]>
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; | |
77a9a338 | 81 | |
2c8c4fb7 AZ |
82 | static DEFINE_PER_CPU(int, exec_actions_level); |
83 | ||
84 | static void action_fifo_init(struct action_fifo *fifo) | |
85 | { | |
86 | fifo->head = 0; | |
87 | fifo->tail = 0; | |
88 | } | |
89 | ||
f1f60b85 | 90 | static bool action_fifo_is_empty(const struct action_fifo *fifo) |
2c8c4fb7 AZ |
91 | { |
92 | return (fifo->head == fifo->tail); | |
93 | } | |
94 | ||
e74d4817 | 95 | static struct deferred_action *action_fifo_get(struct action_fifo *fifo) |
2c8c4fb7 AZ |
96 | { |
97 | if (action_fifo_is_empty(fifo)) | |
98 | return NULL; | |
99 | ||
100 | return &fifo->fifo[fifo->tail++]; | |
101 | } | |
102 | ||
e74d4817 | 103 | static struct deferred_action *action_fifo_put(struct action_fifo *fifo) |
2c8c4fb7 AZ |
104 | { |
105 | if (fifo->head >= DEFERRED_ACTION_FIFO_SIZE - 1) | |
106 | return NULL; | |
107 | ||
108 | return &fifo->fifo[fifo->head++]; | |
109 | } | |
110 | ||
e74d4817 PS |
111 | /* Return queue entry if fifo is not full */ |
112 | static struct deferred_action *add_deferred_actions(struct sk_buff *skb, | |
f1f60b85 | 113 | const struct sw_flow_key *key, |
e74d4817 | 114 | const struct nlattr *attr) |
2c8c4fb7 AZ |
115 | { |
116 | struct action_fifo *fifo; | |
117 | struct deferred_action *da; | |
118 | ||
119 | fifo = this_cpu_ptr(action_fifos); | |
120 | da = action_fifo_put(fifo); | |
121 | if (da) { | |
122 | da->skb = skb; | |
123 | da->actions = attr; | |
e74d4817 | 124 | da->pkt_key = *key; |
2c8c4fb7 AZ |
125 | } |
126 | ||
e74d4817 | 127 | return da; |
e16138e2 PS |
128 | } |
129 | ||
e74d4817 | 130 | static void invalidate_flow_key(struct sw_flow_key *key) |
e16138e2 | 131 | { |
e74d4817 | 132 | key->eth.type = htons(0); |
e16138e2 PS |
133 | } |
134 | ||
f1f60b85 | 135 | static bool is_flow_key_valid(const struct sw_flow_key *key) |
e16138e2 | 136 | { |
e74d4817 | 137 | return !!key->eth.type; |
e16138e2 PS |
138 | } |
139 | ||
b51367aa PS |
140 | static void update_ethertype(struct sk_buff *skb, struct ethhdr *hdr, |
141 | __be16 ethertype) | |
142 | { | |
143 | if (skb->ip_summed == CHECKSUM_COMPLETE) { | |
144 | __be16 diff[] = { ~(hdr->h_proto), ethertype }; | |
145 | ||
146 | skb->csum = ~csum_partial((char *)diff, sizeof(diff), | |
147 | ~skb->csum); | |
148 | } | |
149 | ||
150 | hdr->h_proto = ethertype; | |
151 | } | |
152 | ||
e74d4817 | 153 | static int push_mpls(struct sk_buff *skb, struct sw_flow_key *key, |
ccf43786 SH |
154 | const struct ovs_action_push_mpls *mpls) |
155 | { | |
156 | __be32 *new_mpls_lse; | |
ccf43786 | 157 | |
2baf0e0c | 158 | /* Networking stack do not allow simultaneous Tunnel and MPLS GSO. */ |
8063e095 | 159 | if (skb->encapsulation) |
2baf0e0c PS |
160 | return -ENOTSUPP; |
161 | ||
ccf43786 SH |
162 | if (skb_cow_head(skb, MPLS_HLEN) < 0) |
163 | return -ENOMEM; | |
164 | ||
165 | skb_push(skb, MPLS_HLEN); | |
166 | memmove(skb_mac_header(skb) - MPLS_HLEN, skb_mac_header(skb), | |
167 | skb->mac_len); | |
168 | skb_reset_mac_header(skb); | |
169 | ||
2baf0e0c | 170 | new_mpls_lse = (__be32 *)skb_mpls_header(skb); |
ccf43786 SH |
171 | *new_mpls_lse = mpls->mpls_lse; |
172 | ||
ea3acd7a | 173 | skb_postpush_rcsum(skb, new_mpls_lse, MPLS_HLEN); |
ccf43786 | 174 | |
b51367aa | 175 | update_ethertype(skb, eth_hdr(skb), mpls->mpls_ethertype); |
ccf43786 SH |
176 | if (!ovs_skb_get_inner_protocol(skb)) |
177 | ovs_skb_set_inner_protocol(skb, skb->protocol); | |
178 | skb->protocol = mpls->mpls_ethertype; | |
2baf0e0c | 179 | |
e74d4817 | 180 | invalidate_flow_key(key); |
ccf43786 SH |
181 | return 0; |
182 | } | |
183 | ||
e74d4817 PS |
184 | static int pop_mpls(struct sk_buff *skb, struct sw_flow_key *key, |
185 | const __be16 ethertype) | |
ccf43786 SH |
186 | { |
187 | struct ethhdr *hdr; | |
188 | int err; | |
189 | ||
5cce04b6 | 190 | err = skb_ensure_writable(skb, skb->mac_len + MPLS_HLEN); |
ccf43786 SH |
191 | if (unlikely(err)) |
192 | return err; | |
193 | ||
f021a62f | 194 | skb_postpull_rcsum(skb, skb_mpls_header(skb), MPLS_HLEN); |
ccf43786 SH |
195 | |
196 | memmove(skb_mac_header(skb) + MPLS_HLEN, skb_mac_header(skb), | |
197 | skb->mac_len); | |
198 | ||
199 | __skb_pull(skb, MPLS_HLEN); | |
200 | skb_reset_mac_header(skb); | |
201 | ||
2baf0e0c | 202 | /* skb_mpls_header() is used to locate the ethertype |
ccf43786 SH |
203 | * field correctly in the presence of VLAN tags. |
204 | */ | |
2baf0e0c | 205 | hdr = (struct ethhdr *)(skb_mpls_header(skb) - ETH_HLEN); |
b51367aa | 206 | update_ethertype(skb, hdr, ethertype); |
ccf43786 SH |
207 | if (eth_p_mpls(skb->protocol)) |
208 | skb->protocol = ethertype; | |
2baf0e0c | 209 | |
e74d4817 | 210 | invalidate_flow_key(key); |
ccf43786 SH |
211 | return 0; |
212 | } | |
213 | ||
b940b3d7 JR |
214 | static int set_mpls(struct sk_buff *skb, struct sw_flow_key *flow_key, |
215 | const __be32 *mpls_lse, const __be32 *mask) | |
ccf43786 | 216 | { |
2baf0e0c | 217 | __be32 *stack; |
b940b3d7 | 218 | __be32 lse; |
ccf43786 SH |
219 | int err; |
220 | ||
5cce04b6 | 221 | err = skb_ensure_writable(skb, skb->mac_len + MPLS_HLEN); |
ccf43786 SH |
222 | if (unlikely(err)) |
223 | return err; | |
224 | ||
2baf0e0c | 225 | stack = (__be32 *)skb_mpls_header(skb); |
e281bb23 | 226 | lse = OVS_MASKED(*stack, *mpls_lse, *mask); |
ccf43786 | 227 | if (skb->ip_summed == CHECKSUM_COMPLETE) { |
b940b3d7 JR |
228 | __be32 diff[] = { ~(*stack), lse }; |
229 | ||
ccf43786 SH |
230 | skb->csum = ~csum_partial((char *)diff, sizeof(diff), |
