | ||
267 | static int set_eth_addr(struct sk_buff *skb, struct sw_flow_key *flow_key, | |
268 | const struct ovs_key_ethernet *key, | |
269 | const struct ovs_key_ethernet *mask) | |
ca78c6b6 | 270 | { |
4edb9ae9 | 271 | int err; |
b940b3d7 | 272 | |
5cce04b6 | 273 | err = skb_ensure_writable(skb, ETH_HLEN); |
4edb9ae9 PS |
274 | if (unlikely(err)) |
275 | return err; | |
276 | ||
237c4f2a | 277 | skb_postpull_rcsum(skb, eth_hdr(skb), ETH_ALEN * 2); |
3cfede14 | 278 | |
b940b3d7 JR |
279 | ether_addr_copy_masked(eth_hdr(skb)->h_source, key->eth_src, |
280 | mask->eth_src); | |
281 | ether_addr_copy_masked(eth_hdr(skb)->h_dest, key->eth_dst, | |
282 | mask->eth_dst); | |
4edb9ae9 | 283 | |
237c4f2a | 284 | ovs_skb_postpush_rcsum(skb, eth_hdr(skb), ETH_ALEN * 2); |
3cfede14 | 285 | |
b940b3d7 JR |
286 | ether_addr_copy(flow_key->eth.src, eth_hdr(skb)->h_source); |
287 | ether_addr_copy(flow_key->eth.dst, eth_hdr(skb)->h_dest); | |
4edb9ae9 | 288 | return 0; |
ca78c6b6 BP |
289 | } |
290 | ||
efdb0c9f GG |
291 | static void update_ip_l4_checksum(struct sk_buff *skb, struct iphdr *nh, |
292 | __be32 addr, __be32 new_addr) | |
ca78c6b6 BP |
293 | { |
294 | int transport_len = skb->len - skb_transport_offset(skb); | |
4edb9ae9 | 295 | |
efdb0c9f GG |
296 | if (nh->frag_off & htons(IP_OFFSET)) |
297 | return; | |
298 | ||
4edb9ae9 | 299 | if (nh->protocol == IPPROTO_TCP) { |
ca78c6b6 | 300 | if (likely(transport_len >= sizeof(struct tcphdr))) |
4edb9ae9 | 301 | inet_proto_csum_replace4(&tcp_hdr(skb)->check, skb, |
8063e095 | 302 | addr, new_addr, true); |
4edb9ae9 | 303 | } else if (nh->protocol == IPPROTO_UDP) { |
55ce87bc JG |
304 | if (likely(transport_len >= sizeof(struct udphdr))) { |
305 | struct udphdr *uh = udp_hdr(skb); | |
306 | ||
237c4f2a | 307 | if (uh->check || skb->ip_summed == CHECKSUM_PARTIAL) { |
55ce87bc | 308 | inet_proto_csum_replace4(&uh->check, skb, |
8063e095 | 309 | addr, new_addr, true); |
55ce87bc JG |
310 | if (!uh->check) |
311 | uh->check = CSUM_MANGLED_0; | |
312 | } | |
313 | } | |
ca78c6b6 | 314 | } |
4edb9ae9 | 315 | |
efdb0c9f GG |
316 | } |
317 | ||
318 | static void set_ip_addr(struct sk_buff *skb, struct iphdr *nh, | |
319 | __be32 *addr, __be32 new_addr) | |
320 | { | |
321 | update_ip_l4_checksum(skb, nh, *addr, new_addr); | |
4edb9ae9 | 322 | csum_replace4(&nh->check, *addr, new_addr); |
e2f3178f | 323 | skb_clear_hash(skb); |
4edb9ae9 | 324 | *addr = new_addr; |
ca78c6b6 BP |
325 | } |
326 | ||
bc7a5acd AA |
327 | static void update_ipv6_checksum(struct sk_buff *skb, u8 l4_proto, |
328 | __be32 addr[4], const __be32 new_addr[4]) | |
329 | { | |
330 | int transport_len = skb->len - skb_transport_offset(skb); | |
331 | ||
00894212 | 332 | if (l4_proto == NEXTHDR_TCP) { |
bc7a5acd AA |
333 | if (likely(transport_len >= sizeof(struct tcphdr))) |
334 | inet_proto_csum_replace16(&tcp_hdr(skb)->check, skb, | |
8063e095 | 335 | addr, new_addr, true); |
00894212 | 336 | } else if (l4_proto == NEXTHDR_UDP) { |
bc7a5acd AA |
337 | if (likely(transport_len >= sizeof(struct udphdr))) { |
338 | struct udphdr *uh = udp_hdr(skb); | |
339 | ||
237c4f2a | 340 | if (uh->check || skb->ip_summed == CHECKSUM_PARTIAL) { |
bc7a5acd | 341 | inet_proto_csum_replace16(&uh->check, skb, |
8063e095 | 342 | addr, new_addr, true); |
bc7a5acd AA |
343 | if (!uh->check) |
344 | uh->check = CSUM_MANGLED_0; | |
345 | } | |
346 | } | |
00894212 JG |
347 | } else if (l4_proto == NEXTHDR_ICMP) { |
348 | if (likely(transport_len >= sizeof(struct icmp6hdr))) | |
349 | inet_proto_csum_replace16(&icmp6_hdr(skb)->icmp6_cksum, | |
8063e095 | 350 | skb, addr, new_addr, true); |
bc7a5acd AA |
351 | } |
352 | } | |
353 | ||
b940b3d7 JR |
354 | static void mask_ipv6_addr(const __be32 old[4], const __be32 addr[4], |
355 | const __be32 mask[4], __be32 masked[4]) | |
356 | { | |
e281bb23 JS |
357 | masked[0] = OVS_MASKED(old[0], addr[0], mask[0]); |
358 | masked[1] = OVS_MASKED(old[1], addr[1], mask[1]); | |
359 | masked[2] = OVS_MASKED(old[2], addr[2], mask[2]); | |
360 | masked[3] = OVS_MASKED(old[3], addr[3], mask[3]); | |
b940b3d7 JR |
361 | } |
362 | ||
bc7a5acd AA |
363 | static void set_ipv6_addr(struct sk_buff *skb, u8 l4_proto, |
364 | __be32 addr[4], const __be32 new_addr[4], | |
365 | bool recalculate_csum) | |
366 | { | |
51cf5e71 | 367 | if (likely(recalculate_csum)) |
bc7a5acd AA |
368 | update_ipv6_checksum(skb, l4_proto, addr, new_addr); |
369 | ||
e2f3178f | 370 | skb_clear_hash(skb); |
bc7a5acd AA |
371 | memcpy(addr, new_addr, sizeof(__be32[4])); |
372 | } | |
373 | ||
b940b3d7 | 374 | static void set_ipv6_fl(struct ipv6hdr *nh, u32 fl, u32 mask) |
