1 /* 2 * Copyright (c) 2007-2011 Nicira Networks. 3 * 4 * This program is free software; you can redistribute it and/or 5 * modify it under the terms of version 2 of the GNU General Public 6 * License as published by the Free Software Foundation. 7 * 8 * This program is distributed in the hope that it will be useful, but 9 * WITHOUT ANY WARRANTY; without even the implied warranty of 10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU 11 * General Public License for more details. 12 * 13 * You should have received a copy of the GNU General Public License 14 * along with this program; if not, write to the Free Software 15 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 16 * 02110-1301, USA 17 */ 18 19 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt 20 21 #include <linux/skbuff.h> 22 #include <linux/in.h> 23 #include <linux/ip.h> 24 #include <linux/openvswitch.h> 25 #include <linux/tcp.h> 26 #include <linux/udp.h> 27 #include <linux/in6.h> 28 #include <linux/if_arp.h> 29 #include <linux/if_vlan.h> 30 #include <net/ip.h> 31 #include <net/checksum.h> 32 #include <net/dsfield.h> 33 34 #include "datapath.h" 35 #include "vport.h" 36 37 static int do_execute_actions(struct datapath *dp, struct sk_buff *skb, 38 const struct nlattr *attr, int len, bool keep_skb); 39 40 static int make_writable(struct sk_buff *skb, int write_len) 41 { 42 if (!skb_cloned(skb) || skb_clone_writable(skb, write_len)) 43 return 0; 44 45 return pskb_expand_head(skb, 0, 0, GFP_ATOMIC); 46 } 47 48 /* remove VLAN header from packet and update csum accrodingly. */ 49 static int __pop_vlan_tci(struct sk_buff *skb, __be16 *current_tci) 50 { 51 struct vlan_hdr *vhdr; 52 int err; 53 54 err = make_writable(skb, VLAN_ETH_HLEN); 55 if (unlikely(err)) 56 return err; 57 58 if (skb->ip_summed == CHECKSUM_COMPLETE) 59 skb->csum = csum_sub(skb->csum, csum_partial(skb->data 60 + ETH_HLEN, VLAN_HLEN, 0)); 61 62 vhdr = (struct vlan_hdr *)(skb->data + ETH_HLEN); 63 *current_tci = vhdr->h_vlan_TCI; 64 65 memmove(skb->data + VLAN_HLEN, skb->data, 2 * ETH_ALEN); 66 __skb_pull(skb, VLAN_HLEN); 67 68 vlan_set_encap_proto(skb, vhdr); 69 skb->mac_header += VLAN_HLEN; 70 skb_reset_mac_len(skb); 71 72 return 0; 73 } 74 75 static int pop_vlan(struct sk_buff *skb) 76 { 77 __be16 tci; 78 int err; 79 80 if (likely(vlan_tx_tag_present(skb))) { 81 skb->vlan_tci = 0; 82 } else { 83 if (unlikely(skb->protocol != htons(ETH_P_8021Q) || 84 skb->len < VLAN_ETH_HLEN)) 85 return 0; 86 87 err = __pop_vlan_tci(skb, &tci); 88 if (err) 89 return err; 90 } 91 /* move next vlan tag to hw accel tag */ 92 if (likely(skb->protocol != htons(ETH_P_8021Q) || 93 skb->len < VLAN_ETH_HLEN)) 94 return 0; 95 96 err = __pop_vlan_tci(skb, &tci); 97 if (unlikely(err)) 98 return err; 99 100 __vlan_hwaccel_put_tag(skb, ntohs(tci)); 101 return 0; 102 } 103 104 static int push_vlan(struct sk_buff *skb, const struct ovs_action_push_vlan *vlan) 105 { 106 if (unlikely(vlan_tx_tag_present(skb))) { 107 u16 current_tag; 108 109 /* push down current VLAN tag */ 110 current_tag = vlan_tx_tag_get(skb); 111 112 if (!