231 | ~skb->csum); | |
232 | } | |
233 | ||
b940b3d7 JR |
234 | *stack = lse; |
235 | flow_key->mpls.top_lse = lse; | |
ccf43786 SH |
236 | return 0; |
237 | } | |
238 | ||
e74d4817 | 239 | static int pop_vlan(struct sk_buff *skb, struct sw_flow_key *key) |
064af421 | 240 | { |
d9065a90 | 241 | int err; |
10db8b20 | 242 | |
97894370 | 243 | err = skb_vlan_pop(skb); |
efd8a18e | 244 | if (skb_vlan_tag_present(skb)) |
97894370 TG |
245 | invalidate_flow_key(key); |
246 | else | |
e74d4817 | 247 | key->eth.tci = 0; |
97894370 | 248 | return err; |
d9065a90 PS |
249 | } |
250 | ||
e74d4817 PS |
251 | static int push_vlan(struct sk_buff *skb, struct sw_flow_key *key, |
252 | const struct ovs_action_push_vlan *vlan) | |
d9065a90 | 253 | { |
efd8a18e | 254 | if (skb_vlan_tag_present(skb)) |
e74d4817 | 255 | invalidate_flow_key(key); |
97894370 | 256 | else |
e74d4817 | 257 | key->eth.tci = vlan->vlan_tci; |
97894370 TG |
258 | return skb_vlan_push(skb, vlan->vlan_tpid, |
259 | ntohs(vlan->vlan_tci) & ~VLAN_TAG_PRESENT); | |
064af421 BP |
260 | } |
261 | ||
b940b3d7 JR |
262 | /* 'src' is already properly masked. */ |
263 | static void ether_addr_copy_masked(u8 *dst_, const u8 *src_, const u8 *mask_) | |
264 | { | |
265 | u16 *dst = (u16 *)dst_; | |
266 | const u16 *src = (const u16 *)src_; | |
267 | const u16 *mask = (const u16 *)mask_; | |
268 | ||
e281bb23 JS |
269 | OVS_SET_MASKED(dst[0], src[0], mask[0]); |
270 | OVS_SET_MASKED(dst[1], src[1], mask[1]); | |
271 | OVS_SET_MASKED(dst[2], src[2], mask[2]); | |
b940b3d7 JR |
272 | } |
273 | ||
274 | static int set_eth_addr(struct sk_buff *skb, struct sw_flow_key *flow_key, | |
275 | const struct ovs_key_ethernet *key, | |
276 | const struct ovs_key_ethernet *mask) | |
ca78c6b6 | 277 | { |
4edb9ae9 | 278 | int err; |
b940b3d7 | 279 | |
5cce04b6 | 280 | err = skb_ensure_writable(skb, ETH_HLEN); |
4edb9ae9 PS |
281 | if (unlikely(err)) |
282 | return err; | |
283 | ||
237c4f2a | 284 | skb_postpull_rcsum(skb, eth_hdr(skb), ETH_ALEN * 2); |
3cfede14 | 285 | |
b940b3d7 JR |
286 | ether_addr_copy_masked(eth_hdr(skb)->h_source, key->eth_src, |
287 | mask->eth_src); | |
288 | ether_addr_copy_masked(eth_hdr(skb)->h_dest, key->eth_dst, | |
289 | mask->eth_dst); | |
4edb9ae9 | 290 | |
ea3acd7a | 291 | skb_postpush_rcsum(skb, eth_hdr(skb), ETH_ALEN * 2); |
3cfede14 | 292 | |
b940b3d7 JR |
293 | ether_addr_copy(flow_key->eth.src, eth_hdr(skb)->h_source); |
294 | ether_addr_copy(flow_key->eth.dst, eth_hdr(skb)->h_dest); | |
4edb9ae9 | 295 | return 0; |
ca78c6b6 BP |
296 | } |
297 | ||
efdb0c9f GG |
298 | static void update_ip_l4_checksum(struct sk_buff *skb, struct iphdr *nh, |
299 | __be32 addr, __be32 new_addr) | |
ca78c6b6 BP |
300 | { |
301 | int transport_len = skb->len - skb_transport_offset(skb); | |
4edb9ae9 | 302 | |
efdb0c9f GG |
303 | if (nh->frag_off & htons(IP_OFFSET)) |
304 | return; | |
305 | ||
4edb9ae9 | 306 | if (nh->protocol == IPPROTO_TCP) { |
ca78c6b6 | 307 | if (likely(transport_len >= sizeof(struct tcphdr))) |
4edb9ae9 | 308 | inet_proto_csum_replace4(&tcp_hdr(skb)->check, skb, |
8063e095 | 309 | addr, new_addr, true); |
4edb9ae9 | 310 | } else if (nh->protocol == IPPROTO_UDP) { |
55ce87bc JG |
311 | if (likely(transport_len >= sizeof(struct udphdr))) { |
312 | struct udphdr *uh = udp_hdr(skb); | |
313 | ||
237c4f2a | 314 | if (uh->check || skb->ip_summed == CHECKSUM_PARTIAL) { |
55ce87bc | 315 | inet_proto_csum_replace4(&uh->check, skb, |
8063e095 | 316 | addr, new_addr, true); |
55ce87bc JG |
317 | if (!uh->check) |
318 | uh->check = CSUM_MANGLED_0; | |
319 | } | |
320 | } | |
ca78c6b6 | 321 | } |
4edb9ae9 | 322 | |
efdb0c9f GG |
323 | } |
324 | ||
325 | static void set_ip_addr(struct sk_buff *skb, struct iphdr *nh, | |
326 | __be32 *addr, __be32 new_addr) | |
327 | { | |
328 | update_ip_l4_checksum(skb, nh, *addr, new_addr); | |
4edb9ae9 | 329 | csum_replace4(&nh->check, *addr, new_addr); |
e2f3178f | 330 | skb_clear_hash(skb); |
4edb9ae9 | 331 | *addr = new_addr; |
ca78c6b6 BP |
332 | } |
333 | ||
bc7a5acd AA |
334 | static void update_ipv6_checksum(struct sk_buff *skb, u8 l4_proto, |
335 | __be32 addr[4], const __be32 new_addr[4]) | |
336 | { | |
337 | int transport_len = skb->len - skb_transport_offset(skb); | |
338 | ||
00894212 | 339 | if (l4_proto == NEXTHDR_TCP) { |
bc7a5acd AA |
340 | if (likely(transport_len >= sizeof(struct tcphdr))) |
341 | inet_proto_csum_replace16(&tcp_hdr(skb)->check, skb, | |
8063e095 | 342 | addr, new_addr, true); |
00894212 | 343 | } else if (l4_proto == NEXTHDR_UDP) { |
bc7a5acd AA |
344 | if (likely(transport_len >= sizeof(struct udphdr))) { |
345 | struct udphdr *uh = udp_hdr(skb); | |
346 | ||
237c4f2a | 347 | if (uh->check || skb->ip_summed == CHECKSUM_PARTIAL) { |
bc7a5acd | 348 | inet_proto_csum_replace16(&uh->check, skb, |
8063e095 | 349 | addr, new_addr, true); |
bc7a5acd AA |
350 | if (!uh->check) |
351 | uh->check = CSUM_MANGLED_0; | |
352 | } | |
353 | } | |
00894212 JG |
354 | } else if (l4_proto == NEXTHDR_ICMP) { |
355 | if (likely(transport_len >= sizeof(struct icmp6hdr))) | |
356 | inet_proto_csum_replace16(&icmp6_hdr(skb)->icmp6_cksum, | |
8063e095 | 357 | skb, addr, new_addr, true); |
bc7a5acd AA |
358 | } |
359 | } | |
360 | ||
b940b3d7 JR |
361 | static void mask_ipv6_addr(const __be32 old[4], const __be32 addr[4], |
362 | const __be32 mask[4], __be32 masked[4]) | |
363 | { | |
e281bb23 JS |
364 | masked[0] = OVS_MASKED(old[0], addr[0], mask[0]); |
365 | masked[1] = OVS_MASKED(old[1], addr[1], mask[1]); | |
