]>
Commit | Line | Data |
---|---|---|
ccb1352e | 1 | /* |
971427f3 | 2 | * Copyright (c) 2007-2014 Nicira, Inc. |
ccb1352e JG |
3 | * |
4 | * This program is free software; you can redistribute it and/or | |
5 | * modify it under the terms of version 2 of the GNU General Public | |
6 | * License as published by the Free Software Foundation. | |
7 | * | |
8 | * This program is distributed in the hope that it will be useful, but | |
9 | * WITHOUT ANY WARRANTY; without even the implied warranty of | |
10 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
11 | * General Public License for more details. | |
12 | * | |
13 | * You should have received a copy of the GNU General Public License | |
14 | * along with this program; if not, write to the Free Software | |
15 | * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA | |
16 | * 02110-1301, USA | |
17 | */ | |
18 | ||
ccb1352e JG |
19 | #include <linux/uaccess.h> |
20 | #include <linux/netdevice.h> | |
21 | #include <linux/etherdevice.h> | |
22 | #include <linux/if_ether.h> | |
23 | #include <linux/if_vlan.h> | |
24 | #include <net/llc_pdu.h> | |
25 | #include <linux/kernel.h> | |
26 | #include <linux/jhash.h> | |
27 | #include <linux/jiffies.h> | |
28 | #include <linux/llc.h> | |
29 | #include <linux/module.h> | |
30 | #include <linux/in.h> | |
31 | #include <linux/rcupdate.h> | |
db74a333 | 32 | #include <linux/cpumask.h> |
ccb1352e | 33 | #include <linux/if_arp.h> |
ccb1352e JG |
34 | #include <linux/ip.h> |
35 | #include <linux/ipv6.h> | |
25cd9ba0 | 36 | #include <linux/mpls.h> |
a175a723 | 37 | #include <linux/sctp.h> |
e298e505 | 38 | #include <linux/smp.h> |
ccb1352e JG |
39 | #include <linux/tcp.h> |
40 | #include <linux/udp.h> | |
41 | #include <linux/icmp.h> | |
42 | #include <linux/icmpv6.h> | |
43 | #include <linux/rculist.h> | |
44 | #include <net/ip.h> | |
7d5437c7 | 45 | #include <net/ip_tunnels.h> |
ccb1352e | 46 | #include <net/ipv6.h> |
25cd9ba0 | 47 | #include <net/mpls.h> |
ccb1352e JG |
48 | #include <net/ndisc.h> |
49 | ||
a581b96d | 50 | #include "conntrack.h" |
83c8df26 PS |
51 | #include "datapath.h" |
52 | #include "flow.h" | |
53 | #include "flow_netlink.h" | |
a581b96d | 54 | #include "vport.h" |
83c8df26 | 55 | |
e6445719 | 56 | u64 ovs_flow_used_time(unsigned long flow_jiffies) |
03f0d916 | 57 | { |
e6445719 PS |
58 | struct timespec cur_ts; |
59 | u64 cur_ms, idle_ms; | |
03f0d916 | 60 | |
e6445719 PS |
61 | ktime_get_ts(&cur_ts); |
62 | idle_ms = jiffies_to_msecs(jiffies - flow_jiffies); | |
63 | cur_ms = (u64)cur_ts.tv_sec * MSEC_PER_SEC + | |
64 | cur_ts.tv_nsec / NSEC_PER_MSEC; | |
03f0d916 | 65 | |
e6445719 | 66 | return cur_ms - idle_ms; |
5828cd9a AZ |
67 | } |
68 | ||
df23e9f6 | 69 | #define TCP_FLAGS_BE16(tp) (*(__be16 *)&tcp_flag_word(tp) & htons(0x0FFF)) |
03f0d916 | 70 | |
ad552007 | 71 | void ovs_flow_stats_update(struct sw_flow *flow, __be16 tcp_flags, |
12eb18f7 | 72 | const struct sk_buff *skb) |
03f0d916 | 73 | { |
e298e505 | 74 | struct flow_stats *stats; |
63e7959c | 75 | int node = numa_node_id(); |
db74a333 | 76 | int cpu = smp_processor_id(); |
df8a39de | 77 | int len = skb->len + (skb_vlan_tag_present(skb) ? VLAN_HLEN : 0); |
03f0d916 | 78 | |
db74a333 | 79 | stats = rcu_dereference(flow->stats[cpu]); |
e298e505 | 80 | |
db74a333 | 81 | /* Check if already have CPU-specific stats. */ |
63e7959c JR |
82 | if (likely(stats)) { |
83 | spin_lock(&stats->lock); | |
84 | /* Mark if we write on the pre-allocated stats. */ | |
db74a333 TLSC |
85 | if (cpu == 0 && unlikely(flow->stats_last_writer != cpu)) |
86 | flow->stats_last_writer = cpu; | |
63e7959c JR |
87 | } else { |
88 | stats = rcu_dereference(flow->stats[0]); /* Pre-allocated. */ | |
89 | spin_lock(&stats->lock); | |
90 | ||
db74a333 | 91 | /* If the current CPU is the only writer on the |
63e7959c JR |
92 | * pre-allocated stats keep using them. |
93 | */ | |
db74a333 | 94 | if (unlikely(flow->stats_last_writer != cpu)) { |
63e7959c | 95 | /* A previous locker may have already allocated the |
db74a333 | 96 | * stats, so we need to check again. If CPU-specific |
63e7959c JR |
97 | * stats were already allocated, we update the pre- |
98 | * allocated stats as we have already locked them. | |
