]>
Commit | Line | Data |
---|---|---|
064af421 | 1 | /* |
58afdef1 | 2 | * Copyright (c) 2007-2017 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" |
96b82f6d | 45 | #include "flow_netlink.h" |
064af421 | 46 | |
e74d4817 PS |
47 | static int do_execute_actions(struct datapath *dp, struct sk_buff *skb, |
48 | struct sw_flow_key *key, | |
49 | const struct nlattr *attr, int len); | |
50 | ||
2c8c4fb7 AZ |
51 | struct deferred_action { |
52 | struct sk_buff *skb; | |
53 | const struct nlattr *actions; | |
615fa7ba | 54 | int actions_len; |
2c8c4fb7 AZ |
55 | |
56 | /* Store pkt_key clone when creating deferred action. */ | |
57 | struct sw_flow_key pkt_key; | |
58 | }; | |
59 | ||
a94ebc39 JS |
60 | #define MAX_L2_LEN (VLAN_ETH_HLEN + 3 * MPLS_HLEN) |
61 | struct ovs_frag_data { | |
62 | unsigned long dst; | |
63 | struct vport *vport; | |
86c2eb45 | 64 | struct ovs_gso_cb cb; |
a94ebc39 | 65 | __be16 inner_protocol; |
279ecabe YHW |
66 | u16 network_offset; /* valid only for MPLS */ |
67 | u16 vlan_tci; | |
a94ebc39 JS |
68 | __be16 vlan_proto; |
69 | unsigned int l2_len; | |
f80d0524 | 70 | u8 mac_proto; |
a94ebc39 JS |
71 | u8 l2_data[MAX_L2_LEN]; |
72 | }; | |
73 | ||
74 | static DEFINE_PER_CPU(struct ovs_frag_data, ovs_frag_data_storage); | |
75 | ||
2c8c4fb7 | 76 | #define DEFERRED_ACTION_FIFO_SIZE 10 |
615539ab LR |
77 | #define OVS_RECURSION_LIMIT 4 |
78 | #define OVS_DEFERRED_ACTION_THRESHOLD (OVS_RECURSION_LIMIT - 2) | |
2c8c4fb7 AZ |
79 | struct action_fifo { |
80 | int head; | |
81 | int tail; | |
82 | /* Deferred action fifo queue storage. */ | |
83 | struct deferred_action fifo[DEFERRED_ACTION_FIFO_SIZE]; | |
84 | }; | |
85 | ||
58afdef1 | 86 | struct action_flow_keys { |
615539ab LR |
87 | struct sw_flow_key key[OVS_DEFERRED_ACTION_THRESHOLD]; |
88 | }; | |
77a9a338 | 89 | |
615539ab | 90 | static struct action_fifo __percpu *action_fifos; |
58afdef1 | 91 | static struct action_flow_keys __percpu *flow_keys; |
2c8c4fb7 AZ |
92 | static DEFINE_PER_CPU(int, exec_actions_level); |
93 | ||
58afdef1 AZ |
94 | /* Make a clone of the 'key', using the pre-allocated percpu 'flow_keys' |
95 | * space. Return NULL if out of key spaces. | |
96 | */ | |
97 | static struct sw_flow_key *clone_key(const struct sw_flow_key *key_) | |
98 | { | |
99 | struct action_flow_keys *keys = this_cpu_ptr(flow_keys); | |
100 | int level = this_cpu_read(exec_actions_level); | |
101 | struct sw_flow_key *key = NULL; | |
102 | ||
103 | if (level <= OVS_DEFERRED_ACTION_THRESHOLD) { | |
104 | key = &keys->key[level - 1]; | |
105 | *key = *key_; | |
106 | } | |
107 | ||
108 | return key; | |
109 | } | |
110 | ||
2c8c4fb7 AZ |
111 | static void action_fifo_init(struct action_fifo *fifo) |
112 | { | |
113 | fifo->head = 0; | |
114 | fifo->tail = 0; | |
115 | } | |
116 | ||
f1f60b85 | 117 | static bool action_fifo_is_empty(const struct action_fifo *fifo) |
2c8c4fb7 AZ |
118 | { |
119 | return (fifo->head == fifo->tail); | |
120 | } | |
121 | ||
e74d4817 | 122 | static struct deferred_action *action_fifo_get(struct action_fifo *fifo) |
2c8c4fb7 AZ |
123 | { |
124 | if (action_fifo_is_empty(fifo)) | |
125 | return NULL; | |
126 | ||
127 | return &fifo->fifo[fifo->tail++]; | |
128 | } | |
129 | ||
e74d4817 | 130 | static struct deferred_action *action_fifo_put(struct action_fifo *fifo) |
2c8c4fb7 AZ |
131 | { |
132 | if (fifo->head >= DEFERRED_ACTION_FIFO_SIZE - 1) | |
133 | return NULL; | |
134 | ||
135 | return &fifo->fifo[fifo->head++]; | |
136 | } | |
137 | ||
e74d4817 PS |
138 | /* Return queue entry if fifo is not full */ |
139 | static struct deferred_action *add_deferred_actions(struct sk_buff *skb, | |
615fa7ba AZ |
140 | const struct sw_flow_key *key, |
141 | const struct nlattr *actions, | |
142 | const int actions_len) | |
2c8c4fb7 AZ |
143 | { |
144 | struct action_fifo *fifo; | |
145 | struct deferred_action *da; | |
146 | ||
147 | fifo = this_cpu_ptr(action_fifos); | |
148 | da = action_fifo_put(fifo); | |
149 | if (da) { | |
150 | da->skb = skb; | |
615fa7ba AZ |
151 | da->actions = actions; |
152 | da->actions_len = actions_len; | |
e74d4817 | 153 | da->pkt_key = *key; |
2c8c4fb7 AZ |
154 | } |
155 | ||
e74d4817 | 156 | return da; |
e16138e2 PS |
157 | } |
158 | ||
e74d4817 | 159 | static void invalidate_flow_key(struct sw_flow_key *key) |
e16138e2 | 160 | { |
16d76400 | 161 | key->mac_proto |= SW_FLOW_KEY_INVALID; |
e16138e2 PS |
162 | } |
163 | ||
f1f60b85 | 164 | static bool is_flow_key_valid(const struct sw_flow_key *key) |
e16138e2 | 165 | { |
16d76400 | 166 | return !(key->mac_proto & SW_FLOW_KEY_INVALID); |
e16138e2 PS |
167 | } |
168 | ||
d4004f8d AZ |
169 | static int clone_execute(struct datapath *dp, struct sk_buff *skb, |
170 | struct sw_flow_key *key, | |
171 | u32 recirc_id, | |
172 | const struct nlattr *actions, int len, | |
173 | bool last, bool clone_flow_key); | |
174 | ||
b51367aa PS |
175 | static void update_ethertype(struct sk_buff *skb, struct ethhdr *hdr, |
176 | __be16 ethertype) | |
177 | { | |
178 | if (skb->ip_summed == CHECKSUM_COMPLETE) { | |
179 | __be16 diff[] = { ~(hdr->h_proto), ethertype }; | |
180 | ||
181 | skb->csum = ~csum_partial((char *)diff, sizeof(diff), | |
182 | ~skb->csum); | |
183 | } | |
184 | ||
185 | hdr->h_proto = ethertype; | |
186 | } | |
187 | ||
e74d4817 | 188 | static int push_mpls(struct sk_buff *skb, struct sw_flow_key *key, |
ccf43786 SH |
189 | const struct ovs_action_push_mpls *mpls) |
190 | { | |
65590144 | 191 | struct mpls_shim_hdr *new_mpls_lse; |
ccf43786 | 192 | |
2baf0e0c | 193 | /* Networking stack do not allow simultaneous Tunnel and MPLS GSO. */ |
8063e095 | 194 | if (skb->encapsulation) |
2baf0e0c PS |
195 | return -ENOTSUPP; |
196 | ||
ccf43786 SH |
197 | if (skb_cow_head(skb, MPLS_HLEN) < 0) |
198 | return -ENOMEM; | |
199 | ||
65590144 YHW |
200 | if (!ovs_skb_get_inner_protocol(skb)) { |
201 | skb_set_inner_network_header(skb, skb->mac_len); | |
202 | ovs_skb_set_inner_protocol(skb, skb->protocol); | |
203 | } | |
204 | ||
ccf43786 SH |
205 | skb_push(skb, MPLS_HLEN); |
206 | memmove(skb_mac_header(skb) - MPLS_HLEN, skb_mac_header(skb), | |
207 | skb->mac_len); | |
208 | skb_reset_mac_header(skb); | |
65590144 YHW |
209 | #ifdef MPLS_HEADER_IS_L3 |
210 | skb_set_network_header(skb, skb->mac_len); | |
211 | #endif | |
ccf43786 | 212 | |
65590144 YHW |
213 | new_mpls_lse = mpls_hdr(skb); |
214 | new_mpls_lse->label_stack_entry = mpls->mpls_lse; | |
ccf43786 | 215 | |
ea3acd7a | 216 | skb_postpush_rcsum(skb, new_mpls_lse, MPLS_HLEN); |
ccf43786 | 217 | |
edd6d098 YY |
218 | if (ovs_key_mac_proto(key) == MAC_PROTO_ETHERNET) |
219 | update_ethertype(skb, eth_hdr(skb), mpls->mpls_ethertype); | |
ccf43786 | 220 | skb->protocol = mpls->mpls_ethertype; |
2baf0e0c | 221 | |
e74d4817 | 222 | invalidate_flow_key(key); |
ccf43786 SH |
223 | return 0; |
224 | } | |
225 | ||
e74d4817 PS |
226 | static int pop_mpls(struct sk_buff *skb, struct sw_flow_key *key, |
227 | const __be16 ethertype) | |
ccf43786 | 228 | { |
ccf43786 SH |
229 | int err; |
230 | ||
5cce04b6 | 231 | err = skb_ensure_writable(skb, skb->mac_len + MPLS_HLEN); |
ccf43786 SH |
232 | if (unlikely(err)) |
233 | return err; | |
234 | ||
65590144 | 235 | skb_postpull_rcsum(skb, mpls_hdr(skb), MPLS_HLEN); |
ccf43786 SH |