__vlan_put_tag(skb, current_tag)) 113 return -ENOMEM; 114 115 if (skb->ip_summed == CHECKSUM_COMPLETE) 116 skb->csum = csum_add(skb->csum, csum_partial(skb->data 117 + ETH_HLEN, VLAN_HLEN, 0)); 118 119 } 120 __vlan_hwaccel_put_tag(skb, ntohs(vlan->vlan_tci) & ~VLAN_TAG_PRESENT); 121 return 0; 122 } 123 124 static int set_eth_addr(struct sk_buff *skb, 125 const struct ovs_key_ethernet *eth_key) 126 { 127 int err; 128 err = make_writable(skb, ETH_HLEN); 129 if (unlikely(err)) 130 return err; 131 132 memcpy(eth_hdr(skb)->h_source, eth_key->eth_src, ETH_ALEN); 133 memcpy(eth_hdr(skb)->h_dest, eth_key->eth_dst, ETH_ALEN); 134 135 return 0; 136 } 137 138 static void set_ip_addr(struct sk_buff *skb, struct iphdr *nh, 139 __be32 *addr, __be32 new_addr) 140 { 141 int transport_len = skb->len - skb_transport_offset(skb); 142 143 if (nh->protocol == IPPROTO_TCP) { 144 if (likely(transport_len >= sizeof(struct tcphdr))) 145 inet_proto_csum_replace4(&tcp_hdr(skb)->check, skb, 146 *addr, new_addr, 1); 147 } else if (nh->protocol == IPPROTO_UDP) { 148 if (likely(transport_len >= sizeof(struct udphdr))) 149 inet_proto_csum_replace4(&udp_hdr(skb)->check, skb, 150 *addr, new_addr, 1); 151 } 152 153 csum_replace4(&nh->check, *addr, new_addr); 154 skb->rxhash = 0; 155 *addr = new_addr; 156 } 157 158 static void set_ip_ttl(struct sk_buff *skb, struct iphdr *nh, u8 new_ttl) 159 { 160 csum_replace2(&nh->check, htons(nh->ttl << 8), htons(new_ttl << 8)); 161 nh->ttl = new_ttl; 162 } 163 164 static int set_ipv4(struct sk_buff *skb, const struct ovs_key_ipv4 *ipv4_key) 165 { 166 struct iphdr *nh; 167 int err; 168 169 err = make_writable(skb, skb_network_offset(skb) + 170 sizeof(struct iphdr)); 171 if (unlikely(err)) 172 return err; 173 174 nh = ip_hdr(skb); 175 176 if (ipv4_key->ipv4_src != nh->saddr) 177 set_ip_addr(skb, nh, &nh->saddr, ipv4_key->ipv4_src); 178 179 if (ipv4_key->ipv4_dst != nh->daddr) 180 set_ip_addr(skb, nh, &nh->daddr, ipv4_key->ipv4_dst); 181 182 if (ipv4_key->ipv4_tos != nh->tos) 183 ipv4_change_dsfield(nh, 0, ipv4_key->ipv4_tos); 184 185 if (ipv4_key->ipv4_ttl != nh->ttl) 186 set_ip_ttl(skb, nh, ipv4_key->ipv4_ttl); 187 188 return 0; 189 } 190 191 /* Must follow make_writable() since that can move the skb data. */ 192 static void set_tp_port(struct sk_buff *skb, __be16 *port, 193 __be16 new_port, __sum16 *check) 194 { 195 inet_proto_csum_replace2(check, skb, *port, new_port, 0); 196 *port = new_port; 197 skb->rxhash = 0; 198 } 199 200 static int set_udp_port(struct sk_buff *skb, 201 const struct ovs_key_udp *udp_port_key) 202 { 203 struct udphdr *uh; 204 int err; 205 206 err = make_writable(skb, skb_transport_offset(skb) + 207 sizeof(struct udphdr)); 208 if (unlikely(err)) 209 return err; 210 211 uh = udp_hdr(skb); 212 if (udp_port_key->udp_src != uh->source) 213 set_tp_port(skb, &uh->source, udp_port_key->udp_src, &uh->check); 214 215 if (udp_port_key->udp_dst != uh->dest) 216 set_tp_port(skb, &uh->dest, udp_port_key->udp_dst, &uh->check); 217 218 return 0; 219 } 220 221 static int set_tcp_port(struct sk_buff *skb, 222 const struct ovs_key_tcp *tcp_port_key) 223 { 224 struct tcphdr *th; 225 int err; 226 227 err = make_writable(skb, skb_transport_offset(skb) + 228 sizeof(struct tcphdr)); 229 if (unlikely(err)) 230 return err; 231 232 th = tcp_hdr(skb); 233 if (tcp_port_key->tcp_src != th->source) 234 set_tp_port(skb, &th->source, tcp_port_key->tcp_src, &th->check); 235 236 if (tcp_port_key->tcp_dst != th->dest) 237 set_tp_port(skb, &th->dest, tcp_port_key->tcp_dst, &th->check); 238 239 return 0; 240 } 241 242 static int do_output(struct datapath *dp, struct sk_buff *skb, int out_port) 243 { 244 struct vport *vport; 245 246 if (unlikely(!skb)) 247 return -ENOMEM; 248 249 vport = rcu_dereference(dp->ports[out_port]); 250 if (unlikely(!vport)) { 251 kfree_skb(skb); 252 return -ENODEV; 253 } 254 255 ovs_vport_send(vport, skb); 256 return 0; 257 } 258 259 static int output_userspace(struct datapath *dp, struct sk_buff *skb, 260 const struct nlattr *attr) 261 { 262 struct dp_upcall_info upcall; 263 const struct nlattr *a; 264 int rem; 265 266 upcall.cmd = OVS_PACKET_CMD_ACTION; 267 upcall.key = &OVS_CB(skb)->flow->key; 268 upcall.userdata = NULL; 269 upcall.pid = 0; 270 271 for (a = nla_data(attr), rem = nla_len(attr); rem > 0; 272 a = nla_next(a, &rem)) { 273 switch (nla_type(a)) { 274 case OVS_USERSPACE_ATTR_USERDATA: 275 upcall.userdata = a; 276 break; 277 278 case OVS_USERSPACE_ATTR_PID: 279 upcall.pid = nla_get_u32(a); 280 break; 281 } 282 } 283 284 return ovs_dp_upcall(dp, skb, &upcall); 285 } 286 287 static int sample(struct datapath *dp, struct sk_buff *skb, 288 const struct nlattr *attr) 289 { 290 const struct nlattr *acts_list = NULL; 291 const struct nlattr *a; 292 int rem; 293 294 for (a = nla_data(attr), rem = nla_len(attr); rem > 0; 295 a = nla_next(a, &rem)) { 296 switch (nla_type(a)) { 297 case OVS_SAMPLE_ATTR_PROBABILITY: 298 if (net_random() >= nla_get_u32(a)) 299 return 0; 300 break; 301 302 case OVS_SAMPLE_ATTR_ACTIONS: 303 acts_list = a; 304 break; 305 } 306 } 307 308 return do_execute_actions(dp, skb, nla_data(acts_list), 309 nla_len(acts_list), true); 310 } 311 312 static int execute_set_action(struct sk_buff *skb, 313 const struct nlattr *nested_attr) 314 { 315 int err = 0; 316 317 switch (nla_type(nested_attr)) { 318 case OVS_KEY_ATTR_PRIORITY: 319 skb->priority = nla_get_u32(nested_attr); 320 break; 321 322 case OVS_KEY_ATTR_ETHERNET: 323 err = set_eth_addr(skb, nla_data(nested_attr)); 324 break; 325 326 case OVS_KEY_ATTR_IPV4: 327 err = set_ipv4(skb, nla_data(nested_attr)); 328 break; 329 330 case OVS_KEY_ATTR_TCP: 331 err = set_tcp_port(skb, nla_data(nested_attr)); 332 break; 333 334 case OVS_KEY_ATTR_UDP: 335 err = set_udp_port(skb, nla_data(nested_attr)); 336 break; 337 } 338 339 return err; 340 } 341 342 /* Execute a list of actions against 'skb'. */ 343 static int do_execute_actions(struct datapath *dp, struct sk_buff *skb, 344 const struct nlattr *attr, int len, bool keep_skb) 345 { 346 /* Every output action needs a separate clone of 'skb', but the common 347 * case is just a single output action, so that doing a clone and 348 * then freeing the original skbuff is wasteful. So the following code 349 * is slightly obscure just to avoid that. */ 350 int prev_port = -1; 351 const struct nlattr *a; 352 int rem; 353 354 for (a = attr, rem = len; rem > 0; 355 a = nla_next(a, &rem)) { 356 int err = 0; 357 358 if (prev_port != -1) { 359 do_output(dp, skb_clone(skb, GFP_ATOMIC), prev_port); 360 prev_port = -1; 361 } 362 363 switch (nla_type(a)) { 364 case OVS_ACTION_ATTR_OUTPUT: 365 prev_port = nla_get_u32(a); 366 break; 367 368 case OVS_ACTION_ATTR_USERSPACE: 369 output_userspace(dp, skb, a); 370 break; 371 372 case OVS_ACTION_ATTR_PUSH_VLAN: 373 err = push_vlan(skb, nla_data(a)); 374 if (unlikely(err)) /* skb already freed. */ 375 return err; 376 break; 377 378 case OVS_ACTION_ATTR_POP_VLAN: 379 err = pop_vlan(skb); 380 break; 381 382 case OVS_ACTION_ATTR_SET: 383 err = execute_set_action(skb, nla_data(a)); 384 break; 385 386 case OVS_ACTION_ATTR_SAMPLE: 387 err = sample(dp, skb, a); 388 break; 389 } 390 391 if (unlikely(err)) { 392 kfree_skb(skb); 393 return err; 394 } 395 } 396 397 if (prev_port != -1) { 398 if (keep_skb) 399 skb = skb_clone(skb, GFP_ATOMIC); 400 401 do_output(dp, skb, prev_port); 402 } else if (!keep_skb) 403 consume_skb(skb); 404 405 return 0; 406 } 407 408 /* Execute a list of actions against 'skb'. */ 409 int ovs_execute_actions(struct datapath *dp, struct sk_buff *skb) 410 { 411 struct sw_flow_actions *acts = rcu_dereference(OVS_CB(skb)->flow->sf_acts); 412 413 return do_execute_actions(dp, skb, acts->actions, 414 acts->actions_len, false); 415 } 416