bc7a5acd | 375 | { |
b940b3d7 | 376 | /* Bits 21-24 are always unmasked, so this retains their values. */ |
e281bb23 JS |
377 | OVS_SET_MASKED(nh->flow_lbl[0], (u8)(fl >> 16), (u8)(mask >> 16)); |
378 | OVS_SET_MASKED(nh->flow_lbl[1], (u8)(fl >> 8), (u8)(mask >> 8)); | |
379 | OVS_SET_MASKED(nh->flow_lbl[2], (u8)fl, (u8)mask); | |
bc7a5acd AA |
380 | } |
381 | ||
b940b3d7 JR |
382 | static void set_ip_ttl(struct sk_buff *skb, struct iphdr *nh, u8 new_ttl, |
383 | u8 mask) | |
bc7a5acd | 384 | { |
e281bb23 | 385 | new_ttl = OVS_MASKED(nh->ttl, new_ttl, mask); |
bc7a5acd | 386 | |
a61680c6 JP |
387 | csum_replace2(&nh->check, htons(nh->ttl << 8), htons(new_ttl << 8)); |
388 | nh->ttl = new_ttl; | |
389 | } | |
390 | ||
b940b3d7 JR |
391 | static int set_ipv4(struct sk_buff *skb, struct sw_flow_key *flow_key, |
392 | const struct ovs_key_ipv4 *key, | |
393 | const struct ovs_key_ipv4 *mask) | |
064af421 | 394 | { |
ca78c6b6 | 395 | struct iphdr *nh; |
b940b3d7 | 396 | __be32 new_addr; |
10db8b20 | 397 | int err; |
ca78c6b6 | 398 | |
5cce04b6 TG |
399 | err = skb_ensure_writable(skb, skb_network_offset(skb) + |
400 | sizeof(struct iphdr)); | |
10db8b20 JG |
401 | if (unlikely(err)) |
402 | return err; | |
ca78c6b6 BP |
403 | |
404 | nh = ip_hdr(skb); | |
ca78c6b6 | 405 | |
b940b3d7 JR |
406 | /* Setting an IP addresses is typically only a side effect of |
407 | * matching on them in the current userspace implementation, so it | |
408 | * makes sense to check if the value actually changed. | |
409 | */ | |
410 | if (mask->ipv4_src) { | |
e281bb23 | 411 | new_addr = OVS_MASKED(nh->saddr, key->ipv4_src, mask->ipv4_src); |
ca78c6b6 | 412 | |
b940b3d7 JR |
413 | if (unlikely(new_addr != nh->saddr)) { |
414 | set_ip_addr(skb, nh, &nh->saddr, new_addr); | |
415 | flow_key->ipv4.addr.src = new_addr; | |
416 | } | |
e16138e2 | 417 | } |
b940b3d7 | 418 | if (mask->ipv4_dst) { |
e281bb23 | 419 | new_addr = OVS_MASKED(nh->daddr, key->ipv4_dst, mask->ipv4_dst); |
a4a26436 | 420 | |
b940b3d7 JR |
421 | if (unlikely(new_addr != nh->daddr)) { |
422 | set_ip_addr(skb, nh, &nh->daddr, new_addr); | |
423 | flow_key->ipv4.addr.dst = new_addr; | |
424 | } | |
e16138e2 | 425 | } |
b940b3d7 JR |
426 | if (mask->ipv4_tos) { |
427 | ipv4_change_dsfield(nh, ~mask->ipv4_tos, key->ipv4_tos); | |
428 | flow_key->ip.tos = nh->tos; | |
429 | } | |
430 | if (mask->ipv4_ttl) { | |
431 | set_ip_ttl(skb, nh, key->ipv4_ttl, mask->ipv4_ttl); | |
432 | flow_key->ip.ttl = nh->ttl; | |
e16138e2 | 433 | } |
a61680c6 | 434 | |
10db8b20 | 435 | return 0; |
064af421 BP |
436 | } |
437 | ||
b940b3d7 JR |
438 | static bool is_ipv6_mask_nonzero(const __be32 addr[4]) |
439 | { | |
440 | return !!(addr[0] | addr[1] | addr[2] | addr[3]); | |
441 | } | |
442 | ||
443 | static int set_ipv6(struct sk_buff *skb, struct sw_flow_key *flow_key, | |
444 | const struct ovs_key_ipv6 *key, | |
445 | const struct ovs_key_ipv6 *mask) | |
bc7a5acd AA |
446 | { |
447 | struct ipv6hdr *nh; | |
448 | int err; | |
bc7a5acd | 449 | |
5cce04b6 TG |
450 | err = skb_ensure_writable(skb, skb_network_offset(skb) + |
451 | sizeof(struct ipv6hdr)); | |
bc7a5acd AA |
452 | if (unlikely(err)) |
453 | return err; | |
454 | ||
455 | nh = ipv6_hdr(skb); | |
bc7a5acd | 456 | |
b940b3d7 JR |
457 | /* Setting an IP addresses is typically only a side effect of |
458 | * matching on them in the current userspace implementation, so it | |
459 | * makes sense to check if the value actually changed. | |
460 | */ | |
461 | if (is_ipv6_mask_nonzero(mask->ipv6_src)) { | |
462 | __be32 *saddr = (__be32 *)&nh->saddr; | |
463 | __be32 masked[4]; | |
464 | ||
465 | mask_ipv6_addr(saddr, key->ipv6_src, mask->ipv6_src, masked); | |
466 | ||
467 | if (unlikely(memcmp(saddr, masked, sizeof(masked)))) { | |
49a8eef8 | 468 | set_ipv6_addr(skb, flow_key->ip.proto, saddr, masked, |
b940b3d7 JR |
469 | true); |
470 | memcpy(&flow_key->ipv6.addr.src, masked, | |
471 | sizeof(flow_key->ipv6.addr.src)); | |
472 | } | |
473 | } | |
474 | if (is_ipv6_mask_nonzero(mask->ipv6_dst)) { | |
bc7a5acd | 475 | unsigned int offset = 0; |
8abaa53c | 476 | int flags = IP6_FH_F_SKIP_RH; |
bc7a5acd | 477 | bool recalc_csum = true; |
b940b3d7 JR |
478 | __be32 *daddr = (__be32 *)&nh->daddr; |
479 | __be32 masked[4]; | |
480 | ||
481 | mask_ipv6_addr(daddr, key->ipv6_dst, mask->ipv6_dst, masked); | |
482 | ||
483 | if (unlikely(memcmp(daddr, masked, sizeof(masked)))) { | |
484 | if (ipv6_ext_hdr(nh->nexthdr)) | |
485 | recalc_csum = (ipv6_find_hdr(skb, &offset, | |
486 | NEXTHDR_ROUTING, | |
487 | NULL, &flags) | |