236 | |
237 | memmove(skb_mac_header(skb) + MPLS_HLEN, skb_mac_header(skb), | |
238 | skb->mac_len); | |
239 | ||
240 | __skb_pull(skb, MPLS_HLEN); | |
241 | skb_reset_mac_header(skb); | |
65590144 | 242 | skb_set_network_header(skb, skb->mac_len); |
ccf43786 | 243 | |
edd6d098 YY |
244 | if (ovs_key_mac_proto(key) == MAC_PROTO_ETHERNET) { |
245 | struct ethhdr *hdr; | |
246 | ||
65590144 | 247 | /* mpls_hdr() is used to locate the ethertype |
edd6d098 YY |
248 | * field correctly in the presence of VLAN tags. |
249 | */ | |
65590144 | 250 | hdr = (struct ethhdr *)((void*)mpls_hdr(skb) - ETH_HLEN); |
edd6d098 YY |
251 | update_ethertype(skb, hdr, ethertype); |
252 | } | |
ccf43786 SH |
253 | if (eth_p_mpls(skb->protocol)) |
254 | skb->protocol = ethertype; | |
2baf0e0c | 255 | |
e74d4817 | 256 | invalidate_flow_key(key); |
ccf43786 SH |
257 | return 0; |
258 | } | |
259 | ||
b940b3d7 JR |
260 | static int set_mpls(struct sk_buff *skb, struct sw_flow_key *flow_key, |
261 | const __be32 *mpls_lse, const __be32 *mask) | |
ccf43786 | 262 | { |
65590144 | 263 | struct mpls_shim_hdr *stack; |
b940b3d7 | 264 | __be32 lse; |
ccf43786 SH |
265 | int err; |
266 | ||
5cce04b6 | 267 | err = skb_ensure_writable(skb, skb->mac_len + MPLS_HLEN); |
ccf43786 SH |
268 | if (unlikely(err)) |
269 | return err; | |
270 | ||
65590144 YHW |
271 | stack = mpls_hdr(skb); |
272 | lse = OVS_MASKED(stack->label_stack_entry, *mpls_lse, *mask); | |
ccf43786 | 273 | if (skb->ip_summed == CHECKSUM_COMPLETE) { |
65590144 | 274 | __be32 diff[] = { ~(stack->label_stack_entry), lse }; |
b940b3d7 | 275 | |
ccf43786 SH |
276 | skb->csum = ~csum_partial((char *)diff, sizeof(diff), |
277 | ~skb->csum); | |
278 | } | |
279 | ||
65590144 | 280 | stack->label_stack_entry = lse; |
b940b3d7 | 281 | flow_key->mpls.top_lse = lse; |
ccf43786 SH |
282 | return 0; |
283 | } | |
284 | ||
e74d4817 | 285 | static int pop_vlan(struct sk_buff *skb, struct sw_flow_key *key) |
064af421 | 286 | { |
d9065a90 | 287 | int err; |
10db8b20 | 288 | |
97894370 | 289 | err = skb_vlan_pop(skb); |
d7efce7b | 290 | if (skb_vlan_tag_present(skb)) { |
97894370 | 291 | invalidate_flow_key(key); |
d7efce7b YY |
292 | } else { |
293 | key->eth.vlan.tci = 0; | |
294 | key->eth.vlan.tpid = 0; | |
295 | } | |
97894370 | 296 | return err; |
d9065a90 PS |
297 | } |
298 | ||
e74d4817 PS |
299 | static int push_vlan(struct sk_buff *skb, struct sw_flow_key *key, |
300 | const struct ovs_action_push_vlan *vlan) | |
d9065a90 | 301 | { |
d7efce7b | 302 | if (skb_vlan_tag_present(skb)) { |
e74d4817 | 303 | invalidate_flow_key(key); |
d7efce7b YY |
304 | } else { |
305 | key->eth.vlan.tci = vlan->vlan_tci; | |
306 | key->eth.vlan.tpid = vlan->vlan_tpid; | |
307 | } | |
97894370 TG |
308 | return skb_vlan_push(skb, vlan->vlan_tpid, |
309 | ntohs(vlan->vlan_tci) & ~VLAN_TAG_PRESENT); | |
064af421 BP |
310 | } |
311 | ||
b940b3d7 JR |
312 | /* 'src' is already properly masked. */ |
313 | static void ether_addr_copy_masked(u8 *dst_, const u8 *src_, const u8 *mask_) | |
314 | { | |
315 | u16 *dst = (u16 *)dst_; | |
316 | const u16 *src = (const u16 *)src_; | |
317 | const u16 *mask = (const u16 *)mask_; | |
318 | ||
e281bb23 JS |
319 | OVS_SET_MASKED(dst[0], src[0], mask[0]); |
320 | OVS_SET_MASKED(dst[1], src[1], mask[1]); | |
321 | OVS_SET_MASKED(dst[2], src[2], mask[2]); | |
b940b3d7 JR |
322 | } |
323 | ||
324 | static int set_eth_addr(struct sk_buff *skb, struct sw_flow_key *flow_key, | |
325 | const struct ovs_key_ethernet *key, | |
326 | const struct ovs_key_ethernet *mask) | |
ca78c6b6 | 327 | { |
4edb9ae9 | 328 | int err; |
b940b3d7 | 329 | |
5cce04b6 | 330 | err = skb_ensure_writable(skb, ETH_HLEN); |
4edb9ae9 PS |
331 | if (unlikely(err)) |
332 | return err; | |
333 | ||
237c4f2a | 334 | skb_postpull_rcsum(skb, eth_hdr(skb), ETH_ALEN * 2); |
3cfede14 | 335 | |
b940b3d7 JR |
336 | ether_addr_copy_masked(eth_hdr(skb)->h_source, key->eth_src, |
337 | mask->eth_src); | |
338 | ether_addr_copy_masked(eth_hdr(skb)->h_dest, key->eth_dst, | |
339 | mask->eth_dst); | |
4edb9ae9 | 340 | |
ea3acd7a | 341 | skb_postpush_rcsum(skb, eth_hdr(skb), ETH_ALEN * 2); |
3cfede14 | 342 | |
b940b3d7 JR |
343 | ether_addr_copy(flow_key->eth.src, eth_hdr(skb)->h_source); |
344 | ether_addr_copy(flow_key->eth.dst, eth_hdr(skb)->h_dest); | |
4edb9ae9 | 345 | return 0; |
ca78c6b6 BP |
346 | } |
347 | ||
6fcecb85 YY |
348 | /* pop_eth does not support VLAN packets as this action is never called |
349 | * for them. | |
350 | */ | |
351 | static int pop_eth(struct sk_buff *skb, struct sw_flow_key *key) | |
352 | { | |
353 | skb_pull_rcsum(skb, ETH_HLEN); | |
354 | skb_reset_mac_header(skb); | |
355 | skb_reset_mac_len(skb); | |
356 | ||
357 | /* safe right before invalidate_flow_key */ | |
358 | key->mac_proto = MAC_PROTO_NONE; | |
359 | invalidate_flow_key(key); | |
360 | return 0; | |
361 | } | |
362 | ||
363 | static int push_eth(struct sk_buff *skb, struct sw_flow_key *key, | |
364 | const struct ovs_action_push_eth *ethh) | |
365 | { | |
366 | struct ethhdr *hdr; | |
367 | ||
368 | /* Add the new Ethernet header */ | |
369 | if (skb_cow_head(skb, ETH_HLEN) < 0) | |
370 | return -ENOMEM; | |
371 | ||
372 | skb_push(skb, ETH_HLEN); | |
373 | skb_reset_mac_header(skb); | |
374 | skb_reset_mac_len(skb); | |
375 | ||
376 | hdr = eth_hdr(skb); | |
377 | ether_addr_copy(hdr->h_source, ethh->addresses.eth_src); | |
378 | ether_addr_copy(hdr->h_dest, ethh->addresses.eth_dst); | |
379 | hdr->h_proto = skb->protocol; | |
380 | ||
381 | skb_postpush_rcsum(skb, hdr, ETH_HLEN); | |
382 | ||
383 | /* safe right before invalidate_flow_key */ | |
384 | key->mac_proto = MAC_PROTO_ETHERNET; | |
385 | invalidate_flow_key(key); | |
386 | return 0; | |
387 | } | |
388 | ||
96b82f6d YY |
389 | static int push_nsh(struct sk_buff *skb, struct sw_flow_key *key, |
390 | const struct nshhdr *nh) | |
391 | { | |
392 | int err; | |
393 | ||
394 | err = ovs_nsh_push(skb, nh); | |
395 | if (err) | |
396 | return err; | |
397 | ||
398 | /* safe right before invalidate_flow_key */ | |
399 | key->mac_proto = MAC_PROTO_NONE; | |
400 | invalidate_flow_key(key); | |
401 | return 0; | |
402 | } | |
403 | ||
404 | static int pop_nsh(struct sk_buff *skb, struct sw_flow_key *key) | |
405 | { | |
406 | int err; | |
407 | ||
408 | err = ovs_nsh_pop(skb); | |
409 | if (err) | |
410 | return err; | |
411 | ||
412 | /* safe right before invalidate_flow_key */ | |
413 | if (skb->protocol == htons(ETH_P_TEB)) | |
414 | key->mac_proto = MAC_PROTO_ETHERNET; | |
415 | else | |
416 | key->mac_proto = MAC_PROTO_NONE; | |
417 | invalidate_flow_key(key); | |
418 | return 0; | |
419 | } | |
420 | ||
efdb0c9f GG |
421 | static void update_ip_l4_checksum(struct sk_buff *skb, struct iphdr *nh, |
422 | __be32 addr, __be32 new_addr) | |
ca78c6b6 BP |
423 | { |
424 | int transport_len = skb->len - skb_transport_offset(skb); | |
4edb9ae9 | 425 | |
efdb0c9f GG |
426 | if (nh->frag_off & htons(IP_OFFSET)) |
427 | return; | |
428 | ||
4edb9ae9 | 429 | if (nh->protocol == IPPROTO_TCP) { |
ca78c6b6 | 430 | if (likely(transport_len >= sizeof(struct tcphdr))) |
4edb9ae9 | 431 | inet_proto_csum_replace4(&tcp_hdr(skb)->check, skb, |
8063e095 | 432 | addr, new_addr, true); |
4edb9ae9 | 433 | } else if (nh->protocol == IPPROTO_UDP) { |