366 | masked[2] = OVS_MASKED(old[2], addr[2], mask[2]); | |
367 | masked[3] = OVS_MASKED(old[3], addr[3], mask[3]); | |
b940b3d7 JR |
368 | } |
369 | ||
bc7a5acd AA |
370 | static void set_ipv6_addr(struct sk_buff *skb, u8 l4_proto, |
371 | __be32 addr[4], const __be32 new_addr[4], | |
372 | bool recalculate_csum) | |
373 | { | |
51cf5e71 | 374 | if (likely(recalculate_csum)) |
bc7a5acd AA |
375 | update_ipv6_checksum(skb, l4_proto, addr, new_addr); |
376 | ||
e2f3178f | 377 | skb_clear_hash(skb); |
bc7a5acd AA |
378 | memcpy(addr, new_addr, sizeof(__be32[4])); |
379 | } | |
380 | ||
b940b3d7 | 381 | static void set_ipv6_fl(struct ipv6hdr *nh, u32 fl, u32 mask) |
bc7a5acd | 382 | { |
b940b3d7 | 383 | /* Bits 21-24 are always unmasked, so this retains their values. */ |
e281bb23 JS |
384 | OVS_SET_MASKED(nh->flow_lbl[0], (u8)(fl >> 16), (u8)(mask >> 16)); |
385 | OVS_SET_MASKED(nh->flow_lbl[1], (u8)(fl >> 8), (u8)(mask >> 8)); | |
386 | OVS_SET_MASKED(nh->flow_lbl[2], (u8)fl, (u8)mask); | |
bc7a5acd AA |
387 | } |
388 | ||
b940b3d7 JR |
389 | static void set_ip_ttl(struct sk_buff *skb, struct iphdr *nh, u8 new_ttl, |
390 | u8 mask) | |
bc7a5acd | 391 | { |
e281bb23 | 392 | new_ttl = OVS_MASKED(nh->ttl, new_ttl, mask); |
bc7a5acd | 393 | |
a61680c6 JP |
394 | csum_replace2(&nh->check, htons(nh->ttl << 8), htons(new_ttl << 8)); |
395 | nh->ttl = new_ttl; | |
396 | } | |
397 | ||
b940b3d7 JR |
398 | static int set_ipv4(struct sk_buff *skb, struct sw_flow_key *flow_key, |
399 | const struct ovs_key_ipv4 *key, | |
400 | const struct ovs_key_ipv4 *mask) | |
064af421 | 401 | { |
ca78c6b6 | 402 | struct iphdr *nh; |
b940b3d7 | 403 | __be32 new_addr; |
10db8b20 | 404 | int err; |
ca78c6b6 | 405 | |
5cce04b6 TG |
406 | err = skb_ensure_writable(skb, skb_network_offset(skb) + |
407 | sizeof(struct iphdr)); | |
10db8b20 JG |
408 | if (unlikely(err)) |
409 | return err; | |
ca78c6b6 BP |
410 | |
411 | nh = ip_hdr(skb); | |
ca78c6b6 | 412 | |
b940b3d7 JR |
413 | /* Setting an IP addresses is typically only a side effect of |
414 | * matching on them in the current userspace implementation, so it | |
415 | * makes sense to check if the value actually changed. | |
416 | */ | |
417 | if (mask->ipv4_src) { | |
e281bb23 | 418 | new_addr = OVS_MASKED(nh->saddr, key->ipv4_src, mask->ipv4_src); |
ca78c6b6 | 419 | |
b940b3d7 JR |
420 | if (unlikely(new_addr != nh->saddr)) { |
421 | set_ip_addr(skb, nh, &nh->saddr, new_addr); | |
422 | flow_key->ipv4.addr.src = new_addr; | |
423 | } | |
e16138e2 | 424 | } |
b940b3d7 | 425 | if (mask->ipv4_dst) { |
e281bb23 | 426 | new_addr = OVS_MASKED(nh->daddr, key->ipv4_dst, mask->ipv4_dst); |
a4a26436 | 427 | |
b940b3d7 JR |
428 | if (unlikely(new_addr != nh->daddr)) { |
429 | set_ip_addr(skb, nh, &nh->daddr, new_addr); | |
430 | flow_key->ipv4.addr.dst = new_addr; | |
431 | } | |
e16138e2 | 432 | } |
b940b3d7 JR |
433 | if (mask->ipv4_tos) { |
434 | ipv4_change_dsfield(nh, ~mask->ipv4_tos, key->ipv4_tos); | |
435 | flow_key->ip.tos = nh->tos; | |
436 | } | |
437 | if (mask->ipv4_ttl) { | |
438 | set_ip_ttl(skb, nh, key->ipv4_ttl, mask->ipv4_ttl); | |
439 | flow_key->ip.ttl = nh->ttl; | |
e16138e2 | 440 | } |
a61680c6 | 441 | |
10db8b20 | 442 | return 0; |
064af421 BP |
443 | } |
444 | ||
b940b3d7 JR |
445 | static bool is_ipv6_mask_nonzero(const __be32 addr[4]) |
446 | { | |
447 | return !!(addr[0] | addr[1] | addr[2] | addr[3]); | |
448 | } | |
449 | ||
450 | static int set_ipv6(struct sk_buff *skb, struct sw_flow_key *flow_key, | |
451 | const struct ovs_key_ipv6 *key, | |
452 | const struct ovs_key_ipv6 *mask) | |
bc7a5acd AA |
453 | { |
454 | struct ipv6hdr *nh; | |
455 | int err; | |
bc7a5acd | 456 | |
5cce04b6 TG |
457 | err = skb_ensure_writable(skb, skb_network_offset(skb) + |
458 | sizeof(struct ipv6hdr)); | |
bc7a5acd AA |
459 | if (unlikely(err)) |
460 | return err; | |
461 | ||
462 | nh = ipv6_hdr(skb); | |
bc7a5acd | 463 | |
b940b3d7 JR |
464 | /* Setting an IP addresses is typically only a side effect of |
465 | * matching on them in the current userspace implementation, so it | |
466 | * makes sense to check if the value actually changed. | |
467 | */ | |
468 | if (is_ipv6_mask_nonzero(mask->ipv6_src)) { | |
469 | __be32 *saddr = (__be32 *)&nh->saddr; | |
470 | __be32 masked[4]; | |
471 | ||
472 | mask_ipv6_addr(saddr, key->ipv6_src, mask->ipv6_src, masked); | |
473 | ||
474 | if (unlikely(memcmp(saddr, masked, sizeof(masked)))) { | |
49a8eef8 | 475 | set_ipv6_addr(skb, flow_key->ip.proto, saddr, masked, |
b940b3d7 JR |
476 | true); |
477 | memcpy(&flow_key->ipv6.addr.src, masked, | |
478 | sizeof(flow_key->ipv6.addr.src)); | |
479 | } | |
480 | } | |
481 | if (is_ipv6_mask_nonzero(mask->ipv6_dst)) { | |
bc7a5acd | 482 | unsigned int offset = 0; |
8abaa53c | 483 | int flags = IP6_FH_F_SKIP_RH; |
bc7a5acd | 484 | bool recalc_csum = true; |
b940b3d7 JR |
485 | __be32 *daddr = (__be32 *)&nh->daddr; |
486 | __be32 masked[4]; | |
487 | ||
488 | mask_ipv6_addr(daddr, key->ipv6_dst, mask->ipv6_dst, masked); | |
489 | ||
490 | if (unlikely(memcmp(daddr, masked, sizeof(masked)))) { | |
491 | if (ipv6_ext_hdr(nh->nexthdr)) | |
492 | recalc_csum = (ipv6_find_hdr(skb, &offset, | |
493 | NEXTHDR_ROUTING, | |
494 | NULL, &flags) | |
495 | != NEXTHDR_ROUTING); | |
496 | ||
49a8eef8 | 497 | set_ipv6_addr(skb, flow_key->ip.proto, daddr, masked, |
b940b3d7 JR |
498 | recalc_csum); |
499 | memcpy(&flow_key->ipv6.addr.dst, masked, | |
500 | sizeof(flow_key->ipv6.addr.dst)); | |
501 | } | |
502 | } | |
503 | if (mask->ipv6_tclass) { | |