488 | != NEXTHDR_ROUTING); | |
489 | ||
49a8eef8 | 490 | set_ipv6_addr(skb, flow_key->ip.proto, daddr, masked, |
b940b3d7 JR |
491 | recalc_csum); |
492 | memcpy(&flow_key->ipv6.addr.dst, masked, | |
493 | sizeof(flow_key->ipv6.addr.dst)); | |
494 | } | |
495 | } | |
496 | if (mask->ipv6_tclass) { | |
497 | ipv6_change_dsfield(nh, ~mask->ipv6_tclass, key->ipv6_tclass); | |
498 | flow_key->ip.tos = ipv6_get_dsfield(nh); | |
499 | } | |
500 | if (mask->ipv6_label) { | |
501 | set_ipv6_fl(nh, ntohl(key->ipv6_label), | |
502 | ntohl(mask->ipv6_label)); | |
503 | flow_key->ipv6.label = | |
504 | *(__be32 *)nh & htonl(IPV6_FLOWINFO_FLOWLABEL); | |
505 | } | |
506 | if (mask->ipv6_hlimit) { | |
e281bb23 JS |
507 | OVS_SET_MASKED(nh->hop_limit, key->ipv6_hlimit, |
508 | mask->ipv6_hlimit); | |
b940b3d7 | 509 | flow_key->ip.ttl = nh->hop_limit; |
bc7a5acd | 510 | } |
bc7a5acd AA |
511 | return 0; |
512 | } | |
513 | ||
5cce04b6 | 514 | /* Must follow skb_ensure_writable() since that can move the skb data. */ |
4edb9ae9 | 515 | static void set_tp_port(struct sk_buff *skb, __be16 *port, |
b940b3d7 | 516 | __be16 new_port, __sum16 *check) |
959a2ecd | 517 | { |
8063e095 | 518 | inet_proto_csum_replace2(check, skb, *port, new_port, false); |
4edb9ae9 | 519 | *port = new_port; |
55ce87bc JG |
520 | } |
521 | ||
b940b3d7 JR |
522 | static int set_udp(struct sk_buff *skb, struct sw_flow_key *flow_key, |
523 | const struct ovs_key_udp *key, | |
524 | const struct ovs_key_udp *mask) | |
4edb9ae9 PS |
525 | { |
526 | struct udphdr *uh; | |
b940b3d7 | 527 | __be16 src, dst; |
4edb9ae9 | 528 | int err; |
10db8b20 | 529 | |
5cce04b6 TG |
530 | err = skb_ensure_writable(skb, skb_transport_offset(skb) + |
531 | sizeof(struct udphdr)); | |
10db8b20 JG |
532 | if (unlikely(err)) |
533 | return err; | |
534 | ||
4edb9ae9 | 535 | uh = udp_hdr(skb); |
b940b3d7 | 536 | /* Either of the masks is non-zero, so do not bother checking them. */ |
e281bb23 JS |
537 | src = OVS_MASKED(uh->source, key->udp_src, mask->udp_src); |
538 | dst = OVS_MASKED(uh->dest, key->udp_dst, mask->udp_dst); | |
4edb9ae9 | 539 | |
b940b3d7 JR |
540 | if (uh->check && skb->ip_summed != CHECKSUM_PARTIAL) { |
541 | if (likely(src != uh->source)) { | |
542 | set_tp_port(skb, &uh->source, src, &uh->check); | |
543 | flow_key->tp.src = src; | |
544 | } | |
545 | if (likely(dst != uh->dest)) { | |
546 | set_tp_port(skb, &uh->dest, dst, &uh->check); | |
547 | flow_key->tp.dst = dst; | |
548 | } | |
549 | ||
550 | if (unlikely(!uh->check)) | |
551 | uh->check = CSUM_MANGLED_0; | |
552 | } else { | |
553 | uh->source = src; | |
554 | uh->dest = dst; | |
555 | flow_key->tp.src = src; | |
556 | flow_key->tp.dst = dst; | |
e16138e2 | 557 | } |
10db8b20 | 558 | |
b940b3d7 JR |
559 | skb_clear_hash(skb); |
560 | ||
10db8b20 | 561 | return 0; |
959a2ecd JP |
562 | } |
563 | ||
b940b3d7 JR |
564 | static int set_tcp(struct sk_buff *skb, struct sw_flow_key *flow_key, |
565 | const struct ovs_key_tcp *key, | |
566 | const struct ovs_key_tcp *mask) | |
064af421 | 567 | { |
4edb9ae9 | 568 | struct tcphdr *th; |
b940b3d7 | 569 | __be16 src, dst; |
10db8b20 | 570 | int err; |
064af421 | 571 | |
5cce04b6 TG |
572 | err = skb_ensure_writable(skb, skb_transport_offset(skb) + |
573 | sizeof(struct tcphdr)); | |
10db8b20 JG |
574 | if (unlikely(err)) |
575 | return err; | |
ca78c6b6 | 576 | |
4edb9ae9 | 577 | th = tcp_hdr(skb); |
e281bb23 | 578 | src = OVS_MASKED(th->source, key->tcp_src, mask->tcp_src); |
b940b3d7 JR |
579 | if (likely(src != th->source)) { |
580 | set_tp_port(skb, &th->source, src, &th->check); | |
581 | flow_key->tp.src = src; | |
582 | } | |
e281bb23 | 583 | dst = OVS_MASKED(th->dest, key->tcp_dst, mask->tcp_dst); |
b940b3d7 JR |
584 | if (likely(dst != th->dest)) { |
585 | set_tp_port(skb, &th->dest, dst, &th->check); | |
586 | flow_key->tp.dst = dst; | |
e16138e2 | 587 | } |
b940b3d7 | 588 | skb_clear_hash(skb); |
ca78c6b6 | 589 | |
10db8b20 | 590 | return 0; |
064af421 BP |
591 | } |
592 | ||
b940b3d7 JR |
593 | static int set_sctp(struct sk_buff *skb, struct sw_flow_key *flow_key, |
594 | const struct ovs_key_sctp *key, | |
595 | const struct ovs_key_sctp *mask) | |
10f72e3d | 596 | { |
b940b3d7 | 597 | unsigned int sctphoff = skb_transport_offset(skb); |
10f72e3d | 598 | struct sctphdr *sh; |
b940b3d7 | 599 | __le32 old_correct_csum, new_csum, old_csum; |
10f72e3d | 600 | int err; |
10f72e3d | 601 | |
5cce04b6 | 602 | err = skb_ensure_writable(skb, sctphoff + sizeof(struct sctphdr)); |
10f72e3d JS |
603 | if (unlikely(err)) |
604 | return err; | |
605 | ||
606 | sh = sctp_hdr(skb); | |