55ce87bc JG |
434 | if (likely(transport_len >= sizeof(struct udphdr))) { |
435 | struct udphdr *uh = udp_hdr(skb); | |
436 | ||
237c4f2a | 437 | if (uh->check || skb->ip_summed == CHECKSUM_PARTIAL) { |
55ce87bc | 438 | inet_proto_csum_replace4(&uh->check, skb, |
8063e095 | 439 | addr, new_addr, true); |
55ce87bc JG |
440 | if (!uh->check) |
441 | uh->check = CSUM_MANGLED_0; | |
442 | } | |
443 | } | |
ca78c6b6 | 444 | } |
4edb9ae9 | 445 | |
efdb0c9f GG |
446 | } |
447 | ||
448 | static void set_ip_addr(struct sk_buff *skb, struct iphdr *nh, | |
449 | __be32 *addr, __be32 new_addr) | |
450 | { | |
451 | update_ip_l4_checksum(skb, nh, *addr, new_addr); | |
4edb9ae9 | 452 | csum_replace4(&nh->check, *addr, new_addr); |
e2f3178f | 453 | skb_clear_hash(skb); |
4edb9ae9 | 454 | *addr = new_addr; |
ca78c6b6 BP |
455 | } |
456 | ||
bc7a5acd AA |
457 | static void update_ipv6_checksum(struct sk_buff *skb, u8 l4_proto, |
458 | __be32 addr[4], const __be32 new_addr[4]) | |
459 | { | |
460 | int transport_len = skb->len - skb_transport_offset(skb); | |
461 | ||
00894212 | 462 | if (l4_proto == NEXTHDR_TCP) { |
bc7a5acd AA |
463 | if (likely(transport_len >= sizeof(struct tcphdr))) |
464 | inet_proto_csum_replace16(&tcp_hdr(skb)->check, skb, | |
8063e095 | 465 | addr, new_addr, true); |
00894212 | 466 | } else if (l4_proto == NEXTHDR_UDP) { |
bc7a5acd AA |
467 | if (likely(transport_len >= sizeof(struct udphdr))) { |
468 | struct udphdr *uh = udp_hdr(skb); | |
469 | ||
237c4f2a | 470 | if (uh->check || skb->ip_summed == CHECKSUM_PARTIAL) { |
bc7a5acd | 471 | inet_proto_csum_replace16(&uh->check, skb, |
8063e095 | 472 | addr, new_addr, true); |
bc7a5acd AA |
473 | if (!uh->check) |
474 | uh->check = CSUM_MANGLED_0; | |
475 | } | |
476 | } | |
00894212 JG |
477 | } else if (l4_proto == NEXTHDR_ICMP) { |
478 | if (likely(transport_len >= sizeof(struct icmp6hdr))) | |
479 | inet_proto_csum_replace16(&icmp6_hdr(skb)->icmp6_cksum, | |
8063e095 | 480 | skb, addr, new_addr, true); |
bc7a5acd AA |
481 | } |
482 | } | |
483 | ||
b940b3d7 JR |
484 | static void mask_ipv6_addr(const __be32 old[4], const __be32 addr[4], |
485 | const __be32 mask[4], __be32 masked[4]) | |
486 | { | |
e281bb23 JS |
487 | masked[0] = OVS_MASKED(old[0], addr[0], mask[0]); |
488 | masked[1] = OVS_MASKED(old[1], addr[1], mask[1]); | |
489 | masked[2] = OVS_MASKED(old[2], addr[2], mask[2]); | |
490 | masked[3] = OVS_MASKED(old[3], addr[3], mask[3]); | |
b940b3d7 JR |
491 | } |
492 | ||
bc7a5acd AA |
493 | static void set_ipv6_addr(struct sk_buff *skb, u8 l4_proto, |
494 | __be32 addr[4], const __be32 new_addr[4], | |
495 | bool recalculate_csum) | |
496 | { | |
51cf5e71 | 497 | if (likely(recalculate_csum)) |
bc7a5acd AA |
498 | update_ipv6_checksum(skb, l4_proto, addr, new_addr); |
499 | ||
e2f3178f | 500 | skb_clear_hash(skb); |
bc7a5acd AA |
501 | memcpy(addr, new_addr, sizeof(__be32[4])); |
502 | } | |
503 | ||
b940b3d7 | 504 | static void set_ipv6_fl(struct ipv6hdr *nh, u32 fl, u32 mask) |
bc7a5acd | 505 | { |
b940b3d7 | 506 | /* Bits 21-24 are always unmasked, so this retains their values. */ |
e281bb23 JS |
507 | OVS_SET_MASKED(nh->flow_lbl[0], (u8)(fl >> 16), (u8)(mask >> 16)); |
508 | OVS_SET_MASKED(nh->flow_lbl[1], (u8)(fl >> 8), (u8)(mask >> 8)); | |
509 | OVS_SET_MASKED(nh->flow_lbl[2], (u8)fl, (u8)mask); | |
bc7a5acd AA |
510 | } |
511 | ||
b940b3d7 JR |
512 | static void set_ip_ttl(struct sk_buff *skb, struct iphdr *nh, u8 new_ttl, |
513 | u8 mask) | |
bc7a5acd | 514 | { |
e281bb23 | 515 | new_ttl = OVS_MASKED(nh->ttl, new_ttl, mask); |
bc7a5acd | 516 | |
a61680c6 JP |
517 | csum_replace2(&nh->check, htons(nh->ttl << 8), htons(new_ttl << 8)); |
518 | nh->ttl = new_ttl; | |
519 | } | |
520 | ||
b940b3d7 JR |
521 | static int set_ipv4(struct sk_buff *skb, struct sw_flow_key *flow_key, |
522 | const struct ovs_key_ipv4 *key, | |
523 | const struct ovs_key_ipv4 *mask) | |
064af421 | 524 | { |
ca78c6b6 | 525 | struct iphdr *nh; |
b940b3d7 | 526 | __be32 new_addr; |
10db8b20 | 527 | int err; |
ca78c6b6 | 528 | |
5cce04b6 TG |
529 | err = skb_ensure_writable(skb, skb_network_offset(skb) + |
530 | sizeof(struct iphdr)); | |
10db8b20 JG |
531 | if (unlikely(err)) |
532 | return err; | |
ca78c6b6 BP |
533 | |
534 | nh = ip_hdr(skb); | |
ca78c6b6 | 535 | |
b940b3d7 JR |
536 | /* Setting an IP addresses is typically only a side effect of |
537 | * matching on them in the current userspace implementation, so it | |
538 | * makes sense to check if the value actually changed. | |
539 | */ | |
540 | if (mask->ipv4_src) { | |
e281bb23 | 541 | new_addr = OVS_MASKED(nh->saddr, key->ipv4_src, mask->ipv4_src); |
ca78c6b6 | 542 | |
b940b3d7 JR |
543 | if (unlikely(new_addr != nh->saddr)) { |
544 | set_ip_addr(skb, nh, &nh->saddr, new_addr); | |
545 | flow_key->ipv4.addr.src = new_addr; | |
546 | } | |
e16138e2 | 547 | } |
b940b3d7 | 548 | if (mask->ipv4_dst) { |
e281bb23 | 549 | new_addr = OVS_MASKED(nh->daddr, key->ipv4_dst, mask->ipv4_dst); |
a4a26436 | 550 | |
b940b3d7 JR |
551 | if (unlikely(new_addr != nh->daddr)) { |
552 | set_ip_addr(skb, nh, &nh->daddr, new_addr); | |
553 | flow_key->ipv4.addr.dst = new_addr; | |
554 | } | |
e16138e2 | 555 | } |
b940b3d7 JR |
556 | if (mask->ipv4_tos) { |
557 | ipv4_change_dsfield(nh, ~mask->ipv4_tos, key->ipv4_tos); | |
558 | flow_key->ip.tos = nh->tos; | |
559 | } | |
560 | if (mask->ipv4_ttl) { | |
561 | set_ip_ttl(skb, nh, key->ipv4_ttl, mask->ipv4_ttl); | |
562 | flow_key->ip.ttl = nh->ttl; | |
e16138e2 | 563 | } |
a61680c6 | 564 | |
10db8b20 | 565 | return 0; |
064af421 BP |
566 | } |
567 | ||
b940b3d7 JR |
568 | static bool is_ipv6_mask_nonzero(const __be32 addr[4]) |
569 | { | |
570 | return !!(addr[0] | addr[1] | addr[2] | addr[3]); | |
571 | } | |
572 | ||
573 | static int set_ipv6(struct sk_buff *skb, struct sw_flow_key *flow_key, | |
574 | const struct ovs_key_ipv6 *key, | |
575 | const struct ovs_key_ipv6 *mask) | |
bc7a5acd AA |
576 | { |
577 | struct ipv6hdr *nh; | |
578 | int err; | |
bc7a5acd | 579 | |
5cce04b6 TG |
580 | err = skb_ensure_writable(skb, skb_network_offset(skb) + |
581 | sizeof(struct ipv6hdr)); | |
bc7a5acd AA |
582 | if (unlikely(err)) |
583 | return err; | |
584 | ||
585 | nh = ipv6_hdr(skb); | |
bc7a5acd | 586 | |
b940b3d7 JR |
587 | /* Setting an IP addresses is typically only a side effect of |
588 | * matching on them in the current userspace implementation, so it | |
589 | * makes sense to check if the value actually changed. | |
590 | */ | |
591 | if (is_ipv6_mask_nonzero(mask->ipv6_src)) { | |
592 | __be32 *saddr = (__be32 *)&nh->saddr; | |
593 | __be32 masked[4]; | |
594 | ||
595 | mask_ipv6_addr(saddr, key->ipv6_src, mask->ipv6_src, masked); | |
596 | ||
597 | if (unlikely(memcmp(saddr, masked, sizeof(masked)))) { | |
49a8eef8 | 598 | set_ipv6_addr(skb, flow_key->ip.proto, saddr, masked, |
b940b3d7 JR |
599 | true); |
600 | memcpy(&flow_key->ipv6.addr.src, masked, | |
601 | sizeof(flow_key->ipv6.addr.src)); | |
602 | } | |
603 | } | |
604 | if (is_ipv6_mask_nonzero(mask->ipv6_dst)) { | |
bc7a5acd | 605 | unsigned int offset = 0; |
8abaa53c | 606 | int flags = IP6_FH_F_SKIP_RH; |
bc7a5acd | 607 | bool recalc_csum = true; |
b940b3d7 JR |
608 | __be32 *daddr = (__be32 *)&nh->daddr; |