504 | ipv6_change_dsfield(nh, ~mask->ipv6_tclass, key->ipv6_tclass); | |
505 | flow_key->ip.tos = ipv6_get_dsfield(nh); | |
506 | } | |
507 | if (mask->ipv6_label) { | |
508 | set_ipv6_fl(nh, ntohl(key->ipv6_label), | |
509 | ntohl(mask->ipv6_label)); | |
510 | flow_key->ipv6.label = | |
511 | *(__be32 *)nh & htonl(IPV6_FLOWINFO_FLOWLABEL); | |
512 | } | |
513 | if (mask->ipv6_hlimit) { | |
e281bb23 JS |
514 | OVS_SET_MASKED(nh->hop_limit, key->ipv6_hlimit, |
515 | mask->ipv6_hlimit); | |
b940b3d7 | 516 | flow_key->ip.ttl = nh->hop_limit; |
bc7a5acd | 517 | } |
bc7a5acd AA |
518 | return 0; |
519 | } | |
520 | ||
5cce04b6 | 521 | /* Must follow skb_ensure_writable() since that can move the skb data. */ |
4edb9ae9 | 522 | static void set_tp_port(struct sk_buff *skb, __be16 *port, |
b940b3d7 | 523 | __be16 new_port, __sum16 *check) |
959a2ecd | 524 | { |
8063e095 | 525 | inet_proto_csum_replace2(check, skb, *port, new_port, false); |
4edb9ae9 | 526 | *port = new_port; |
55ce87bc JG |
527 | } |
528 | ||
b940b3d7 JR |
529 | static int set_udp(struct sk_buff *skb, struct sw_flow_key *flow_key, |
530 | const struct ovs_key_udp *key, | |
531 | const struct ovs_key_udp *mask) | |
4edb9ae9 PS |
532 | { |
533 | struct udphdr *uh; | |
b940b3d7 | 534 | __be16 src, dst; |
4edb9ae9 | 535 | int err; |
10db8b20 | 536 | |
5cce04b6 TG |
537 | err = skb_ensure_writable(skb, skb_transport_offset(skb) + |
538 | sizeof(struct udphdr)); | |
10db8b20 JG |
539 | if (unlikely(err)) |
540 | return err; | |
541 | ||
4edb9ae9 | 542 | uh = udp_hdr(skb); |
b940b3d7 | 543 | /* Either of the masks is non-zero, so do not bother checking them. */ |
e281bb23 JS |
544 | src = OVS_MASKED(uh->source, key->udp_src, mask->udp_src); |
545 | dst = OVS_MASKED(uh->dest, key->udp_dst, mask->udp_dst); | |
4edb9ae9 | 546 | |
b940b3d7 JR |
547 | if (uh->check && skb->ip_summed != CHECKSUM_PARTIAL) { |
548 | if (likely(src != uh->source)) { | |
549 | set_tp_port(skb, &uh->source, src, &uh->check); | |
550 | flow_key->tp.src = src; | |
551 | } | |
552 | if (likely(dst != uh->dest)) { | |
553 | set_tp_port(skb, &uh->dest, dst, &uh->check); | |
554 | flow_key->tp.dst = dst; | |
555 | } | |
556 | ||
557 | if (unlikely(!uh->check)) | |
558 | uh->check = CSUM_MANGLED_0; | |
559 | } else { | |
560 | uh->source = src; | |
561 | uh->dest = dst; | |
562 | flow_key->tp.src = src; | |
563 | flow_key->tp.dst = dst; | |
e16138e2 | 564 | } |
10db8b20 | 565 | |
b940b3d7 JR |
566 | skb_clear_hash(skb); |
567 | ||
10db8b20 | 568 | return 0; |
959a2ecd JP |
569 | } |
570 | ||
b940b3d7 JR |
571 | static int set_tcp(struct sk_buff *skb, struct sw_flow_key *flow_key, |
572 | const struct ovs_key_tcp *key, | |
573 | const struct ovs_key_tcp *mask) | |
064af421 | 574 | { |
4edb9ae9 | 575 | struct tcphdr *th; |
b940b3d7 | 576 | __be16 src, dst; |
10db8b20 | 577 | int err; |
064af421 | 578 | |
5cce04b6 TG |
579 | err = skb_ensure_writable(skb, skb_transport_offset(skb) + |
580 | sizeof(struct tcphdr)); | |
10db8b20 JG |
581 | if (unlikely(err)) |
582 | return err; | |
ca78c6b6 | 583 | |
4edb9ae9 | 584 | th = tcp_hdr(skb); |
e281bb23 | 585 | src = OVS_MASKED(th->source, key->tcp_src, mask->tcp_src); |
b940b3d7 JR |
586 | if (likely(src != th->source)) { |
587 | set_tp_port(skb, &th->source, src, &th->check); | |
588 | flow_key->tp.src = src; | |
589 | } | |
e281bb23 | 590 | dst = OVS_MASKED(th->dest, key->tcp_dst, mask->tcp_dst); |
b940b3d7 JR |
591 | if (likely(dst != th->dest)) { |
592 | set_tp_port(skb, &th->dest, dst, &th->check); | |
593 | flow_key->tp.dst = dst; | |
e16138e2 | 594 | } |
b940b3d7 | 595 | skb_clear_hash(skb); |
ca78c6b6 | 596 | |
10db8b20 | 597 | return 0; |
064af421 BP |
598 | } |
599 | ||
b940b3d7 JR |
600 | static int set_sctp(struct sk_buff *skb, struct sw_flow_key *flow_key, |
601 | const struct ovs_key_sctp *key, | |
602 | const struct ovs_key_sctp *mask) | |
10f72e3d | 603 | { |
b940b3d7 | 604 | unsigned int sctphoff = skb_transport_offset(skb); |
10f72e3d | 605 | struct sctphdr *sh; |
b940b3d7 | 606 | __le32 old_correct_csum, new_csum, old_csum; |
10f72e3d | 607 | int err; |
10f72e3d | 608 | |
5cce04b6 | 609 | err = skb_ensure_writable(skb, sctphoff + sizeof(struct sctphdr)); |
10f72e3d JS |
610 | if (unlikely(err)) |
611 | return err; | |
612 | ||
613 | sh = sctp_hdr(skb); | |
b940b3d7 JR |
614 | old_csum = sh->checksum; |
615 | old_correct_csum = sctp_compute_cksum(skb, sctphoff); | |
10f72e3d | 616 | |
e281bb23 JS |
617 | sh->source = OVS_MASKED(sh->source, key->sctp_src, mask->sctp_src); |
618 | sh->dest = OVS_MASKED(sh->dest, key->sctp_dst, mask->sctp_dst); | |
10f72e3d | 619 | |
b940b3d7 | 620 | new_csum = sctp_compute_cksum(skb, sctphoff); |
10f72e3d | 621 | |
b940b3d7 JR |
622 | /* Carry any checksum errors through. */ |
623 | sh->checksum = old_csum ^ old_correct_csum ^ new_csum; | |
10f72e3d | 624 | |
b940b3d7 JR |
625 | skb_clear_hash(skb); |
626 | flow_key->tp.src = sh->source; | |
627 | flow_key->tp.dst = sh->dest; | |
10f72e3d JS |
628 | |
629 | return 0; | |
630 | } | |
631 | ||
a94ebc39 JS |
632 | static int ovs_vport_output(OVS_VPORT_OUTPUT_PARAMS) |
633 | { | |
e9326797 | 634 | struct ovs_frag_data *data = this_cpu_ptr(&ovs_frag_data_storage); |
a94ebc39 JS |
635 | struct vport *vport = data->vport; |
636 | ||
637 | if (skb_cow_head(skb, data->l2_len) < 0) { | |
638 | kfree_skb(skb); | |
639 | return -ENOMEM; | |
640 | } | |
641 | ||
642 | __skb_dst_copy(skb, data->dst); | |
86c2eb45 | 643 | *OVS_GSO_CB(skb) = data->cb; |
a94ebc39 JS |
644 | ovs_skb_set_inner_protocol(skb, data->inner_protocol); |