b940b3d7 JR |
607 | old_csum = sh->checksum; |
608 | old_correct_csum = sctp_compute_cksum(skb, sctphoff); | |
10f72e3d | 609 | |
e281bb23 JS |
610 | sh->source = OVS_MASKED(sh->source, key->sctp_src, mask->sctp_src); |
611 | sh->dest = OVS_MASKED(sh->dest, key->sctp_dst, mask->sctp_dst); | |
10f72e3d | 612 | |
b940b3d7 | 613 | new_csum = sctp_compute_cksum(skb, sctphoff); |
10f72e3d | 614 | |
b940b3d7 JR |
615 | /* Carry any checksum errors through. */ |
616 | sh->checksum = old_csum ^ old_correct_csum ^ new_csum; | |
10f72e3d | 617 | |
b940b3d7 JR |
618 | skb_clear_hash(skb); |
619 | flow_key->tp.src = sh->source; | |
620 | flow_key->tp.dst = sh->dest; | |
10f72e3d JS |
621 | |
622 | return 0; | |
623 | } | |
624 | ||
a94ebc39 JS |
625 | static int ovs_vport_output(OVS_VPORT_OUTPUT_PARAMS) |
626 | { | |
627 | struct ovs_frag_data *data = get_pcpu_ptr(ovs_frag_data_storage); | |
628 | struct vport *vport = data->vport; | |
629 | ||
630 | if (skb_cow_head(skb, data->l2_len) < 0) { | |
631 | kfree_skb(skb); | |
632 | return -ENOMEM; | |
633 | } | |
634 | ||
635 | __skb_dst_copy(skb, data->dst); | |
86c2eb45 | 636 | *OVS_GSO_CB(skb) = data->cb; |
a94ebc39 JS |
637 | ovs_skb_set_inner_protocol(skb, data->inner_protocol); |
638 | skb->vlan_tci = data->vlan_tci; | |
639 | skb->vlan_proto = data->vlan_proto; | |
640 | ||
641 | /* Reconstruct the MAC header. */ | |
642 | skb_push(skb, data->l2_len); | |
643 | memcpy(skb->data, &data->l2_data, data->l2_len); | |
644 | ovs_skb_postpush_rcsum(skb, skb->data, data->l2_len); | |
645 | skb_reset_mac_header(skb); | |
646 | ||
647 | ovs_vport_send(vport, skb); | |
648 | return 0; | |
649 | } | |
650 | ||
651 | static unsigned int | |
652 | ovs_dst_get_mtu(const struct dst_entry *dst) | |
653 | { | |
654 | return dst->dev->mtu; | |
655 | } | |
656 | ||
657 | static struct dst_ops ovs_dst_ops = { | |
658 | .family = AF_UNSPEC, | |
659 | .mtu = ovs_dst_get_mtu, | |
660 | }; | |
661 | ||
662 | /* prepare_frag() is called once per (larger-than-MTU) frame; its inverse is | |
663 | * ovs_vport_output(), which is called once per fragmented packet. | |
664 | */ | |
665 | static void prepare_frag(struct vport *vport, struct sk_buff *skb) | |
666 | { | |
667 | unsigned int hlen = skb_network_offset(skb); | |
668 | struct ovs_frag_data *data; | |
669 | ||
670 | data = get_pcpu_ptr(ovs_frag_data_storage); | |
671 | data->dst = (unsigned long) skb_dst(skb); | |
672 | data->vport = vport; | |
86c2eb45 | 673 | data->cb = *OVS_GSO_CB(skb); |
a94ebc39 JS |
674 | data->inner_protocol = ovs_skb_get_inner_protocol(skb); |
675 | data->vlan_tci = skb->vlan_tci; | |
676 | data->vlan_proto = skb->vlan_proto; | |
677 | data->l2_len = hlen; | |
678 | memcpy(&data->l2_data, skb->data, hlen); | |
679 | ||
680 | memset(IPCB(skb), 0, sizeof(struct inet_skb_parm)); | |
681 | skb_pull(skb, hlen); | |
682 | } | |
683 | ||
684 | static void ovs_fragment(struct vport *vport, struct sk_buff *skb, u16 mru, | |
685 | __be16 ethertype) | |
686 | { | |
687 | if (skb_network_offset(skb) > MAX_L2_LEN) { | |
688 | OVS_NLERR(1, "L2 header too long to fragment"); | |
c05e2094 | 689 | goto err; |
a94ebc39 JS |
690 | } |
691 | ||
692 | if (ethertype == htons(ETH_P_IP)) { | |
693 | struct dst_entry ovs_dst; | |
694 | unsigned long orig_dst; | |
695 | ||
696 | prepare_frag(vport, skb); | |
697 | dst_init(&ovs_dst, &ovs_dst_ops, NULL, 1, | |
698 | DST_OBSOLETE_NONE, DST_NOCOUNT); | |
699 | ovs_dst.dev = vport->dev; | |
700 | ||
701 | orig_dst = (unsigned long) skb_dst(skb); | |
702 | skb_dst_set_noref(skb, &ovs_dst); | |
703 | IPCB(skb)->frag_max_size = mru; | |
704 | ||
705 | ip_do_fragment(skb->sk, skb, ovs_vport_output); | |
706 | refdst_drop(orig_dst); | |
707 | } else if (ethertype == htons(ETH_P_IPV6)) { | |
708 | const struct nf_ipv6_ops *v6ops = nf_get_ipv6_ops(); | |
709 | unsigned long orig_dst; | |
710 | struct rt6_info ovs_rt; | |
711 | ||
712 | if (!v6ops) { | |
c05e2094 | 713 | goto err; |
a94ebc39 JS |
714 | } |
715 | ||
716 | prepare_frag(vport, skb); | |
717 | memset(&ovs_rt, 0, sizeof(ovs_rt)); | |
718 | dst_init(&ovs_rt.dst, &ovs_dst_ops, NULL, 1, | |
719 | DST_OBSOLETE_NONE, DST_NOCOUNT); | |
720 | ovs_rt.dst.dev = vport->dev; | |
721 | ||
722 | orig_dst = (unsigned long) skb_dst(skb); | |
723 | skb_dst_set_noref(skb, &ovs_rt.dst); | |
724 | IP6CB(skb)->frag_max_size = mru; | |
725 | ||
726 | v6ops->fragment(skb->sk, skb, ovs_vport_output); | |
727 | refdst_drop(orig_dst); | |
728 | } else { | |
729 | WARN_ONCE(1, "Failed fragment ->%s: eth=%04x, MRU=%d, MTU=%d.", | |
730 | ovs_vport_name(vport), ntohs(ethertype), mru, | |