609 | __be32 masked[4]; | |
610 | ||
611 | mask_ipv6_addr(daddr, key->ipv6_dst, mask->ipv6_dst, masked); | |
612 | ||
613 | if (unlikely(memcmp(daddr, masked, sizeof(masked)))) { | |
614 | if (ipv6_ext_hdr(nh->nexthdr)) | |
615 | recalc_csum = (ipv6_find_hdr(skb, &offset, | |
616 | NEXTHDR_ROUTING, | |
617 | NULL, &flags) | |
618 | != NEXTHDR_ROUTING); | |
619 | ||
49a8eef8 | 620 | set_ipv6_addr(skb, flow_key->ip.proto, daddr, masked, |
b940b3d7 JR |
621 | recalc_csum); |
622 | memcpy(&flow_key->ipv6.addr.dst, masked, | |
623 | sizeof(flow_key->ipv6.addr.dst)); | |
624 | } | |
625 | } | |
626 | if (mask->ipv6_tclass) { | |
627 | ipv6_change_dsfield(nh, ~mask->ipv6_tclass, key->ipv6_tclass); | |
628 | flow_key->ip.tos = ipv6_get_dsfield(nh); | |
629 | } | |
630 | if (mask->ipv6_label) { | |
631 | set_ipv6_fl(nh, ntohl(key->ipv6_label), | |
632 | ntohl(mask->ipv6_label)); | |
633 | flow_key->ipv6.label = | |
634 | *(__be32 *)nh & htonl(IPV6_FLOWINFO_FLOWLABEL); | |
635 | } | |
636 | if (mask->ipv6_hlimit) { | |
e281bb23 JS |
637 | OVS_SET_MASKED(nh->hop_limit, key->ipv6_hlimit, |
638 | mask->ipv6_hlimit); | |
b940b3d7 | 639 | flow_key->ip.ttl = nh->hop_limit; |
bc7a5acd | 640 | } |
bc7a5acd AA |
641 | return 0; |
642 | } | |
643 | ||
96b82f6d YY |
644 | static int set_nsh(struct sk_buff *skb, struct sw_flow_key *flow_key, |
645 | const struct nlattr *a) | |
646 | { | |
647 | struct nshhdr *nh; | |
648 | size_t length; | |
649 | int err; | |
650 | u8 flags; | |
651 | u8 ttl; | |
652 | int i; | |
653 | ||
654 | struct ovs_key_nsh key; | |
655 | struct ovs_key_nsh mask; | |
656 | ||
657 | err = nsh_key_from_nlattr(a, &key, &mask); | |
658 | if (err) | |
659 | return err; | |
660 | ||
661 | /* Make sure the NSH base header is there */ | |
662 | if (!pskb_may_pull(skb, skb_network_offset(skb) + NSH_BASE_HDR_LEN)) | |
663 | return -ENOMEM; | |
664 | ||
665 | nh = nsh_hdr(skb); | |
666 | length = nsh_hdr_len(nh); | |
667 | ||
668 | /* Make sure the whole NSH header is there */ | |
669 | err = skb_ensure_writable(skb, skb_network_offset(skb) + | |
670 | length); | |
671 | if (unlikely(err)) | |
672 | return err; | |
673 | ||
674 | nh = nsh_hdr(skb); | |
675 | skb_postpull_rcsum(skb, nh, length); | |
676 | flags = nsh_get_flags(nh); | |
677 | flags = OVS_MASKED(flags, key.base.flags, mask.base.flags); | |
678 | flow_key->nsh.base.flags = flags; | |
679 | ttl = nsh_get_ttl(nh); | |
680 | ttl = OVS_MASKED(ttl, key.base.ttl, mask.base.ttl); | |
681 | flow_key->nsh.base.ttl = ttl; | |
682 | nsh_set_flags_and_ttl(nh, flags, ttl); | |
683 | nh->path_hdr = OVS_MASKED(nh->path_hdr, key.base.path_hdr, | |
684 | mask.base.path_hdr); | |
685 | flow_key->nsh.base.path_hdr = nh->path_hdr; | |
686 | switch (nh->mdtype) { | |
687 | case NSH_M_TYPE1: | |
688 | for (i = 0; i < NSH_MD1_CONTEXT_SIZE; i++) { | |
689 | nh->md1.context[i] = | |
690 | OVS_MASKED(nh->md1.context[i], key.context[i], | |
691 | mask.context[i]); | |
692 | } | |
693 | memcpy(flow_key->nsh.context, nh->md1.context, | |
694 | sizeof(nh->md1.context)); | |
695 | break; | |
696 | case NSH_M_TYPE2: | |
697 | memset(flow_key->nsh.context, 0, | |
698 | sizeof(flow_key->nsh.context)); | |
699 | break; | |
700 | default: | |
701 | return -EINVAL; | |
702 | } | |
703 | skb_postpush_rcsum(skb, nh, length); | |
704 | return 0; | |
705 | } | |
706 | ||
5cce04b6 | 707 | /* Must follow skb_ensure_writable() since that can move the skb data. */ |
4edb9ae9 | 708 | static void set_tp_port(struct sk_buff *skb, __be16 *port, |
b940b3d7 | 709 | __be16 new_port, __sum16 *check) |
959a2ecd | 710 | { |
8063e095 | 711 | inet_proto_csum_replace2(check, skb, *port, new_port, false); |
4edb9ae9 | 712 | *port = new_port; |
55ce87bc JG |
713 | } |
714 | ||
b940b3d7 JR |
715 | static int set_udp(struct sk_buff *skb, struct sw_flow_key *flow_key, |
716 | const struct ovs_key_udp *key, | |
717 | const struct ovs_key_udp *mask) | |
4edb9ae9 PS |
718 | { |
719 | struct udphdr *uh; | |
b940b3d7 | 720 | __be16 src, dst; |
4edb9ae9 | 721 | int err; |
10db8b20 | 722 | |
5cce04b6 TG |
723 | err = skb_ensure_writable(skb, skb_transport_offset(skb) + |
724 | sizeof(struct udphdr)); | |
10db8b20 JG |
725 | if (unlikely(err)) |
726 | return err; | |
727 | ||
4edb9ae9 | 728 | uh = udp_hdr(skb); |
b940b3d7 | 729 | /* Either of the masks is non-zero, so do not bother checking them. */ |
e281bb23 JS |
730 | src = OVS_MASKED(uh->source, key->udp_src, mask->udp_src); |
731 | dst = OVS_MASKED(uh->dest, key->udp_dst, mask->udp_dst); | |
4edb9ae9 | 732 | |
b940b3d7 JR |
733 | if (uh->check && skb->ip_summed != CHECKSUM_PARTIAL) { |
734 | if (likely(src != uh->source)) { | |
735 | set_tp_port(skb, &uh->source, src, &uh->check); | |
736 | flow_key->tp.src = src; | |
737 | } | |
738 | if (likely(dst != uh->dest)) { | |
739 | set_tp_port(skb, &uh->dest, dst, &uh->check); | |
740 | flow_key->tp.dst = dst; | |
741 | } | |
742 | ||
743 | if (unlikely(!uh->check)) | |
744 | uh->check = CSUM_MANGLED_0; | |
745 | } else { | |
746 | uh->source = src; | |
747 | uh->dest = dst; | |
748 | flow_key->tp.src = src; | |
749 | flow_key->tp.dst = dst; | |
e16138e2 | 750 | } |
10db8b20 | 751 | |
b940b3d7 JR |
752 | skb_clear_hash(skb); |
753 | ||
10db8b20 | 754 | return 0; |
959a2ecd JP |
755 | } |
756 | ||
b940b3d7 JR |
757 | static int set_tcp(struct sk_buff *skb, struct sw_flow_key *flow_key, |
758 | const struct ovs_key_tcp *key, | |
759 | const struct ovs_key_tcp *mask) | |
064af421 | 760 | { |
4edb9ae9 | 761 | struct tcphdr *th; |
b940b3d7 | 762 | __be16 src, dst; |
10db8b20 | 763 | int err; |
064af421 | 764 | |
5cce04b6 TG |
765 | err = skb_ensure_writable(skb, skb_transport_offset(skb) + |
766 | sizeof(struct tcphdr)); | |
10db8b20 JG |
767 | if (unlikely(err)) |
768 | return err; | |
ca78c6b6 | 769 | |
4edb9ae9 | 770 | th = tcp_hdr(skb); |
e281bb23 | 771 | src = OVS_MASKED(th->source, key->tcp_src, mask->tcp_src); |
b940b3d7 JR |
772 | if (likely(src != th->source)) { |
773 | set_tp_port(skb, &th->source, src, &th->check); | |
774 | flow_key->tp.src = src; | |
775 | } | |
e281bb23 | 776 | dst = OVS_MASKED(th->dest, key->tcp_dst, mask->tcp_dst); |
b940b3d7 JR |
777 | if (likely(dst != th->dest)) { |
778 | set_tp_port(skb, &th->dest, dst, &th->check); | |
779 | flow_key->tp.dst = dst; | |
e16138e2 | 780 | } |
b940b3d7 | 781 | skb_clear_hash(skb); |
ca78c6b6 | 782 | |
10db8b20 | 783 | return 0; |
064af421 BP |
784 | } |
785 | ||
b940b3d7 JR |
786 | static int set_sctp(struct sk_buff *skb, struct sw_flow_key *flow_key, |
787 | const struct ovs_key_sctp *key, | |
788 | const struct ovs_key_sctp *mask) | |
10f72e3d | 789 | { |
b940b3d7 | 790 | unsigned int sctphoff = skb_transport_offset(skb); |
10f72e3d | 791 | struct sctphdr *sh; |
b940b3d7 | 792 | __le32 old_correct_csum, new_csum, old_csum; |
10f72e3d | 793 | int err; |
10f72e3d | 794 | |
5cce04b6 | 795 | err = skb_ensure_writable(skb, sctphoff + sizeof(struct sctphdr)); |
10f72e3d JS |
796 | if (unlikely(err)) |
797 | return err; | |
798 | ||
799 | sh = sctp_hdr(skb); | |
b940b3d7 JR |
800 | old_csum = sh->checksum; |
801 | old_correct_csum = sctp_compute_cksum(skb, sctphoff); | |
10f72e3d | 802 | |
e281bb23 JS |
803 | sh->source = OVS_MASKED(sh->source, key->sctp_src, mask->sctp_src); |