645 | skb->vlan_tci = data->vlan_tci; | |
646 | skb->vlan_proto = data->vlan_proto; | |
647 | ||
648 | /* Reconstruct the MAC header. */ | |
649 | skb_push(skb, data->l2_len); | |
650 | memcpy(skb->data, &data->l2_data, data->l2_len); | |
ea3acd7a | 651 | skb_postpush_rcsum(skb, skb->data, data->l2_len); |
a94ebc39 JS |
652 | skb_reset_mac_header(skb); |
653 | ||
654 | ovs_vport_send(vport, skb); | |
655 | return 0; | |
656 | } | |
657 | ||
658 | static unsigned int | |
659 | ovs_dst_get_mtu(const struct dst_entry *dst) | |
660 | { | |
661 | return dst->dev->mtu; | |
662 | } | |
663 | ||
664 | static struct dst_ops ovs_dst_ops = { | |
665 | .family = AF_UNSPEC, | |
666 | .mtu = ovs_dst_get_mtu, | |
667 | }; | |
668 | ||
669 | /* prepare_frag() is called once per (larger-than-MTU) frame; its inverse is | |
670 | * ovs_vport_output(), which is called once per fragmented packet. | |
671 | */ | |
672 | static void prepare_frag(struct vport *vport, struct sk_buff *skb) | |
673 | { | |
674 | unsigned int hlen = skb_network_offset(skb); | |
675 | struct ovs_frag_data *data; | |
676 | ||
e9326797 | 677 | data = this_cpu_ptr(&ovs_frag_data_storage); |
a94ebc39 JS |
678 | data->dst = (unsigned long) skb_dst(skb); |
679 | data->vport = vport; | |
86c2eb45 | 680 | data->cb = *OVS_GSO_CB(skb); |
a94ebc39 JS |
681 | data->inner_protocol = ovs_skb_get_inner_protocol(skb); |
682 | data->vlan_tci = skb->vlan_tci; | |
683 | data->vlan_proto = skb->vlan_proto; | |
684 | data->l2_len = hlen; | |
685 | memcpy(&data->l2_data, skb->data, hlen); | |
686 | ||
687 | memset(IPCB(skb), 0, sizeof(struct inet_skb_parm)); | |
688 | skb_pull(skb, hlen); | |
689 | } | |
690 | ||
025813bb EB |
691 | static void ovs_fragment(struct net *net, struct vport *vport, |
692 | struct sk_buff *skb, u16 mru, __be16 ethertype) | |
a94ebc39 JS |
693 | { |
694 | if (skb_network_offset(skb) > MAX_L2_LEN) { | |
695 | OVS_NLERR(1, "L2 header too long to fragment"); | |
c05e2094 | 696 | goto err; |
a94ebc39 JS |
697 | } |
698 | ||
699 | if (ethertype == htons(ETH_P_IP)) { | |
700 | struct dst_entry ovs_dst; | |
701 | unsigned long orig_dst; | |
702 | ||
703 | prepare_frag(vport, skb); | |
704 | dst_init(&ovs_dst, &ovs_dst_ops, NULL, 1, | |
705 | DST_OBSOLETE_NONE, DST_NOCOUNT); | |
706 | ovs_dst.dev = vport->dev; | |
707 | ||
708 | orig_dst = (unsigned long) skb_dst(skb); | |
709 | skb_dst_set_noref(skb, &ovs_dst); | |
710 | IPCB(skb)->frag_max_size = mru; | |
711 | ||
0374bcbe | 712 | ip_do_fragment(net, skb->sk, skb, ovs_vport_output); |
a94ebc39 JS |
713 | refdst_drop(orig_dst); |
714 | } else if (ethertype == htons(ETH_P_IPV6)) { | |
715 | const struct nf_ipv6_ops *v6ops = nf_get_ipv6_ops(); | |
716 | unsigned long orig_dst; | |
717 | struct rt6_info ovs_rt; | |
718 | ||
719 | if (!v6ops) { | |
c05e2094 | 720 | goto err; |
a94ebc39 JS |
721 | } |
722 | ||
723 | prepare_frag(vport, skb); | |
724 | memset(&ovs_rt, 0, sizeof(ovs_rt)); | |
725 | dst_init(&ovs_rt.dst, &ovs_dst_ops, NULL, 1, | |
726 | DST_OBSOLETE_NONE, DST_NOCOUNT); | |
727 | ovs_rt.dst.dev = vport->dev; | |
728 | ||
729 | orig_dst = (unsigned long) skb_dst(skb); | |
730 | skb_dst_set_noref(skb, &ovs_rt.dst); | |
731 | IP6CB(skb)->frag_max_size = mru; | |
0643a78b PS |
732 | #ifdef HAVE_IP_LOCAL_OUT_TAKES_NET |
733 | v6ops->fragment(net, skb->sk, skb, ovs_vport_output); | |
734 | #else | |
a94ebc39 | 735 | v6ops->fragment(skb->sk, skb, ovs_vport_output); |
0643a78b | 736 | #endif |
a94ebc39 JS |
737 | refdst_drop(orig_dst); |
738 | } else { | |
739 | WARN_ONCE(1, "Failed fragment ->%s: eth=%04x, MRU=%d, MTU=%d.", | |
740 | ovs_vport_name(vport), ntohs(ethertype), mru, | |
741 | vport->dev->mtu); | |
c05e2094 | 742 | goto err; |
a94ebc39 | 743 | } |
c05e2094 JS |
744 | |
745 | return; | |
746 | err: | |
747 | kfree_skb(skb); | |
a94ebc39 | 748 | } |
a94ebc39 JS |
749 | |
750 | static void do_output(struct datapath *dp, struct sk_buff *skb, int out_port, | |
751 | struct sw_flow_key *key) | |
064af421 | 752 | { |
fe90efd9 | 753 | struct vport *vport = ovs_vport_rcu(dp, out_port); |
064af421 | 754 | |
a94ebc39 JS |
755 | if (likely(vport)) { |
756 | u16 mru = OVS_CB(skb)->mru; | |
4c7804f1 WT |
757 | u32 cutlen = OVS_CB(skb)->cutlen; |
758 | ||
759 | if (unlikely(cutlen > 0)) { | |
760 | if (skb->len - cutlen > ETH_HLEN) | |
761 | pskb_trim(skb, skb->len - cutlen); | |
762 | else | |
763 | pskb_trim(skb, ETH_HLEN); | |
764 | } | |
a94ebc39 JS |
765 | |
766 | if (likely(!mru || (skb->len <= mru + ETH_HLEN))) { | |
767 | ovs_vport_send(vport, skb); | |
768 | } else if (mru <= vport->dev->mtu) { | |
025813bb | 769 | struct net *net = ovs_dp_get_net(dp); |
a94ebc39 JS |
770 | __be16 ethertype = key->eth.type; |
771 | ||
772 | if (!is_flow_key_valid(key)) { | |
773 | if (eth_p_mpls(skb->protocol)) | |
774 | ethertype = ovs_skb_get_inner_protocol(skb); | |
775 | else | |
776 | ethertype = vlan_get_protocol(skb); | |
777 | } | |
778 | ||
025813bb | 779 | ovs_fragment(net, vport, skb, mru, ethertype); |
a94ebc39 JS |
780 | } else { |
781 | OVS_NLERR(true, "Cannot fragment IP frames"); | |
782 | kfree_skb(skb); | |
783 | } | |
784 | } else { | |
f15c8639 | 785 | kfree_skb(skb); |
a94ebc39 | 786 | } |
064af421 | 787 | } |
aad7cb91 | 788 | |
98403001 | 789 | static int output_userspace(struct datapath *dp, struct sk_buff *skb, |
0e469d3b | 790 | struct sw_flow_key *key, const struct nlattr *attr, |
4c7804f1 WT |
791 | const struct nlattr *actions, int actions_len, |
792 | uint32_t cutlen) | |
064af421 | 793 | { |