731 | vport->dev->mtu); | |
c05e2094 | 732 | goto err; |
a94ebc39 | 733 | } |
c05e2094 JS |
734 | |
735 | return; | |
736 | err: | |
737 | kfree_skb(skb); | |
a94ebc39 | 738 | } |
a94ebc39 JS |
739 | |
740 | static void do_output(struct datapath *dp, struct sk_buff *skb, int out_port, | |
741 | struct sw_flow_key *key) | |
064af421 | 742 | { |
fe90efd9 | 743 | struct vport *vport = ovs_vport_rcu(dp, out_port); |
064af421 | 744 | |
a94ebc39 JS |
745 | if (likely(vport)) { |
746 | u16 mru = OVS_CB(skb)->mru; | |
4c7804f1 WT |
747 | u32 cutlen = OVS_CB(skb)->cutlen; |
748 | ||
749 | if (unlikely(cutlen > 0)) { | |
750 | if (skb->len - cutlen > ETH_HLEN) | |
751 | pskb_trim(skb, skb->len - cutlen); | |
752 | else | |
753 | pskb_trim(skb, ETH_HLEN); | |
754 | } | |
a94ebc39 JS |
755 | |
756 | if (likely(!mru || (skb->len <= mru + ETH_HLEN))) { | |
757 | ovs_vport_send(vport, skb); | |
758 | } else if (mru <= vport->dev->mtu) { | |
759 | __be16 ethertype = key->eth.type; | |
760 | ||
761 | if (!is_flow_key_valid(key)) { | |
762 | if (eth_p_mpls(skb->protocol)) | |
763 | ethertype = ovs_skb_get_inner_protocol(skb); | |
764 | else | |
765 | ethertype = vlan_get_protocol(skb); | |
766 | } | |
767 | ||
768 | ovs_fragment(vport, skb, mru, ethertype); | |
769 | } else { | |
770 | OVS_NLERR(true, "Cannot fragment IP frames"); | |
771 | kfree_skb(skb); | |
772 | } | |
773 | } else { | |
f15c8639 | 774 | kfree_skb(skb); |
a94ebc39 | 775 | } |
064af421 | 776 | } |
98403001 | 777 | static int output_userspace(struct datapath *dp, struct sk_buff *skb, |
0e469d3b | 778 | struct sw_flow_key *key, const struct nlattr *attr, |
4c7804f1 WT |
779 | const struct nlattr *actions, int actions_len, |
780 | uint32_t cutlen) | |
064af421 | 781 | { |
e23775f2 | 782 | struct ip_tunnel_info info; |
856081f6 | 783 | struct dp_upcall_info upcall; |
98403001 BP |
784 | const struct nlattr *a; |
785 | int rem; | |
856081f6 | 786 | |
0e469d3b | 787 | memset(&upcall, 0, sizeof(upcall)); |
df2c07f4 | 788 | upcall.cmd = OVS_PACKET_CMD_ACTION; |
a94ebc39 | 789 | upcall.mru = OVS_CB(skb)->mru; |
98403001 BP |
790 | |
791 | for (a = nla_data(attr), rem = nla_len(attr); rem > 0; | |
792 | a = nla_next(a, &rem)) { | |
793 | switch (nla_type(a)) { | |
794 | case OVS_USERSPACE_ATTR_USERDATA: | |
795 | upcall.userdata = a; | |
796 | break; | |
797 | ||
798 | case OVS_USERSPACE_ATTR_PID: | |
28aea917 | 799 | upcall.portid = nla_get_u32(a); |
98403001 | 800 | break; |
8b7ea2d4 WZ |
801 | |
802 | case OVS_USERSPACE_ATTR_EGRESS_TUN_PORT: { | |
803 | /* Get out tunnel info. */ | |
804 | struct vport *vport; | |
805 | ||
806 | vport = ovs_vport_rcu(dp, nla_get_u32(a)); | |
807 | if (vport) { | |
808 | int err; | |
809 | ||
e23775f2 | 810 | upcall.egress_tun_info = &info; |
8b7ea2d4 | 811 | err = ovs_vport_get_egress_tun_info(vport, skb, |
e23775f2 PS |
812 | &upcall); |
813 | if (err) | |
814 | upcall.egress_tun_info = NULL; | |
8b7ea2d4 | 815 | } |
e23775f2 | 816 | |
8b7ea2d4 | 817 | break; |
98403001 | 818 | } |
8b7ea2d4 | 819 | |
0e469d3b NM |
820 | case OVS_USERSPACE_ATTR_ACTIONS: { |
821 | /* Include actions. */ | |
822 | upcall.actions = actions; | |
823 | upcall.actions_len = actions_len; | |
824 | break; | |
825 | } | |
826 | ||
8b7ea2d4 | 827 | } /* End of switch. */ |
98403001 BP |
828 | } |
829 | ||
4c7804f1 | 830 | return ovs_dp_upcall(dp, skb, key, &upcall, cutlen); |
064af421 BP |
831 | } |
832 | ||
6ff686f2 | 833 | static int sample(struct datapath *dp, struct sk_buff *skb, |
0e469d3b NM |
834 | struct sw_flow_key *key, const struct nlattr *attr, |
835 | const struct nlattr *actions, int actions_len) | |
6ff686f2 PS |
836 | { |
837 | const struct nlattr *acts_list = NULL; | |
838 | const struct nlattr *a; | |
839 | int rem; | |
4c7804f1 | 840 | u32 cutlen = 0; |
6ff686f2 PS |
841 | |
842 | for (a = nla_data(attr), rem = nla_len(attr); rem > 0; | |
843 | a = nla_next(a, &rem)) { | |
c02c4967 WZ |
844 | u32 probability; |
845 | ||
6ff686f2 PS |
846 | switch (nla_type(a)) { |
847 | case OVS_SAMPLE_ATTR_PROBABILITY: | |
c02c4967 WZ |
848 | probability = nla_get_u32(a); |
849 | if (!probability || prandom_u32() > probability) | |
6ff686f2 PS |
850 | return 0; |
851 | break; | |
852 | ||
853 | case OVS_SAMPLE_ATTR_ACTIONS: | |
854 | acts_list = a; | |
855 | break; | |
856 | } | |
857 | } | |
858 | ||
fbf4f74d SH |
859 | rem = nla_len(acts_list); |
860 | a = nla_data(acts_list); | |
861 | ||
d7ff93d7 AZ |
862 | /* Actions list is empty, do nothing */ |
863 | if (unlikely(!rem)) | |
864 | return 0; | |
e16138e2 | 865 | |