804 | sh->dest = OVS_MASKED(sh->dest, key->sctp_dst, mask->sctp_dst); | |
10f72e3d | 805 | |
b940b3d7 | 806 | new_csum = sctp_compute_cksum(skb, sctphoff); |
10f72e3d | 807 | |
b940b3d7 JR |
808 | /* Carry any checksum errors through. */ |
809 | sh->checksum = old_csum ^ old_correct_csum ^ new_csum; | |
10f72e3d | 810 | |
b940b3d7 JR |
811 | skb_clear_hash(skb); |
812 | flow_key->tp.src = sh->source; | |
813 | flow_key->tp.dst = sh->dest; | |
10f72e3d JS |
814 | |
815 | return 0; | |
816 | } | |
817 | ||
a94ebc39 JS |
818 | static int ovs_vport_output(OVS_VPORT_OUTPUT_PARAMS) |
819 | { | |
e9326797 | 820 | struct ovs_frag_data *data = this_cpu_ptr(&ovs_frag_data_storage); |
a94ebc39 JS |
821 | struct vport *vport = data->vport; |
822 | ||
823 | if (skb_cow_head(skb, data->l2_len) < 0) { | |
824 | kfree_skb(skb); | |
825 | return -ENOMEM; | |
826 | } | |
827 | ||
828 | __skb_dst_copy(skb, data->dst); | |
86c2eb45 | 829 | *OVS_GSO_CB(skb) = data->cb; |
a94ebc39 JS |
830 | ovs_skb_set_inner_protocol(skb, data->inner_protocol); |
831 | skb->vlan_tci = data->vlan_tci; | |
832 | skb->vlan_proto = data->vlan_proto; | |
833 | ||
834 | /* Reconstruct the MAC header. */ | |
835 | skb_push(skb, data->l2_len); | |
836 | memcpy(skb->data, &data->l2_data, data->l2_len); | |
ea3acd7a | 837 | skb_postpush_rcsum(skb, skb->data, data->l2_len); |
a94ebc39 JS |
838 | skb_reset_mac_header(skb); |
839 | ||
279ecabe YHW |
840 | if (eth_p_mpls(skb->protocol)) { |
841 | skb->inner_network_header = skb->network_header; | |
842 | skb_set_network_header(skb, data->network_offset); | |
843 | skb_reset_mac_len(skb); | |
844 | } | |
845 | ||
f80d0524 | 846 | ovs_vport_send(vport, skb, data->mac_proto); |
a94ebc39 JS |
847 | return 0; |
848 | } | |
849 | ||
850 | static unsigned int | |
851 | ovs_dst_get_mtu(const struct dst_entry *dst) | |
852 | { | |
853 | return dst->dev->mtu; | |
854 | } | |
855 | ||
856 | static struct dst_ops ovs_dst_ops = { | |
857 | .family = AF_UNSPEC, | |
858 | .mtu = ovs_dst_get_mtu, | |
859 | }; | |
860 | ||
861 | /* prepare_frag() is called once per (larger-than-MTU) frame; its inverse is | |
862 | * ovs_vport_output(), which is called once per fragmented packet. | |
863 | */ | |
f80d0524 | 864 | static void prepare_frag(struct vport *vport, struct sk_buff *skb, |
279ecabe | 865 | u16 orig_network_offset, u8 mac_proto) |
a94ebc39 JS |
866 | { |
867 | unsigned int hlen = skb_network_offset(skb); | |
868 | struct ovs_frag_data *data; | |
869 | ||
e9326797 | 870 | data = this_cpu_ptr(&ovs_frag_data_storage); |
a94ebc39 JS |
871 | data->dst = (unsigned long) skb_dst(skb); |
872 | data->vport = vport; | |
86c2eb45 | 873 | data->cb = *OVS_GSO_CB(skb); |
a94ebc39 | 874 | data->inner_protocol = ovs_skb_get_inner_protocol(skb); |
279ecabe | 875 | data->network_offset = orig_network_offset; |
a94ebc39 JS |
876 | data->vlan_tci = skb->vlan_tci; |
877 | data->vlan_proto = skb->vlan_proto; | |
f80d0524 | 878 | data->mac_proto = mac_proto; |
a94ebc39 JS |
879 | data->l2_len = hlen; |
880 | memcpy(&data->l2_data, skb->data, hlen); | |
881 | ||
882 | memset(IPCB(skb), 0, sizeof(struct inet_skb_parm)); | |
883 | skb_pull(skb, hlen); | |
884 | } | |
885 | ||
025813bb | 886 | static void ovs_fragment(struct net *net, struct vport *vport, |
f80d0524 YY |
887 | struct sk_buff *skb, u16 mru, |
888 | struct sw_flow_key *key) | |
a94ebc39 | 889 | { |
279ecabe YHW |
890 | u16 orig_network_offset = 0; |
891 | ||
892 | if (eth_p_mpls(skb->protocol)) { | |
893 | orig_network_offset = skb_network_offset(skb); | |
894 | skb->network_header = skb->inner_network_header; | |
895 | } | |
896 | ||
a94ebc39 JS |
897 | if (skb_network_offset(skb) > MAX_L2_LEN) { |
898 | OVS_NLERR(1, "L2 header too long to fragment"); | |
c05e2094 | 899 | goto err; |
a94ebc39 JS |
900 | } |
901 | ||
f80d0524 | 902 | if (key->eth.type == htons(ETH_P_IP)) { |
a94ebc39 JS |
903 | struct dst_entry ovs_dst; |
904 | unsigned long orig_dst; | |
905 | ||
279ecabe YHW |
906 | prepare_frag(vport, skb, orig_network_offset, |
907 | ovs_key_mac_proto(key)); | |
a94ebc39 JS |
908 | dst_init(&ovs_dst, &ovs_dst_ops, NULL, 1, |
909 | DST_OBSOLETE_NONE, DST_NOCOUNT); | |
910 | ovs_dst.dev = vport->dev; | |
911 | ||
912 | orig_dst = (unsigned long) skb_dst(skb); | |
913 | skb_dst_set_noref(skb, &ovs_dst); | |
914 | IPCB(skb)->frag_max_size = mru; | |
915 | ||
0374bcbe | 916 | ip_do_fragment(net, skb->sk, skb, ovs_vport_output); |
a94ebc39 | 917 | refdst_drop(orig_dst); |
f80d0524 | 918 | } else if (key->eth.type == htons(ETH_P_IPV6)) { |
a94ebc39 JS |
919 | const struct nf_ipv6_ops *v6ops = nf_get_ipv6_ops(); |
920 | unsigned long orig_dst; | |
921 | struct rt6_info ovs_rt; | |
922 | ||
09b26ccb | 923 | if (!v6ops) |
c05e2094 | 924 | goto err; |
a94ebc39 | 925 | |
279ecabe | 926 | prepare_frag(vport, skb, orig_network_offset, |
f80d0524 | 927 | ovs_key_mac_proto(key)); |
a94ebc39 JS |
928 | memset(&ovs_rt, 0, sizeof(ovs_rt)); |
929 | dst_init(&ovs_rt.dst, &ovs_dst_ops, NULL, 1, | |
930 | DST_OBSOLETE_NONE, DST_NOCOUNT); | |
931 | ovs_rt.dst.dev = vport->dev; | |
932 | ||
933 | orig_dst = (unsigned long) skb_dst(skb); | |
934 | skb_dst_set_noref(skb, &ovs_rt.dst); | |
935 | IP6CB(skb)->frag_max_size = mru; | |
0643a78b PS |
936 | #ifdef HAVE_IP_LOCAL_OUT_TAKES_NET |
937 | v6ops->fragment(net, skb->sk, skb, ovs_vport_output); | |
938 | #else | |
a94ebc39 | 939 | v6ops->fragment(skb->sk, skb, ovs_vport_output); |
0643a78b | 940 | #endif |
a94ebc39 JS |
941 | refdst_drop(orig_dst); |
942 | } else { | |
943 | WARN_ONCE(1, "Failed fragment ->%s: eth=%04x, MRU=%d, MTU=%d.", | |
f80d0524 | 944 | ovs_vport_name(vport), ntohs(key->eth.type), mru, |
a94ebc39 | 945 | vport->dev->mtu); |
c05e2094 | 946 | goto err; |
a94ebc39 | 947 | } |
c05e2094 JS |
948 | |
949 | return; | |
950 | err: | |
951 | kfree_skb(skb); | |
a94ebc39 | 952 | } |
a94ebc39 JS |
953 | |
954 | static void do_output(struct datapath *dp, struct sk_buff *skb, int out_port, | |
955 | struct sw_flow_key *key) | |
064af421 | 956 | { |
fe90efd9 | 957 | struct vport *vport = ovs_vport_rcu(dp, out_port); |
064af421 | 958 | |
a94ebc39 JS |
959 | if (likely(vport)) { |
960 | u16 mru = OVS_CB(skb)->mru; | |
4c7804f1 WT |
961 | u32 cutlen = OVS_CB(skb)->cutlen; |
962 | ||
963 | if (unlikely(cutlen > 0)) { | |
f80d0524 | 964 | if (skb->len - cutlen > ovs_mac_header_len(key)) |
4c7804f1 WT |
965 | pskb_trim(skb, skb->len - cutlen); |
966 | else | |
f80d0524 | 967 | pskb_trim(skb, ovs_mac_header_len(key)); |
4c7804f1 | 968 | } |
a94ebc39 | 969 | |
f0b0bdd2 YY |
970 | if (likely(!mru || |
971 | (skb->len <= mru + vport->dev->hard_header_len))) { | |
f80d0524 | 972 | ovs_vport_send(vport, skb, ovs_key_mac_proto(key)); |
a94ebc39 | 973 | } else if (mru <= vport->dev->mtu) { |
025813bb | 974 | struct net *net = ovs_dp_get_net(dp); |
a94ebc39 | 975 | |
f80d0524 | 976 | ovs_fragment(net, vport, skb, mru, key); |
a94ebc39 JS |
977 | } else { |
978 | OVS_NLERR(true, "Cannot fragment IP frames"); | |
979 | kfree_skb(skb); | |
980 | } | |
981 | } else { | |
f15c8639 | 982 | kfree_skb(skb); |
a94ebc39 | 983 | } |
064af421 | 984 | } |
aad7cb91 | 985 | |
98403001 | 986 | static int output_userspace(struct datapath *dp, struct sk_buff *skb, |
0e469d3b | 987 | struct sw_flow_key *key, const struct nlattr *attr, |