856081f6 | 794 | struct dp_upcall_info upcall; |
98403001 | 795 | const struct nlattr *a; |
aad7cb91 | 796 | int rem, err; |
856081f6 | 797 | |
0e469d3b | 798 | memset(&upcall, 0, sizeof(upcall)); |
df2c07f4 | 799 | upcall.cmd = OVS_PACKET_CMD_ACTION; |
a94ebc39 | 800 | upcall.mru = OVS_CB(skb)->mru; |
98403001 | 801 | |
aad7cb91 | 802 | SKB_INIT_FILL_METADATA_DST(skb); |
98403001 BP |
803 | for (a = nla_data(attr), rem = nla_len(attr); rem > 0; |
804 | a = nla_next(a, &rem)) { | |
805 | switch (nla_type(a)) { | |
806 | case OVS_USERSPACE_ATTR_USERDATA: | |
807 | upcall.userdata = a; | |
808 | break; | |
809 | ||
810 | case OVS_USERSPACE_ATTR_PID: | |
28aea917 | 811 | upcall.portid = nla_get_u32(a); |
98403001 | 812 | break; |
8b7ea2d4 WZ |
813 | |
814 | case OVS_USERSPACE_ATTR_EGRESS_TUN_PORT: { | |
815 | /* Get out tunnel info. */ | |
816 | struct vport *vport; | |
817 | ||
818 | vport = ovs_vport_rcu(dp, nla_get_u32(a)); | |
819 | if (vport) { | |
aad7cb91 PS |
820 | err = dev_fill_metadata_dst(vport->dev, skb); |
821 | if (!err) | |
822 | upcall.egress_tun_info = skb_tunnel_info(skb); | |
8b7ea2d4 | 823 | } |
e23775f2 | 824 | |
8b7ea2d4 | 825 | break; |
98403001 | 826 | } |
8b7ea2d4 | 827 | |
0e469d3b NM |
828 | case OVS_USERSPACE_ATTR_ACTIONS: { |
829 | /* Include actions. */ | |
830 | upcall.actions = actions; | |
831 | upcall.actions_len = actions_len; | |
832 | break; | |
833 | } | |
834 | ||
8b7ea2d4 | 835 | } /* End of switch. */ |
98403001 BP |
836 | } |
837 | ||
aad7cb91 PS |
838 | err = ovs_dp_upcall(dp, skb, key, &upcall, cutlen); |
839 | SKB_RESTORE_FILL_METADATA_DST(skb); | |
840 | return err; | |
064af421 BP |
841 | } |
842 | ||
6ff686f2 | 843 | static int sample(struct datapath *dp, struct sk_buff *skb, |
0e469d3b NM |
844 | struct sw_flow_key *key, const struct nlattr *attr, |
845 | const struct nlattr *actions, int actions_len) | |
6ff686f2 PS |
846 | { |
847 | const struct nlattr *acts_list = NULL; | |
848 | const struct nlattr *a; | |
849 | int rem; | |
4c7804f1 | 850 | u32 cutlen = 0; |
6ff686f2 PS |
851 | |
852 | for (a = nla_data(attr), rem = nla_len(attr); rem > 0; | |
853 | a = nla_next(a, &rem)) { | |
c02c4967 WZ |
854 | u32 probability; |
855 | ||
6ff686f2 PS |
856 | switch (nla_type(a)) { |
857 | case OVS_SAMPLE_ATTR_PROBABILITY: | |
c02c4967 WZ |
858 | probability = nla_get_u32(a); |
859 | if (!probability || prandom_u32() > probability) | |
6ff686f2 PS |
860 | return 0; |
861 | break; | |
862 | ||
863 | case OVS_SAMPLE_ATTR_ACTIONS: | |
864 | acts_list = a; | |
865 | break; | |
866 | } | |
867 | } | |
868 | ||
fbf4f74d SH |
869 | rem = nla_len(acts_list); |
870 | a = nla_data(acts_list); | |
871 | ||
d7ff93d7 AZ |
872 | /* Actions list is empty, do nothing */ |
873 | if (unlikely(!rem)) | |
874 | return 0; | |
e16138e2 | 875 | |
d7ff93d7 | 876 | /* The only known usage of sample action is having a single user-space |
4c7804f1 | 877 | * action, or having a truncate action followed by a single user-space |
d7ff93d7 AZ |
878 | * action. Treat this usage as a special case. |
879 | * The output_userspace() should clone the skb to be sent to the | |
e74d4817 PS |
880 | * user space. This skb will be consumed by its caller. |
881 | */ | |
4c7804f1 WT |
882 | if (unlikely(nla_type(a) == OVS_ACTION_ATTR_TRUNC)) { |
883 | struct ovs_action_trunc *trunc = nla_data(a); | |
884 | ||
885 | if (skb->len > trunc->max_len) | |
886 | cutlen = skb->len - trunc->max_len; | |
887 | ||
888 | a = nla_next(a, &rem); | |
889 | } | |
890 | ||
d7ff93d7 | 891 | if (likely(nla_type(a) == OVS_ACTION_ATTR_USERSPACE && |
684b5f5d | 892 | nla_is_last(a, rem))) |
4c7804f1 WT |
893 | return output_userspace(dp, skb, key, a, actions, |
894 | actions_len, cutlen); | |
d7ff93d7 AZ |
895 | |
896 | skb = skb_clone(skb, GFP_ATOMIC); | |
897 | if (!skb) | |
898 | /* Skip the sample action when out of memory. */ | |
899 | return 0; | |
900 | ||
e74d4817 | 901 | if (!add_deferred_actions(skb, key, a)) { |
2c8c4fb7 AZ |
902 | if (net_ratelimit()) |
903 | pr_warn("%s: deferred actions limit reached, dropping sample action\n", | |
904 | ovs_dp_name(dp)); | |
fbf4f74d | 905 | |
2c8c4fb7 AZ |
906 | kfree_skb(skb); |
907 | } | |
2c8c4fb7 | 908 | return 0; |
6ff686f2 PS |
909 | } |
910 | ||
e74d4817 PS |
911 | static void execute_hash(struct sk_buff *skb, struct sw_flow_key *key, |
912 | const struct nlattr *attr) | |
7804df20 | 913 | { |
7804df20 AZ |
914 | struct ovs_action_hash *hash_act = nla_data(attr); |
915 | u32 hash = 0; | |
916 | ||
917 | /* OVS_HASH_ALG_L4 is the only possible hash algorithm. */ | |
e2f3178f | 918 | hash = skb_get_hash(skb); |
7804df20 AZ |
919 | hash = jhash_1word(hash, hash_act->hash_basis); |
920 | if (!hash) | |
921 | hash = 0x1; | |
922 | ||
923 | key->ovs_flow_hash = hash; | |
924 | } | |
925 | ||
b940b3d7 JR |
926 | static int execute_set_action(struct sk_buff *skb, |
927 | struct sw_flow_key *flow_key, | |
928 | const struct nlattr *a) | |
929 | { | |
930 | /* Only tunnel set execution is supported without a mask. */ | |
931 | if (nla_type(a) == OVS_KEY_ATTR_TUNNEL_INFO) { | |
e23775f2 PS |
932 | struct ovs_tunnel_info *tun = nla_data(a); |
933 | ||
934 | ovs_skb_dst_drop(skb); | |
935 | ovs_dst_hold((struct dst_entry *)tun->tun_dst); | |
936 | ovs_skb_dst_set(skb, (struct dst_entry *)tun->tun_dst); | |
b940b3d7 JR |
937 | return 0; |
938 | } | |
939 | ||
940 | return -EINVAL; | |
b940b3d7 JR |
941 | } |
942 | ||
943 | /* Mask is at the midpoint of the data. */ | |
944 | #define get_mask(a, type) ((const type)nla_data(a) + 1) | |