d7ff93d7 | 866 | /* The only known usage of sample action is having a single user-space |
4c7804f1 | 867 | * action, or having a truncate action followed by a single user-space |
d7ff93d7 AZ |
868 | * action. Treat this usage as a special case. |
869 | * The output_userspace() should clone the skb to be sent to the | |
e74d4817 PS |
870 | * user space. This skb will be consumed by its caller. |
871 | */ | |
4c7804f1 WT |
872 | if (unlikely(nla_type(a) == OVS_ACTION_ATTR_TRUNC)) { |
873 | struct ovs_action_trunc *trunc = nla_data(a); | |
874 | ||
875 | if (skb->len > trunc->max_len) | |
876 | cutlen = skb->len - trunc->max_len; | |
877 | ||
878 | a = nla_next(a, &rem); | |
879 | } | |
880 | ||
d7ff93d7 | 881 | if (likely(nla_type(a) == OVS_ACTION_ATTR_USERSPACE && |
684b5f5d | 882 | nla_is_last(a, rem))) |
4c7804f1 WT |
883 | return output_userspace(dp, skb, key, a, actions, |
884 | actions_len, cutlen); | |
d7ff93d7 AZ |
885 | |
886 | skb = skb_clone(skb, GFP_ATOMIC); | |
887 | if (!skb) | |
888 | /* Skip the sample action when out of memory. */ | |
889 | return 0; | |
890 | ||
e74d4817 | 891 | if (!add_deferred_actions(skb, key, a)) { |
2c8c4fb7 AZ |
892 | if (net_ratelimit()) |
893 | pr_warn("%s: deferred actions limit reached, dropping sample action\n", | |
894 | ovs_dp_name(dp)); | |
fbf4f74d | 895 | |
2c8c4fb7 AZ |
896 | kfree_skb(skb); |
897 | } | |
2c8c4fb7 | 898 | return 0; |
6ff686f2 PS |
899 | } |
900 | ||
e74d4817 PS |
901 | static void execute_hash(struct sk_buff *skb, struct sw_flow_key *key, |
902 | const struct nlattr *attr) | |
7804df20 | 903 | { |
7804df20 AZ |
904 | struct ovs_action_hash *hash_act = nla_data(attr); |
905 | u32 hash = 0; | |
906 | ||
907 | /* OVS_HASH_ALG_L4 is the only possible hash algorithm. */ | |
e2f3178f | 908 | hash = skb_get_hash(skb); |
7804df20 AZ |
909 | hash = jhash_1word(hash, hash_act->hash_basis); |
910 | if (!hash) | |
911 | hash = 0x1; | |
912 | ||
913 | key->ovs_flow_hash = hash; | |
914 | } | |
915 | ||
b940b3d7 JR |
916 | static int execute_set_action(struct sk_buff *skb, |
917 | struct sw_flow_key *flow_key, | |
918 | const struct nlattr *a) | |
919 | { | |
920 | /* Only tunnel set execution is supported without a mask. */ | |
921 | if (nla_type(a) == OVS_KEY_ATTR_TUNNEL_INFO) { | |
e23775f2 PS |
922 | struct ovs_tunnel_info *tun = nla_data(a); |
923 | ||
924 | ovs_skb_dst_drop(skb); | |
925 | ovs_dst_hold((struct dst_entry *)tun->tun_dst); | |
926 | ovs_skb_dst_set(skb, (struct dst_entry *)tun->tun_dst); | |
b940b3d7 JR |
927 | return 0; |
928 | } | |
929 | ||
930 | return -EINVAL; | |
b940b3d7 JR |
931 | } |
932 | ||
933 | /* Mask is at the midpoint of the data. */ | |
934 | #define get_mask(a, type) ((const type)nla_data(a) + 1) | |
935 | ||
936 | static int execute_masked_set_action(struct sk_buff *skb, | |
937 | struct sw_flow_key *flow_key, | |
938 | const struct nlattr *a) | |
4edb9ae9 | 939 | { |
15c39847 | 940 | int err = 0; |
4edb9ae9 | 941 | |
b940b3d7 | 942 | switch (nla_type(a)) { |
abff858b | 943 | case OVS_KEY_ATTR_PRIORITY: |
e281bb23 JS |
944 | OVS_SET_MASKED(skb->priority, nla_get_u32(a), |
945 | *get_mask(a, u32 *)); | |
b940b3d7 | 946 | flow_key->phy.priority = skb->priority; |
abff858b PS |
947 | break; |
948 | ||
72e8bf28 | 949 | case OVS_KEY_ATTR_SKB_MARK: |
e281bb23 | 950 | OVS_SET_MASKED(skb->mark, nla_get_u32(a), *get_mask(a, u32 *)); |
b940b3d7 | 951 | flow_key->phy.skb_mark = skb->mark; |
72e8bf28 AA |
952 | break; |
953 | ||
f0cd669f | 954 | case OVS_KEY_ATTR_TUNNEL_INFO: |
b940b3d7 JR |
955 | /* Masked data not supported for tunnel. */ |
956 | err = -EINVAL; | |
4edb9ae9 PS |
957 | break; |
958 | ||
959 | case OVS_KEY_ATTR_ETHERNET: | |
b940b3d7 JR |
960 | err = set_eth_addr(skb, flow_key, nla_data(a), |
961 | get_mask(a, struct ovs_key_ethernet *)); | |
4edb9ae9 PS |
962 | break; |
963 | ||
964 | case OVS_KEY_ATTR_IPV4: | |
b940b3d7 JR |
965 | err = set_ipv4(skb, flow_key, nla_data(a), |
966 | get_mask(a, struct ovs_key_ipv4 *)); | |
4edb9ae9 PS |
967 | break; |
968 | ||
bc7a5acd | 969 | case OVS_KEY_ATTR_IPV6: |
b940b3d7 JR |
970 | err = set_ipv6(skb, flow_key, nla_data(a), |
971 | get_mask(a, struct ovs_key_ipv6 *)); | |
bc7a5acd AA |
972 | break; |
973 | ||
4edb9ae9 | 974 | case OVS_KEY_ATTR_TCP: |
b940b3d7 JR |
975 | err = set_tcp(skb, flow_key, nla_data(a), |
976 | get_mask(a, struct ovs_key_tcp *)); | |
4edb9ae9 PS |
977 | break; |
978 | ||
979 | case OVS_KEY_ATTR_UDP: | |
b940b3d7 JR |
980 | err = set_udp(skb, flow_key, nla_data(a), |
981 | get_mask(a, struct ovs_key_udp *)); | |
4edb9ae9 | 982 | break; |
10f72e3d JS |
983 | |
984 | case OVS_KEY_ATTR_SCTP: | |