4c7804f1 WT |
988 | const struct nlattr *actions, int actions_len, |
989 | uint32_t cutlen) | |
064af421 | 990 | { |
856081f6 | 991 | struct dp_upcall_info upcall; |
98403001 | 992 | const struct nlattr *a; |
aad7cb91 | 993 | int rem, err; |
856081f6 | 994 | |
0e469d3b | 995 | memset(&upcall, 0, sizeof(upcall)); |
df2c07f4 | 996 | upcall.cmd = OVS_PACKET_CMD_ACTION; |
a94ebc39 | 997 | upcall.mru = OVS_CB(skb)->mru; |
98403001 | 998 | |
aad7cb91 | 999 | SKB_INIT_FILL_METADATA_DST(skb); |
98403001 BP |
1000 | for (a = nla_data(attr), rem = nla_len(attr); rem > 0; |
1001 | a = nla_next(a, &rem)) { | |
1002 | switch (nla_type(a)) { | |
1003 | case OVS_USERSPACE_ATTR_USERDATA: | |
1004 | upcall.userdata = a; | |
1005 | break; | |
1006 | ||
1007 | case OVS_USERSPACE_ATTR_PID: | |
28aea917 | 1008 | upcall.portid = nla_get_u32(a); |
98403001 | 1009 | break; |
8b7ea2d4 WZ |
1010 | |
1011 | case OVS_USERSPACE_ATTR_EGRESS_TUN_PORT: { | |
1012 | /* Get out tunnel info. */ | |
1013 | struct vport *vport; | |
1014 | ||
1015 | vport = ovs_vport_rcu(dp, nla_get_u32(a)); | |
1016 | if (vport) { | |
aad7cb91 PS |
1017 | err = dev_fill_metadata_dst(vport->dev, skb); |
1018 | if (!err) | |
1019 | upcall.egress_tun_info = skb_tunnel_info(skb); | |
8b7ea2d4 | 1020 | } |
e23775f2 | 1021 | |
8b7ea2d4 | 1022 | break; |
98403001 | 1023 | } |
8b7ea2d4 | 1024 | |
0e469d3b NM |
1025 | case OVS_USERSPACE_ATTR_ACTIONS: { |
1026 | /* Include actions. */ | |
1027 | upcall.actions = actions; | |
1028 | upcall.actions_len = actions_len; | |
1029 | break; | |
1030 | } | |
1031 | ||
8b7ea2d4 | 1032 | } /* End of switch. */ |
98403001 BP |
1033 | } |
1034 | ||
aad7cb91 PS |
1035 | err = ovs_dp_upcall(dp, skb, key, &upcall, cutlen); |
1036 | SKB_RESTORE_FILL_METADATA_DST(skb); | |
1037 | return err; | |
064af421 BP |
1038 | } |
1039 | ||
7d9595e3 AZ |
1040 | /* When 'last' is true, sample() should always consume the 'skb'. |
1041 | * Otherwise, sample() should keep 'skb' intact regardless what | |
1042 | * actions are executed within sample(). | |
1043 | */ | |
6ff686f2 | 1044 | static int sample(struct datapath *dp, struct sk_buff *skb, |
0e469d3b | 1045 | struct sw_flow_key *key, const struct nlattr *attr, |
7d9595e3 | 1046 | bool last) |
6ff686f2 | 1047 | { |
7d9595e3 AZ |
1048 | struct nlattr *actions; |
1049 | struct nlattr *sample_arg; | |
7d9595e3 | 1050 | int rem = nla_len(attr); |
7d9595e3 | 1051 | const struct sample_arg *arg; |
d4004f8d | 1052 | bool clone_flow_key; |
6ff686f2 | 1053 | |
7d9595e3 AZ |
1054 | /* The first action is always 'OVS_SAMPLE_ATTR_ARG'. */ |
1055 | sample_arg = nla_data(attr); | |
1056 | arg = nla_data(sample_arg); | |
1057 | actions = nla_next(sample_arg, &rem); | |
c02c4967 | 1058 | |
7d9595e3 AZ |
1059 | if ((arg->probability != U32_MAX) && |
1060 | (!arg->probability || prandom_u32() > arg->probability)) { | |
1061 | if (last) | |
1062 | consume_skb(skb); | |
1063 | return 0; | |
6ff686f2 PS |
1064 | } |
1065 | ||
d4004f8d AZ |
1066 | clone_flow_key = !arg->exec; |
1067 | return clone_execute(dp, skb, key, 0, actions, rem, last, | |
1068 | clone_flow_key); | |
6ff686f2 PS |
1069 | } |
1070 | ||
5a77cf96 YS |
1071 | /* When 'last' is true, clone() should always consume the 'skb'. |
1072 | * Otherwise, clone() should keep 'skb' intact regardless what | |
1073 | * actions are executed within clone(). | |
1074 | */ | |
1075 | static int clone(struct datapath *dp, struct sk_buff *skb, | |
1076 | struct sw_flow_key *key, const struct nlattr *attr, | |
1077 | bool last) | |
1078 | { | |
1079 | struct nlattr *actions; | |
1080 | struct nlattr *clone_arg; | |
1081 | int rem = nla_len(attr); | |
1082 | bool dont_clone_flow_key; | |
1083 | ||
1084 | /* The first action is always 'OVS_CLONE_ATTR_ARG'. */ | |
1085 | clone_arg = nla_data(attr); | |
1086 | dont_clone_flow_key = nla_get_u32(clone_arg); | |
1087 | actions = nla_next(clone_arg, &rem); | |
1088 | ||
1089 | return clone_execute(dp, skb, key, 0, actions, rem, last, | |
1090 | !dont_clone_flow_key); | |
1091 | } | |
1092 | ||
e74d4817 PS |
1093 | static void execute_hash(struct sk_buff *skb, struct sw_flow_key *key, |
1094 | const struct nlattr *attr) | |
7804df20 | 1095 | { |
7804df20 AZ |
1096 | struct ovs_action_hash *hash_act = nla_data(attr); |
1097 | u32 hash = 0; | |
1098 | ||
1099 | /* OVS_HASH_ALG_L4 is the only possible hash algorithm. */ | |
e2f3178f | 1100 | hash = skb_get_hash(skb); |
7804df20 AZ |
1101 | hash = jhash_1word(hash, hash_act->hash_basis); |
1102 | if (!hash) | |
1103 | hash = 0x1; | |
1104 | ||
1105 | key->ovs_flow_hash = hash; | |
1106 | } | |
1107 | ||
b940b3d7 JR |
1108 | static int execute_set_action(struct sk_buff *skb, |
1109 | struct sw_flow_key *flow_key, | |
1110 | const struct nlattr *a) | |
1111 | { | |
1112 | /* Only tunnel set execution is supported without a mask. */ | |
1113 | if (nla_type(a) == OVS_KEY_ATTR_TUNNEL_INFO) { | |
e23775f2 PS |
1114 | struct ovs_tunnel_info *tun = nla_data(a); |
1115 | ||
1116 | ovs_skb_dst_drop(skb); | |
1117 | ovs_dst_hold((struct dst_entry *)tun->tun_dst); | |
1118 | ovs_skb_dst_set(skb, (struct dst_entry *)tun->tun_dst); | |
b940b3d7 JR |
1119 | return 0; |
1120 | } | |
1121 | ||
1122 | return -EINVAL; | |
b940b3d7 JR |
1123 | } |
1124 | ||
1125 | /* Mask is at the midpoint of the data. */ | |
1126 | #define get_mask(a, type) ((const type)nla_data(a) + 1) | |
1127 | ||
1128 | static int execute_masked_set_action(struct sk_buff *skb, | |
1129 | struct sw_flow_key *flow_key, | |
1130 | const struct nlattr *a) | |
4edb9ae9 | 1131 | { |
15c39847 | 1132 | int err = 0; |
4edb9ae9 | 1133 | |
b940b3d7 | 1134 | switch (nla_type(a)) { |
abff858b | 1135 | case OVS_KEY_ATTR_PRIORITY: |
e281bb23 JS |
1136 | OVS_SET_MASKED(skb->priority, nla_get_u32(a), |
1137 | *get_mask(a, u32 *)); | |
b940b3d7 | 1138 | flow_key->phy.priority = skb->priority; |
abff858b PS |
1139 | break; |
1140 | ||
72e8bf28 | 1141 | case OVS_KEY_ATTR_SKB_MARK: |
e281bb23 | 1142 | OVS_SET_MASKED(skb->mark, nla_get_u32(a), *get_mask(a, u32 *)); |
b940b3d7 | 1143 | flow_key->phy.skb_mark = skb->mark; |
72e8bf28 AA |
1144 | break; |
1145 | ||
f0cd669f | 1146 | case OVS_KEY_ATTR_TUNNEL_INFO: |
b940b3d7 JR |
1147 | /* Masked data not supported for tunnel. */ |
1148 | err = -EINVAL; | |
4edb9ae9 PS |
1149 | break; |
1150 | ||
1151 | case OVS_KEY_ATTR_ETHERNET: | |
b940b3d7 JR |
1152 | err = set_eth_addr(skb, flow_key, nla_data(a), |
1153 | get_mask(a, struct ovs_key_ethernet *)); | |
4edb9ae9 PS |
1154 | break; |
1155 | ||
96b82f6d YY |
1156 | case OVS_KEY_ATTR_NSH: |
1157 | err = set_nsh(skb, flow_key, a); | |
1158 | break; | |
1159 | ||
4edb9ae9 | 1160 | case OVS_KEY_ATTR_IPV4: |
b940b3d7 JR |
1161 | err = set_ipv4(skb, flow_key, nla_data(a), |
1162 | get_mask(a, struct ovs_key_ipv4 *)); | |
4edb9ae9 PS |
1163 | break; |
1164 | ||
bc7a5acd | 1165 | case OVS_KEY_ATTR_IPV6: |
b940b3d7 JR |
1166 | err = set_ipv6(skb, flow_key, nla_data(a), |
1167 | get_mask(a, struct ovs_key_ipv6 *)); | |
bc7a5acd AA |
1168 | break; |
1169 | ||
4edb9ae9 | 1170 | case OVS_KEY_ATTR_TCP: |
b940b3d7 JR |
1171 | err = set_tcp(skb, flow_key, nla_data(a), |
1172 | get_mask(a, struct ovs_key_tcp *)); | |
4edb9ae9 PS |
1173 | break; |
1174 | ||
1175 | case OVS_KEY_ATTR_UDP: | |