945 | ||
946 | static int execute_masked_set_action(struct sk_buff *skb, | |
947 | struct sw_flow_key *flow_key, | |
948 | const struct nlattr *a) | |
4edb9ae9 | 949 | { |
15c39847 | 950 | int err = 0; |
4edb9ae9 | 951 | |
b940b3d7 | 952 | switch (nla_type(a)) { |
abff858b | 953 | case OVS_KEY_ATTR_PRIORITY: |
e281bb23 JS |
954 | OVS_SET_MASKED(skb->priority, nla_get_u32(a), |
955 | *get_mask(a, u32 *)); | |
b940b3d7 | 956 | flow_key->phy.priority = skb->priority; |
abff858b PS |
957 | break; |
958 | ||
72e8bf28 | 959 | case OVS_KEY_ATTR_SKB_MARK: |
e281bb23 | 960 | OVS_SET_MASKED(skb->mark, nla_get_u32(a), *get_mask(a, u32 *)); |
b940b3d7 | 961 | flow_key->phy.skb_mark = skb->mark; |
72e8bf28 AA |
962 | break; |
963 | ||
f0cd669f | 964 | case OVS_KEY_ATTR_TUNNEL_INFO: |
b940b3d7 JR |
965 | /* Masked data not supported for tunnel. */ |
966 | err = -EINVAL; | |
4edb9ae9 PS |
967 | break; |
968 | ||
969 | case OVS_KEY_ATTR_ETHERNET: | |
b940b3d7 JR |
970 | err = set_eth_addr(skb, flow_key, nla_data(a), |
971 | get_mask(a, struct ovs_key_ethernet *)); | |
4edb9ae9 PS |
972 | break; |
973 | ||
974 | case OVS_KEY_ATTR_IPV4: | |
b940b3d7 JR |
975 | err = set_ipv4(skb, flow_key, nla_data(a), |
976 | get_mask(a, struct ovs_key_ipv4 *)); | |
4edb9ae9 PS |
977 | break; |
978 | ||
bc7a5acd | 979 | case OVS_KEY_ATTR_IPV6: |
b940b3d7 JR |
980 | err = set_ipv6(skb, flow_key, nla_data(a), |
981 | get_mask(a, struct ovs_key_ipv6 *)); | |
bc7a5acd AA |
982 | break; |
983 | ||
4edb9ae9 | 984 | case OVS_KEY_ATTR_TCP: |
b940b3d7 JR |
985 | err = set_tcp(skb, flow_key, nla_data(a), |
986 | get_mask(a, struct ovs_key_tcp *)); | |
4edb9ae9 PS |
987 | break; |
988 | ||
989 | case OVS_KEY_ATTR_UDP: | |
b940b3d7 JR |
990 | err = set_udp(skb, flow_key, nla_data(a), |
991 | get_mask(a, struct ovs_key_udp *)); | |
4edb9ae9 | 992 | break; |
10f72e3d JS |
993 | |
994 | case OVS_KEY_ATTR_SCTP: | |
b940b3d7 JR |
995 | err = set_sctp(skb, flow_key, nla_data(a), |
996 | get_mask(a, struct ovs_key_sctp *)); | |
10f72e3d | 997 | break; |
ccf43786 SH |
998 | |
999 | case OVS_KEY_ATTR_MPLS: | |
b940b3d7 JR |
1000 | err = set_mpls(skb, flow_key, nla_data(a), get_mask(a, |
1001 | __be32 *)); | |
ccf43786 | 1002 | break; |
a94ebc39 JS |
1003 | |
1004 | case OVS_KEY_ATTR_CT_STATE: | |
1005 | case OVS_KEY_ATTR_CT_ZONE: | |
372ce973 | 1006 | case OVS_KEY_ATTR_CT_MARK: |
c05e2094 | 1007 | case OVS_KEY_ATTR_CT_LABELS: |
a94ebc39 JS |
1008 | err = -EINVAL; |
1009 | break; | |
4edb9ae9 | 1010 | } |
15c39847 | 1011 | |
4edb9ae9 PS |
1012 | return err; |
1013 | } | |
1014 | ||
a6059080 | 1015 | static int execute_recirc(struct datapath *dp, struct sk_buff *skb, |
7d16c847 PS |
1016 | struct sw_flow_key *key, |
1017 | const struct nlattr *a, int rem) | |
a6059080 | 1018 | { |
e74d4817 PS |
1019 | struct deferred_action *da; |
1020 | ||
1021 | if (!is_flow_key_valid(key)) { | |
867e37ba AZ |
1022 | int err; |
1023 | ||
e74d4817 | 1024 | err = ovs_flow_key_update(skb, key); |
867e37ba AZ |
1025 | if (err) |
1026 | return err; | |
867e37ba | 1027 | } |
e74d4817 | 1028 | BUG_ON(!is_flow_key_valid(key)); |
a6059080 | 1029 | |
684b5f5d | 1030 | if (!nla_is_last(a, rem)) { |
e16138e2 | 1031 | /* Recirc action is the not the last action |
e74d4817 PS |
1032 | * of the action list, need to clone the skb. |
1033 | */ | |
e16138e2 PS |
1034 | skb = skb_clone(skb, GFP_ATOMIC); |
1035 | ||
1036 | /* Skip the recirc action when out of memory, but | |
e74d4817 PS |
1037 | * continue on with the rest of the action list. |
1038 | */ | |
e16138e2 PS |
1039 | if (!skb) |
1040 | return 0; | |
2c8c4fb7 | 1041 | } |
a6059080 | 1042 | |
e74d4817 PS |
1043 | da = add_deferred_actions(skb, key, NULL); |
1044 | if (da) { | |
1045 | da->pkt_key.recirc_id = nla_get_u32(a); | |
2c8c4fb7 AZ |
1046 | } else { |
1047 | kfree_skb(skb); | |
1048 | ||
1049 | if (net_ratelimit()) | |
1050 | pr_warn("%s: deferred action limit reached, drop recirc action\n", | |
1051 | ovs_dp_name(dp)); | |
867e37ba | 1052 | } |
a6059080 AZ |
1053 | |
1054 | return 0; | |
1055 | } | |
1056 | ||
064af421 | 1057 | /* Execute a list of actions against 'skb'. */ |
871dfe07 | 1058 | static int do_execute_actions(struct datapath *dp, struct sk_buff *skb, |
e74d4817 PS |
1059 | struct sw_flow_key *key, |
1060 | const struct nlattr *attr, int len) | |
064af421 BP |
1061 | { |
1062 | /* Every output action needs a separate clone of 'skb', but the common | |
1063 | * case is just a single output action, so that doing a clone and | |
1064 | * then freeing the original skbuff is wasteful. So the following code | |
e74d4817 PS |
1065 | * is slightly obscure just to avoid that. |
1066 | */ | |
064af421 | 1067 | int prev_port = -1; |
cdee00fd | 1068 | const struct nlattr *a; |
10db8b20 | 1069 | int rem; |
72b06300 | 1070 | |
6ff686f2 | 1071 | for (a = attr, rem = len; rem > 0; |
a4af2475 | 1072 | a = nla_next(a, &rem)) { |
10db8b20 JG |
1073 | int err = 0; |
1074 | ||
fe90efd9 AZ |
1075 | if (unlikely(prev_port != -1)) { |
1076 | struct sk_buff *out_skb = skb_clone(skb, GFP_ATOMIC); | |
1077 | ||
1078 | if (out_skb) | |
a94ebc39 | 1079 | do_output(dp, out_skb, prev_port, key); |
fe90efd9 | 1080 | |
4c7804f1 | 1081 | OVS_CB(skb)->cutlen = 0; |
064af421 BP |
1082 | prev_port = -1; |
1083 | } | |
1084 | ||
cdee00fd | 1085 | switch (nla_type(a)) { |
df2c07f4 | 1086 | case OVS_ACTION_ATTR_OUTPUT: |
cdee00fd | 1087 | prev_port = nla_get_u32(a); |
064af421 BP |
1088 | break; |
1089 | ||
4c7804f1 WT |
1090 | case OVS_ACTION_ATTR_TRUNC: { |