b940b3d7 JR |
985 | err = set_sctp(skb, flow_key, nla_data(a), |
986 | get_mask(a, struct ovs_key_sctp *)); | |
10f72e3d | 987 | break; |
ccf43786 SH |
988 | |
989 | case OVS_KEY_ATTR_MPLS: | |
b940b3d7 JR |
990 | err = set_mpls(skb, flow_key, nla_data(a), get_mask(a, |
991 | __be32 *)); | |
ccf43786 | 992 | break; |
a94ebc39 JS |
993 | |
994 | case OVS_KEY_ATTR_CT_STATE: | |
995 | case OVS_KEY_ATTR_CT_ZONE: | |
372ce973 | 996 | case OVS_KEY_ATTR_CT_MARK: |
c05e2094 | 997 | case OVS_KEY_ATTR_CT_LABELS: |
a94ebc39 JS |
998 | err = -EINVAL; |
999 | break; | |
4edb9ae9 | 1000 | } |
15c39847 | 1001 | |
4edb9ae9 PS |
1002 | return err; |
1003 | } | |
1004 | ||
a6059080 | 1005 | static int execute_recirc(struct datapath *dp, struct sk_buff *skb, |
7d16c847 PS |
1006 | struct sw_flow_key *key, |
1007 | const struct nlattr *a, int rem) | |
a6059080 | 1008 | { |
e74d4817 PS |
1009 | struct deferred_action *da; |
1010 | ||
1011 | if (!is_flow_key_valid(key)) { | |
867e37ba AZ |
1012 | int err; |
1013 | ||
e74d4817 | 1014 | err = ovs_flow_key_update(skb, key); |
867e37ba AZ |
1015 | if (err) |
1016 | return err; | |
867e37ba | 1017 | } |
e74d4817 | 1018 | BUG_ON(!is_flow_key_valid(key)); |
a6059080 | 1019 | |
684b5f5d | 1020 | if (!nla_is_last(a, rem)) { |
e16138e2 | 1021 | /* Recirc action is the not the last action |
e74d4817 PS |
1022 | * of the action list, need to clone the skb. |
1023 | */ | |
e16138e2 PS |
1024 | skb = skb_clone(skb, GFP_ATOMIC); |
1025 | ||
1026 | /* Skip the recirc action when out of memory, but | |
e74d4817 PS |
1027 | * continue on with the rest of the action list. |
1028 | */ | |
e16138e2 PS |
1029 | if (!skb) |
1030 | return 0; | |
2c8c4fb7 | 1031 | } |
a6059080 | 1032 | |
e74d4817 PS |
1033 | da = add_deferred_actions(skb, key, NULL); |
1034 | if (da) { | |
1035 | da->pkt_key.recirc_id = nla_get_u32(a); | |
2c8c4fb7 AZ |
1036 | } else { |
1037 | kfree_skb(skb); | |
1038 | ||
1039 | if (net_ratelimit()) | |
1040 | pr_warn("%s: deferred action limit reached, drop recirc action\n", | |
1041 | ovs_dp_name(dp)); | |
867e37ba | 1042 | } |
a6059080 AZ |
1043 | |
1044 | return 0; | |
1045 | } | |
1046 | ||
064af421 | 1047 | /* Execute a list of actions against 'skb'. */ |
871dfe07 | 1048 | static int do_execute_actions(struct datapath *dp, struct sk_buff *skb, |
e74d4817 PS |
1049 | struct sw_flow_key *key, |
1050 | const struct nlattr *attr, int len) | |
064af421 BP |
1051 | { |
1052 | /* Every output action needs a separate clone of 'skb', but the common | |
1053 | * case is just a single output action, so that doing a clone and | |
1054 | * then freeing the original skbuff is wasteful. So the following code | |
e74d4817 PS |
1055 | * is slightly obscure just to avoid that. |
1056 | */ | |
064af421 | 1057 | int prev_port = -1; |
cdee00fd | 1058 | const struct nlattr *a; |
10db8b20 | 1059 | int rem; |
72b06300 | 1060 | |
6ff686f2 | 1061 | for (a = attr, rem = len; rem > 0; |
a4af2475 | 1062 | a = nla_next(a, &rem)) { |
10db8b20 JG |
1063 | int err = 0; |
1064 | ||
fe90efd9 AZ |
1065 | if (unlikely(prev_port != -1)) { |
1066 | struct sk_buff *out_skb = skb_clone(skb, GFP_ATOMIC); | |
1067 | ||
1068 | if (out_skb) | |
a94ebc39 | 1069 | do_output(dp, out_skb, prev_port, key); |
fe90efd9 | 1070 | |
4c7804f1 | 1071 | OVS_CB(skb)->cutlen = 0; |
064af421 BP |
1072 | prev_port = -1; |
1073 | } | |
1074 | ||
cdee00fd | 1075 | switch (nla_type(a)) { |
df2c07f4 | 1076 | case OVS_ACTION_ATTR_OUTPUT: |
cdee00fd | 1077 | prev_port = nla_get_u32(a); |
064af421 BP |
1078 | break; |
1079 | ||
4c7804f1 WT |
1080 | case OVS_ACTION_ATTR_TRUNC: { |
1081 | struct ovs_action_trunc *trunc = nla_data(a); | |
1082 | ||
1083 | if (skb->len > trunc->max_len) | |
1084 | OVS_CB(skb)->cutlen = skb->len - trunc->max_len; | |
1085 | break; | |
1086 | } | |
1087 | ||
df2c07f4 | 1088 | case OVS_ACTION_ATTR_USERSPACE: |
4c7804f1 WT |
1089 | output_userspace(dp, skb, key, a, attr, |
1090 | len, OVS_CB(skb)->cutlen); | |
1091 | OVS_CB(skb)->cutlen = 0; | |
064af421 | 1092 | break; |
7804df20 AZ |
1093 | |
1094 | case OVS_ACTION_ATTR_HASH: | |
e74d4817 | 1095 | execute_hash(skb, key, a); |
7804df20 | 1096 | break; |
064af421 | 1097 | |
ccf43786 | 1098 | case OVS_ACTION_ATTR_PUSH_MPLS: |
e74d4817 | 1099 | err = push_mpls(skb, key, nla_data(a)); |
ccf43786 SH |
1100 | break; |
1101 | ||
1102 | case OVS_ACTION_ATTR_POP_MPLS: | |
e74d4817 | 1103 | err = pop_mpls(skb, key, nla_get_be16(a)); |
ccf43786 SH |
1104 | break; |
1105 | ||
fea393b1 | 1106 | case OVS_ACTION_ATTR_PUSH_VLAN: |
e74d4817 | 1107 | err = push_vlan(skb, key, nla_data(a)); |