b940b3d7 JR |
1176 | err = set_udp(skb, flow_key, nla_data(a), |
1177 | get_mask(a, struct ovs_key_udp *)); | |
4edb9ae9 | 1178 | break; |
10f72e3d JS |
1179 | |
1180 | case OVS_KEY_ATTR_SCTP: | |
b940b3d7 JR |
1181 | err = set_sctp(skb, flow_key, nla_data(a), |
1182 | get_mask(a, struct ovs_key_sctp *)); | |
10f72e3d | 1183 | break; |
ccf43786 SH |
1184 | |
1185 | case OVS_KEY_ATTR_MPLS: | |
b940b3d7 JR |
1186 | err = set_mpls(skb, flow_key, nla_data(a), get_mask(a, |
1187 | __be32 *)); | |
ccf43786 | 1188 | break; |
a94ebc39 JS |
1189 | |
1190 | case OVS_KEY_ATTR_CT_STATE: | |
1191 | case OVS_KEY_ATTR_CT_ZONE: | |
372ce973 | 1192 | case OVS_KEY_ATTR_CT_MARK: |
c05e2094 | 1193 | case OVS_KEY_ATTR_CT_LABELS: |
c30b4cea JR |
1194 | case OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV4: |
1195 | case OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6: | |
a94ebc39 JS |
1196 | err = -EINVAL; |
1197 | break; | |
4edb9ae9 | 1198 | } |
15c39847 | 1199 | |
4edb9ae9 PS |
1200 | return err; |
1201 | } | |
1202 | ||
a6059080 | 1203 | static int execute_recirc(struct datapath *dp, struct sk_buff *skb, |
7d16c847 | 1204 | struct sw_flow_key *key, |
d4004f8d | 1205 | const struct nlattr *a, bool last) |
a6059080 | 1206 | { |
d4004f8d | 1207 | u32 recirc_id; |
e74d4817 PS |
1208 | |
1209 | if (!is_flow_key_valid(key)) { | |
867e37ba AZ |
1210 | int err; |
1211 | ||
e74d4817 | 1212 | err = ovs_flow_key_update(skb, key); |
867e37ba AZ |
1213 | if (err) |
1214 | return err; | |
867e37ba | 1215 | } |
e74d4817 | 1216 | BUG_ON(!is_flow_key_valid(key)); |
a6059080 | 1217 | |
d4004f8d AZ |
1218 | recirc_id = nla_get_u32(a); |
1219 | return clone_execute(dp, skb, key, recirc_id, NULL, 0, last, true); | |
a6059080 AZ |
1220 | } |
1221 | ||
faa3beb4 NS |
1222 | static int execute_check_pkt_len(struct datapath *dp, struct sk_buff *skb, |
1223 | struct sw_flow_key *key, | |
1224 | const struct nlattr *attr, bool last) | |
1225 | { | |
1226 | const struct nlattr *actions, *cpl_arg; | |
1227 | const struct check_pkt_len_arg *arg; | |
1228 | int rem = nla_len(attr); | |
1229 | bool clone_flow_key; | |
1230 | ||
1231 | /* The first netlink attribute in 'attr' is always | |
1232 | * 'OVS_CHECK_PKT_LEN_ATTR_ARG'. | |
1233 | */ | |
1234 | cpl_arg = nla_data(attr); | |
1235 | arg = nla_data(cpl_arg); | |
1236 | ||
1237 | if (skb->len <= arg->pkt_len) { | |
1238 | /* Second netlink attribute in 'attr' is always | |
1239 | * 'OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_LESS_EQUAL'. | |
1240 | */ | |
1241 | actions = nla_next(cpl_arg, &rem); | |
1242 | clone_flow_key = !arg->exec_for_lesser_equal; | |
1243 | } else { | |
1244 | /* Third netlink attribute in 'attr' is always | |
1245 | * 'OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_GREATER'. | |
1246 | */ | |
1247 | actions = nla_next(cpl_arg, &rem); | |
1248 | actions = nla_next(actions, &rem); | |
1249 | clone_flow_key = !arg->exec_for_greater; | |
1250 | } | |
1251 | ||
1252 | return clone_execute(dp, skb, key, 0, nla_data(actions), | |
1253 | nla_len(actions), last, clone_flow_key); | |
1254 | } | |
1255 | ||
064af421 | 1256 | /* Execute a list of actions against 'skb'. */ |
871dfe07 | 1257 | static int do_execute_actions(struct datapath *dp, struct sk_buff *skb, |
e74d4817 PS |
1258 | struct sw_flow_key *key, |
1259 | const struct nlattr *attr, int len) | |
064af421 | 1260 | { |
cdee00fd | 1261 | const struct nlattr *a; |
10db8b20 | 1262 | int rem; |
72b06300 | 1263 | |
6ff686f2 | 1264 | for (a = attr, rem = len; rem > 0; |
a4af2475 | 1265 | a = nla_next(a, &rem)) { |
10db8b20 JG |
1266 | int err = 0; |
1267 | ||
2b6a82da AZ |
1268 | switch (nla_type(a)) { |
1269 | case OVS_ACTION_ATTR_OUTPUT: { | |
1270 | int port = nla_get_u32(a); | |
1271 | struct sk_buff *clone; | |
1272 | ||
1273 | /* Every output action needs a separate clone | |
1274 | * of 'skb', In case the output action is the | |
1275 | * last action, cloning can be avoided. | |
1276 | */ | |
1277 | if (nla_is_last(a, rem)) { | |
1278 | do_output(dp, skb, port, key); | |
1279 | /* 'skb' has been used for output. | |
1280 | */ | |
1281 | return 0; | |
1282 | } | |
fe90efd9 | 1283 | |
2b6a82da AZ |
1284 | clone = skb_clone(skb, GFP_ATOMIC); |
1285 | if (clone) | |
1286 | do_output(dp, clone, port, key); | |
4c7804f1 | 1287 | OVS_CB(skb)->cutlen = 0; |
064af421 | 1288 | break; |
2b6a82da | 1289 | } |
064af421 | 1290 | |
4c7804f1 WT |
1291 | case OVS_ACTION_ATTR_TRUNC: { |
1292 | struct ovs_action_trunc *trunc = nla_data(a); | |
1293 | ||
1294 | if (skb->len > trunc->max_len) | |
1295 | OVS_CB(skb)->cutlen = skb->len - trunc->max_len; | |
1296 | break; | |
1297 | } | |
1298 | ||
df2c07f4 | 1299 | case OVS_ACTION_ATTR_USERSPACE: |
4c7804f1 WT |
1300 | output_userspace(dp, skb, key, a, attr, |
1301 | len, OVS_CB(skb)->cutlen); | |
1302 | OVS_CB(skb)->cutlen = 0; | |
064af421 | 1303 | break; |
7804df20 AZ |
1304 | |
1305 | case OVS_ACTION_ATTR_HASH: | |
e74d4817 | 1306 | execute_hash(skb, key, a); |
7804df20 | 1307 | break; |
064af421 | 1308 | |
ccf43786 | 1309 | case OVS_ACTION_ATTR_PUSH_MPLS: |
e74d4817 | 1310 | err = push_mpls(skb, key, nla_data(a)); |
ccf43786 SH |
1311 | break; |
1312 | ||
1313 | case OVS_ACTION_ATTR_POP_MPLS: | |
e74d4817 | 1314 | err = pop_mpls(skb, key, nla_get_be16(a)); |
ccf43786 SH |
1315 | break; |
1316 | ||
fea393b1 | 1317 | case OVS_ACTION_ATTR_PUSH_VLAN: |
e74d4817 | 1318 | err = push_vlan(skb, key, nla_data(a)); |
064af421 BP |
1319 | break; |
1320 | ||
fea393b1 | 1321 | case OVS_ACTION_ATTR_POP_VLAN: |
e74d4817 | 1322 | err = pop_vlan(skb, key); |
064af421 BP |
1323 | break; |
1324 | ||
d4004f8d AZ |
1325 | case OVS_ACTION_ATTR_RECIRC: { |
1326 | bool last = nla_is_last(a, rem); | |
1327 | ||
1328 | err = execute_recirc(dp, skb, key, a, last); | |
1329 | if (last) { | |
867e37ba AZ |
1330 | /* If this is the last action, the skb has |
1331 | * been consumed or freed. | |
e74d4817 PS |
1332 | * Return immediately. |
1333 | */ | |
867e37ba AZ |
1334 | return err; |
1335 | } | |
a6059080 | 1336 | break; |
d4004f8d | 1337 | } |
a6059080 | 1338 | |
4edb9ae9 | 1339 | case OVS_ACTION_ATTR_SET: |
e74d4817 | 1340 | err = execute_set_action(skb, key, nla_data(a)); |
064af421 | 1341 | break; |
c1c9c9c4 | 1342 | |
b940b3d7 JR |
1343 | case OVS_ACTION_ATTR_SET_MASKED: |
1344 | case OVS_ACTION_ATTR_SET_TO_MASKED: | |
1345 | err = execute_masked_set_action(skb, key, nla_data(a)); | |
1346 | break; | |
1347 | ||
7d9595e3 AZ |
1348 | case OVS_ACTION_ATTR_SAMPLE: { |
1349 | bool last = nla_is_last(a, rem); | |
1350 | ||
1351 | err = sample(dp, skb, key, a, last); | |
1352 | if (last) | |
1353 | return err; | |
1354 | ||
6ff686f2 | 1355 | break; |
7d9595e3 | 1356 | } |
a94ebc39 JS |
1357 | |
1358 | case OVS_ACTION_ATTR_CT: | |
c05e2094 JS |
1359 | if (!is_flow_key_valid(key)) { |
1360 | err = ovs_flow_key_update(skb, key); | |
1361 | if (err) | |
1362 | return err; | |
1363 | } | |
1364 | ||
a94ebc39 JS |
1365 | err = ovs_ct_execute(ovs_dp_get_net(dp), skb, key, |
1366 | nla_data(a)); | |
1367 | ||
1368 | /* Hide stolen IP fragments from user space. */ | |
c05e2094 JS |
1369 | if (err) |
1370 | return err == -EINPROGRESS ? 0 : err; | |
a94ebc39 | 1371 | break; |
6fcecb85 | 1372 | |
0cdfdddd EG |