1091 | struct ovs_action_trunc *trunc = nla_data(a); | |
1092 | ||
1093 | if (skb->len > trunc->max_len) | |
1094 | OVS_CB(skb)->cutlen = skb->len - trunc->max_len; | |
1095 | break; | |
1096 | } | |
1097 | ||
df2c07f4 | 1098 | case OVS_ACTION_ATTR_USERSPACE: |
4c7804f1 WT |
1099 | output_userspace(dp, skb, key, a, attr, |
1100 | len, OVS_CB(skb)->cutlen); | |
1101 | OVS_CB(skb)->cutlen = 0; | |
064af421 | 1102 | break; |
7804df20 AZ |
1103 | |
1104 | case OVS_ACTION_ATTR_HASH: | |
e74d4817 | 1105 | execute_hash(skb, key, a); |
7804df20 | 1106 | break; |
064af421 | 1107 | |
ccf43786 | 1108 | case OVS_ACTION_ATTR_PUSH_MPLS: |
e74d4817 | 1109 | err = push_mpls(skb, key, nla_data(a)); |
ccf43786 SH |
1110 | break; |
1111 | ||
1112 | case OVS_ACTION_ATTR_POP_MPLS: | |
e74d4817 | 1113 | err = pop_mpls(skb, key, nla_get_be16(a)); |
ccf43786 SH |
1114 | break; |
1115 | ||
fea393b1 | 1116 | case OVS_ACTION_ATTR_PUSH_VLAN: |
e74d4817 | 1117 | err = push_vlan(skb, key, nla_data(a)); |
064af421 BP |
1118 | break; |
1119 | ||
fea393b1 | 1120 | case OVS_ACTION_ATTR_POP_VLAN: |
e74d4817 | 1121 | err = pop_vlan(skb, key); |
064af421 BP |
1122 | break; |
1123 | ||
e16138e2 | 1124 | case OVS_ACTION_ATTR_RECIRC: |
e74d4817 | 1125 | err = execute_recirc(dp, skb, key, a, rem); |
684b5f5d | 1126 | if (nla_is_last(a, rem)) { |
867e37ba AZ |
1127 | /* If this is the last action, the skb has |
1128 | * been consumed or freed. | |
e74d4817 PS |
1129 | * Return immediately. |
1130 | */ | |
867e37ba AZ |
1131 | return err; |
1132 | } | |
a6059080 | 1133 | break; |
a6059080 | 1134 | |
4edb9ae9 | 1135 | case OVS_ACTION_ATTR_SET: |
e74d4817 | 1136 | err = execute_set_action(skb, key, nla_data(a)); |
064af421 | 1137 | break; |
c1c9c9c4 | 1138 | |
b940b3d7 JR |
1139 | case OVS_ACTION_ATTR_SET_MASKED: |
1140 | case OVS_ACTION_ATTR_SET_TO_MASKED: | |
1141 | err = execute_masked_set_action(skb, key, nla_data(a)); | |
1142 | break; | |
1143 | ||
6ff686f2 | 1144 | case OVS_ACTION_ATTR_SAMPLE: |
0e469d3b | 1145 | err = sample(dp, skb, key, a, attr, len); |
6ff686f2 | 1146 | break; |
a94ebc39 JS |
1147 | |
1148 | case OVS_ACTION_ATTR_CT: | |
c05e2094 JS |
1149 | if (!is_flow_key_valid(key)) { |
1150 | err = ovs_flow_key_update(skb, key); | |
1151 | if (err) | |
1152 | return err; | |
1153 | } | |
1154 | ||
a94ebc39 JS |
1155 | err = ovs_ct_execute(ovs_dp_get_net(dp), skb, key, |
1156 | nla_data(a)); | |
1157 | ||
1158 | /* Hide stolen IP fragments from user space. */ | |
c05e2094 JS |
1159 | if (err) |
1160 | return err == -EINPROGRESS ? 0 : err; | |
a94ebc39 | 1161 | break; |
6ff686f2 | 1162 | } |
15c39847 | 1163 | |
10db8b20 JG |
1164 | if (unlikely(err)) { |
1165 | kfree_skb(skb); | |
1166 | return err; | |
1167 | } | |
064af421 | 1168 | } |
6c222e55 | 1169 | |
fbf4f74d | 1170 | if (prev_port != -1) |
a94ebc39 | 1171 | do_output(dp, skb, prev_port, key); |
fbf4f74d | 1172 | else |
5b95ab0e | 1173 | consume_skb(skb); |
10db8b20 | 1174 | |
a5225dd6 | 1175 | return 0; |
064af421 | 1176 | } |
871dfe07 | 1177 | |
2c8c4fb7 AZ |
1178 | static void process_deferred_actions(struct datapath *dp) |
1179 | { | |
1180 | struct action_fifo *fifo = this_cpu_ptr(action_fifos); | |
1181 | ||
1182 | /* Do not touch the FIFO in case there is no deferred actions. */ | |
1183 | if (action_fifo_is_empty(fifo)) | |
1184 | return; | |
1185 | ||
1186 | /* Finishing executing all deferred actions. */ | |
1187 | do { | |
1188 | struct deferred_action *da = action_fifo_get(fifo); | |
1189 | struct sk_buff *skb = da->skb; | |
7d16c847 | 1190 | struct sw_flow_key *key = &da->pkt_key; |
2c8c4fb7 AZ |
1191 | const struct nlattr *actions = da->actions; |
1192 | ||
1193 | if (actions) | |
7d16c847 | 1194 | do_execute_actions(dp, skb, key, actions, |
2c8c4fb7 AZ |
1195 | nla_len(actions)); |
1196 | else | |
7d16c847 | 1197 | ovs_dp_process_packet(skb, key); |
2c8c4fb7 AZ |
1198 | } while (!action_fifo_is_empty(fifo)); |
1199 | ||
1200 | /* Reset FIFO for the next packet. */ | |
1201 | action_fifo_init(fifo); | |
1202 | } | |
1203 | ||
871dfe07 | 1204 | /* Execute a list of actions against 'skb'. */ |
2c8c4fb7 | 1205 | int ovs_execute_actions(struct datapath *dp, struct sk_buff *skb, |
7d16c847 PS |
1206 | const struct sw_flow_actions *acts, |
1207 | struct sw_flow_key *key) | |
2c8c4fb7 | 1208 | { |
77a9a338 PS |
1209 | static const int ovs_recursion_limit = 4; |
1210 | int err, level; | |
2c8c4fb7 | 1211 | |
77a9a338 PS |
1212 | level = __this_cpu_inc_return(exec_actions_level); |
1213 | if (unlikely(level > ovs_recursion_limit)) { | |
1214 | net_crit_ratelimited("ovs: recursion limit reached on datapath %s, probable configuration error\n", | |
1215 | ovs_dp_name(dp)); | |
2c8c4fb7 | 1216 | kfree_skb(skb); |
77a9a338 PS |
1217 | err = -ENETDOWN; |
1218 | goto out; | |
2c8c4fb7 AZ |
1219 | } |
1220 | ||
7d16c847 PS |
1221 | err = do_execute_actions(dp, skb, key, |
1222 | acts->actions, acts->actions_len); | |
2c8c4fb7 | 1223 | |
77a9a338 | 1224 | if (level == 1) |
2c8c4fb7 AZ |
1225 | process_deferred_actions(dp); |
1226 | ||
77a9a338 PS |
1227 | out: |
1228 | __this_cpu_dec(exec_actions_level); | |
2c8c4fb7 AZ |
1229 | return err; |
1230 | } | |
1231 | ||
1232 | int action_fifos_init(void) | |
1233 | { | |
1234 | action_fifos = alloc_percpu(struct action_fifo); | |
1235 | if (!action_fifos) | |
1236 | return -ENOMEM; | |
1237 | ||
1238 | return 0; | |
1239 | } | |
1240 | ||
1241 | void action_fifos_exit(void) | |
60759b2b | 1242 | { |
2c8c4fb7 | 1243 | free_percpu(action_fifos); |
871dfe07 | 1244 | } |