064af421 BP |
1108 | break; |
1109 | ||
fea393b1 | 1110 | case OVS_ACTION_ATTR_POP_VLAN: |
e74d4817 | 1111 | err = pop_vlan(skb, key); |
064af421 BP |
1112 | break; |
1113 | ||
e16138e2 | 1114 | case OVS_ACTION_ATTR_RECIRC: |
e74d4817 | 1115 | err = execute_recirc(dp, skb, key, a, rem); |
684b5f5d | 1116 | if (nla_is_last(a, rem)) { |
867e37ba AZ |
1117 | /* If this is the last action, the skb has |
1118 | * been consumed or freed. | |
e74d4817 PS |
1119 | * Return immediately. |
1120 | */ | |
867e37ba AZ |
1121 | return err; |
1122 | } | |
a6059080 | 1123 | break; |
a6059080 | 1124 | |
4edb9ae9 | 1125 | case OVS_ACTION_ATTR_SET: |
e74d4817 | 1126 | err = execute_set_action(skb, key, nla_data(a)); |
064af421 | 1127 | break; |
c1c9c9c4 | 1128 | |
b940b3d7 JR |
1129 | case OVS_ACTION_ATTR_SET_MASKED: |
1130 | case OVS_ACTION_ATTR_SET_TO_MASKED: | |
1131 | err = execute_masked_set_action(skb, key, nla_data(a)); | |
1132 | break; | |
1133 | ||
6ff686f2 | 1134 | case OVS_ACTION_ATTR_SAMPLE: |
0e469d3b | 1135 | err = sample(dp, skb, key, a, attr, len); |
6ff686f2 | 1136 | break; |
a94ebc39 JS |
1137 | |
1138 | case OVS_ACTION_ATTR_CT: | |
c05e2094 JS |
1139 | if (!is_flow_key_valid(key)) { |
1140 | err = ovs_flow_key_update(skb, key); | |
1141 | if (err) | |
1142 | return err; | |
1143 | } | |
1144 | ||
a94ebc39 JS |
1145 | err = ovs_ct_execute(ovs_dp_get_net(dp), skb, key, |
1146 | nla_data(a)); | |
1147 | ||
1148 | /* Hide stolen IP fragments from user space. */ | |
c05e2094 JS |
1149 | if (err) |
1150 | return err == -EINPROGRESS ? 0 : err; | |
a94ebc39 | 1151 | break; |
6ff686f2 | 1152 | } |
15c39847 | 1153 | |
10db8b20 JG |
1154 | if (unlikely(err)) { |
1155 | kfree_skb(skb); | |
1156 | return err; | |
1157 | } | |
064af421 | 1158 | } |
6c222e55 | 1159 | |
fbf4f74d | 1160 | if (prev_port != -1) |
a94ebc39 | 1161 | do_output(dp, skb, prev_port, key); |
fbf4f74d | 1162 | else |
5b95ab0e | 1163 | consume_skb(skb); |
10db8b20 | 1164 | |
a5225dd6 | 1165 | return 0; |
064af421 | 1166 | } |
871dfe07 | 1167 | |
2c8c4fb7 AZ |
1168 | static void process_deferred_actions(struct datapath *dp) |
1169 | { | |
1170 | struct action_fifo *fifo = this_cpu_ptr(action_fifos); | |
1171 | ||
1172 | /* Do not touch the FIFO in case there is no deferred actions. */ | |
1173 | if (action_fifo_is_empty(fifo)) | |
1174 | return; | |
1175 | ||
1176 | /* Finishing executing all deferred actions. */ | |
1177 | do { | |
1178 | struct deferred_action *da = action_fifo_get(fifo); | |
1179 | struct sk_buff *skb = da->skb; | |
7d16c847 | 1180 | struct sw_flow_key *key = &da->pkt_key; |
2c8c4fb7 AZ |
1181 | const struct nlattr *actions = da->actions; |
1182 | ||
1183 | if (actions) | |
7d16c847 | 1184 | do_execute_actions(dp, skb, key, actions, |
2c8c4fb7 AZ |
1185 | nla_len(actions)); |
1186 | else | |
7d16c847 | 1187 | ovs_dp_process_packet(skb, key); |
2c8c4fb7 AZ |
1188 | } while (!action_fifo_is_empty(fifo)); |
1189 | ||
1190 | /* Reset FIFO for the next packet. */ | |
1191 | action_fifo_init(fifo); | |
1192 | } | |
1193 | ||
871dfe07 | 1194 | /* Execute a list of actions against 'skb'. */ |
2c8c4fb7 | 1195 | int ovs_execute_actions(struct datapath *dp, struct sk_buff *skb, |
7d16c847 PS |
1196 | const struct sw_flow_actions *acts, |
1197 | struct sw_flow_key *key) | |
2c8c4fb7 AZ |
1198 | { |
1199 | int level = this_cpu_read(exec_actions_level); | |
1200 | int err; | |
1201 | ||
1202 | if (unlikely(level >= EXEC_ACTIONS_LEVEL_LIMIT)) { | |
1203 | if (net_ratelimit()) | |
1204 | pr_warn("%s: packet loop detected, dropping.\n", | |
1205 | ovs_dp_name(dp)); | |
1206 | ||
1207 | kfree_skb(skb); | |
1208 | return -ELOOP; | |
1209 | } | |
1210 | ||
1211 | this_cpu_inc(exec_actions_level); | |
7d16c847 PS |
1212 | err = do_execute_actions(dp, skb, key, |
1213 | acts->actions, acts->actions_len); | |
2c8c4fb7 AZ |
1214 | |
1215 | if (!level) | |
1216 | process_deferred_actions(dp); | |
1217 | ||
1218 | this_cpu_dec(exec_actions_level); | |
1219 | ||
1220 | /* This return status currently does not reflect the errors | |
1221 | * encounted during deferred actions execution. Probably needs to | |
e74d4817 PS |
1222 | * be fixed in the future. |
1223 | */ | |
2c8c4fb7 AZ |
1224 | return err; |
1225 | } | |
1226 | ||
1227 | int action_fifos_init(void) | |
1228 | { | |
1229 | action_fifos = alloc_percpu(struct action_fifo); | |
1230 | if (!action_fifos) | |
1231 | return -ENOMEM; | |
1232 | ||
1233 | return 0; | |
1234 | } | |
1235 | ||
1236 | void action_fifos_exit(void) | |
60759b2b | 1237 | { |
2c8c4fb7 | 1238 | free_percpu(action_fifos); |
871dfe07 | 1239 | } |