1373 | case OVS_ACTION_ATTR_CT_CLEAR: |
1374 | err = ovs_ct_clear(skb, key); | |
1375 | break; | |
1376 | ||
6fcecb85 YY |
1377 | case OVS_ACTION_ATTR_PUSH_ETH: |
1378 | err = push_eth(skb, key, nla_data(a)); | |
1379 | break; | |
1380 | ||
1381 | case OVS_ACTION_ATTR_POP_ETH: | |
1382 | err = pop_eth(skb, key); | |
1383 | break; | |
96b82f6d YY |
1384 | |
1385 | case OVS_ACTION_ATTR_PUSH_NSH: { | |
1386 | u8 buffer[NSH_HDR_MAX_LEN]; | |
1387 | struct nshhdr *nh = (struct nshhdr *)buffer; | |
1388 | ||
1389 | err = nsh_hdr_from_nlattr(nla_data(a), nh, | |
1390 | NSH_HDR_MAX_LEN); | |
1391 | if (unlikely(err)) | |
1392 | break; | |
1393 | err = push_nsh(skb, key, nh); | |
1394 | break; | |
1395 | } | |
1396 | ||
1397 | case OVS_ACTION_ATTR_POP_NSH: | |
1398 | err = pop_nsh(skb, key); | |
1399 | break; | |
e02b6f81 AZ |
1400 | |
1401 | case OVS_ACTION_ATTR_METER: | |
1402 | if (ovs_meter_execute(dp, skb, key, nla_get_u32(a))) { | |
1403 | consume_skb(skb); | |
1404 | return 0; | |
1405 | } | |
5a77cf96 YS |
1406 | break; |
1407 | ||
1408 | case OVS_ACTION_ATTR_CLONE: { | |
1409 | bool last = nla_is_last(a, rem); | |
1410 | ||
1411 | err = clone(dp, skb, key, a, last); | |
1412 | if (last) | |
1413 | return err; | |
1414 | break; | |
1415 | } | |
faa3beb4 NS |
1416 | |
1417 | case OVS_ACTION_ATTR_CHECK_PKT_LEN: { | |
1418 | bool last = nla_is_last(a, rem); | |
1419 | ||
1420 | err = execute_check_pkt_len(dp, skb, key, a, last); | |
1421 | if (last) | |
1422 | return err; | |
1423 | ||
1424 | break; | |
1425 | } | |
6ff686f2 | 1426 | } |
15c39847 | 1427 | |
10db8b20 JG |
1428 | if (unlikely(err)) { |
1429 | kfree_skb(skb); | |
1430 | return err; | |
1431 | } | |
064af421 | 1432 | } |
6c222e55 | 1433 | |
2b6a82da | 1434 | consume_skb(skb); |
a5225dd6 | 1435 | return 0; |
064af421 | 1436 | } |
871dfe07 | 1437 | |
d4004f8d AZ |
1438 | /* Execute the actions on the clone of the packet. The effect of the |
1439 | * execution does not affect the original 'skb' nor the original 'key'. | |
1440 | * | |
1441 | * The execution may be deferred in case the actions can not be executed | |
1442 | * immediately. | |
1443 | */ | |
1444 | static int clone_execute(struct datapath *dp, struct sk_buff *skb, | |
1445 | struct sw_flow_key *key, u32 recirc_id, | |
1446 | const struct nlattr *actions, int len, | |
1447 | bool last, bool clone_flow_key) | |
1448 | { | |
1449 | struct deferred_action *da; | |
1450 | struct sw_flow_key *clone; | |
1451 | ||
1452 | skb = last ? skb : skb_clone(skb, GFP_ATOMIC); | |
1453 | if (!skb) { | |
1454 | /* Out of memory, skip this action. | |
1455 | */ | |
1456 | return 0; | |
1457 | } | |
1458 | ||
1459 | /* When clone_flow_key is false, the 'key' will not be change | |
1460 | * by the actions, then the 'key' can be used directly. | |
1461 | * Otherwise, try to clone key from the next recursion level of | |
1462 | * 'flow_keys'. If clone is successful, execute the actions | |
1463 | * without deferring. | |
1464 | */ | |
1465 | clone = clone_flow_key ? clone_key(key) : key; | |
1466 | if (clone) { | |
1467 | int err = 0; | |
1468 | ||
1469 | if (actions) { /* Sample action */ | |
1470 | if (clone_flow_key) | |
1471 | __this_cpu_inc(exec_actions_level); | |
1472 | ||
1473 | err = do_execute_actions(dp, skb, clone, | |
1474 | actions, len); | |
1475 | ||
1476 | if (clone_flow_key) | |
1477 | __this_cpu_dec(exec_actions_level); | |
1478 | } else { /* Recirc action */ | |
1479 | clone->recirc_id = recirc_id; | |
1480 | ovs_dp_process_packet(skb, clone); | |
1481 | } | |
1482 | return err; | |
1483 | } | |
1484 | ||
1485 | /* Out of 'flow_keys' space. Defer actions */ | |
1486 | da = add_deferred_actions(skb, key, actions, len); | |
1487 | if (da) { | |
1488 | if (!actions) { /* Recirc action */ | |
1489 | key = &da->pkt_key; | |
1490 | key->recirc_id = recirc_id; | |
1491 | } | |
1492 | } else { | |
1493 | /* Out of per CPU action FIFO space. Drop the 'skb' and | |
1494 | * log an error. | |
1495 | */ | |
1496 | kfree_skb(skb); | |
1497 | ||
1498 | if (net_ratelimit()) { | |
1499 | if (actions) { /* Sample action */ | |
1500 | pr_warn("%s: deferred action limit reached, drop sample action\n", | |
1501 | ovs_dp_name(dp)); | |
1502 | } else { /* Recirc action */ | |
1503 | pr_warn("%s: deferred action limit reached, drop recirc action\n", | |
1504 | ovs_dp_name(dp)); | |
1505 | } | |
1506 | } | |
1507 | } | |
1508 | return 0; | |
1509 | } | |
1510 | ||
2c8c4fb7 AZ |
1511 | static void process_deferred_actions(struct datapath *dp) |
1512 | { | |
1513 | struct action_fifo *fifo = this_cpu_ptr(action_fifos); | |
1514 | ||
1515 | /* Do not touch the FIFO in case there is no deferred actions. */ | |
1516 | if (action_fifo_is_empty(fifo)) | |
1517 | return; | |
1518 | ||
1519 | /* Finishing executing all deferred actions. */ | |
1520 | do { | |
1521 | struct deferred_action *da = action_fifo_get(fifo); | |
1522 | struct sk_buff *skb = da->skb; | |
7d16c847 | 1523 | struct sw_flow_key *key = &da->pkt_key; |
2c8c4fb7 | 1524 | const struct nlattr *actions = da->actions; |
615fa7ba | 1525 | int actions_len = da->actions_len; |
2c8c4fb7 AZ |
1526 | |
1527 | if (actions) | |
615fa7ba | 1528 | do_execute_actions(dp, skb, key, actions, actions_len); |
2c8c4fb7 | 1529 | else |
7d16c847 | 1530 | ovs_dp_process_packet(skb, key); |
2c8c4fb7 AZ |
1531 | } while (!action_fifo_is_empty(fifo)); |
1532 | ||
1533 | /* Reset FIFO for the next packet. */ | |
1534 | action_fifo_init(fifo); | |
1535 | } | |
1536 | ||
871dfe07 | 1537 | /* Execute a list of actions against 'skb'. */ |
2c8c4fb7 | 1538 | int ovs_execute_actions(struct datapath *dp, struct sk_buff *skb, |
7d16c847 PS |
1539 | const struct sw_flow_actions *acts, |
1540 | struct sw_flow_key *key) | |
2c8c4fb7 | 1541 | { |
77a9a338 | 1542 | int err, level; |
2c8c4fb7 | 1543 | |
77a9a338 | 1544 | level = __this_cpu_inc_return(exec_actions_level); |
615539ab | 1545 | if (unlikely(level > OVS_RECURSION_LIMIT)) { |
77a9a338 PS |
1546 | net_crit_ratelimited("ovs: recursion limit reached on datapath %s, probable configuration error\n", |
1547 | ovs_dp_name(dp)); | |
2c8c4fb7 | 1548 | kfree_skb(skb); |
77a9a338 PS |
1549 | err = -ENETDOWN; |
1550 | goto out; | |
2c8c4fb7 AZ |
1551 | } |
1552 | ||
6b330a60 | 1553 | OVS_CB(skb)->acts_origlen = acts->orig_len; |
7d16c847 PS |
1554 | err = do_execute_actions(dp, skb, key, |
1555 | acts->actions, acts->actions_len); | |
2c8c4fb7 | 1556 | |
77a9a338 | 1557 | if (level == 1) |
2c8c4fb7 AZ |
1558 | process_deferred_actions(dp); |
1559 | ||
77a9a338 PS |
1560 | out: |
1561 | __this_cpu_dec(exec_actions_level); | |
2c8c4fb7 AZ |
1562 | return err; |
1563 | } | |
1564 | ||
1565 | int action_fifos_init(void) | |
1566 | { | |
1567 | action_fifos = alloc_percpu(struct action_fifo); | |
1568 | if (!action_fifos) | |
1569 | return -ENOMEM; | |
1570 | ||
58afdef1 AZ |
1571 | flow_keys = alloc_percpu(struct action_flow_keys); |
1572 | if (!flow_keys) { | |
615539ab LR |
1573 | free_percpu(action_fifos); |
1574 | return -ENOMEM; | |
1575 | } | |
1576 | ||
2c8c4fb7 AZ |
1577 | return 0; |
1578 | } | |
1579 | ||
1580 | void action_fifos_exit(void) | |
60759b2b | 1581 | { |
2c8c4fb7 | 1582 | free_percpu(action_fifos); |
58afdef1 | 1583 | free_percpu(flow_keys); |
871dfe07 | 1584 | } |