99 | */ | |
db74a333 TLSC |
100 | if (likely(flow->stats_last_writer != -1) && |
101 | likely(!rcu_access_pointer(flow->stats[cpu]))) { | |
102 | /* Try to allocate CPU-specific stats. */ | |
63e7959c JR |
103 | struct flow_stats *new_stats; |
104 | ||
105 | new_stats = | |
106 | kmem_cache_alloc_node(flow_stats_cache, | |
4167e9b2 DR |
107 | GFP_NOWAIT | |
108 | __GFP_THISNODE | | |
109 | __GFP_NOWARN | | |
63e7959c JR |
110 | __GFP_NOMEMALLOC, |
111 | node); | |
112 | if (likely(new_stats)) { | |
113 | new_stats->used = jiffies; | |
114 | new_stats->packet_count = 1; | |
24cc59d1 | 115 | new_stats->byte_count = len; |
63e7959c JR |
116 | new_stats->tcp_flags = tcp_flags; |
117 | spin_lock_init(&new_stats->lock); | |
118 | ||
db74a333 | 119 | rcu_assign_pointer(flow->stats[cpu], |
63e7959c JR |
120 | new_stats); |
121 | goto unlock; | |
122 | } | |
123 | } | |
db74a333 | 124 | flow->stats_last_writer = cpu; |
63e7959c JR |
125 | } |
126 | } | |
127 | ||
e298e505 PS |
128 | stats->used = jiffies; |
129 | stats->packet_count++; | |
24cc59d1 | 130 | stats->byte_count += len; |
e298e505 | 131 | stats->tcp_flags |= tcp_flags; |
63e7959c | 132 | unlock: |
e298e505 PS |
133 | spin_unlock(&stats->lock); |
134 | } | |
135 | ||
86ec8dba JR |
136 | /* Must be called with rcu_read_lock or ovs_mutex. */ |
137 | void ovs_flow_stats_get(const struct sw_flow *flow, | |
138 | struct ovs_flow_stats *ovs_stats, | |
e298e505 PS |
139 | unsigned long *used, __be16 *tcp_flags) |
140 | { | |
db74a333 | 141 | int cpu; |
e298e505 PS |
142 | |
143 | *used = 0; | |
144 | *tcp_flags = 0; | |
145 | memset(ovs_stats, 0, sizeof(*ovs_stats)); | |
146 | ||
db74a333 TLSC |
147 | /* We open code this to make sure cpu 0 is always considered */ |
148 | for (cpu = 0; cpu < nr_cpu_ids; cpu = cpumask_next(cpu, cpu_possible_mask)) { | |
149 | struct flow_stats *stats = rcu_dereference_ovsl(flow->stats[cpu]); | |
63e7959c JR |
150 | |
151 | if (stats) { | |
152 | /* Local CPU may write on non-local stats, so we must | |
153 | * block bottom-halves here. | |
154 | */ | |
155 | spin_lock_bh(&stats->lock); | |
156 | if (!*used || time_after(stats->used, *used)) | |
157 | *used = stats->used; | |
158 | *tcp_flags |= stats->tcp_flags; | |
159 | ovs_stats->n_packets += stats->packet_count; | |
160 | ovs_stats->n_bytes += stats->byte_count; | |
161 | spin_unlock_bh(&stats->lock); | |
162 | } | |
e298e505 | 163 | } |
e298e505 PS |
164 | } |
165 | ||
86ec8dba | 166 | /* Called with ovs_mutex. */ |
e298e505 PS |
167 | void ovs_flow_stats_clear(struct sw_flow *flow) |
168 | { | |
db74a333 | 169 | int cpu; |
63e7959c | 170 | |
db74a333 TLSC |
171 | /* We open code this to make sure cpu 0 is always considered */ |
172 | for (cpu = 0; cpu < nr_cpu_ids; cpu = cpumask_next(cpu, cpu_possible_mask)) { | |
173 | struct flow_stats *stats = ovsl_dereference(flow->stats[cpu]); | |
63e7959c JR |
174 | |
175 | if (stats) { | |
176 | spin_lock_bh(&stats->lock); | |
177 | stats->used = 0; | |
178 | stats->packet_count = 0; | |
179 | stats->byte_count = 0; | |
180 | stats->tcp_flags = 0; | |
181 | spin_unlock_bh(&stats->lock); | |
182 | } | |
183 | } | |
03f0d916 AZ |
184 | } |
185 | ||
ccb1352e JG |
186 | static int check_header(struct sk_buff *skb, int len) |
187 | { | |
188 | if (unlikely(skb->len < len)) | |
189 | return -EINVAL; | |
190 | if (unlikely(!pskb_may_pull(skb, len))) | |
191 | return -ENOMEM; | |
192 | return 0; | |
193 | } | |
194 | ||
195 | static bool arphdr_ok(struct sk_buff *skb) | |
196 | { | |
197 | return pskb_may_pull(skb, skb_network_offset(skb) + | |
198 | sizeof(struct arp_eth_header)); | |
199 | } | |
200 | ||
201 | static int check_iphdr(struct sk_buff *skb) | |
202 | { | |
203 | unsigned int nh_ofs = skb_network_offset(skb); | |
204 | unsigned int ip_len; | |
205 | int err; | |
206 | ||
207 | err = check_header(skb, nh_ofs + sizeof(struct iphdr)); | |
208 | if (unlikely(err)) | |
209 | return err; | |
210 | ||
211 | ip_len = ip_hdrlen(skb); | |
212 | if (unlikely(ip_len < sizeof(struct iphdr) || | |
213 | skb->len < nh_ofs + ip_len)) | |
214 | return -EINVAL; | |
215 | ||
216 | skb_set_transport_header(skb, nh_ofs + ip_len); | |
217 | return 0; | |
218 | } | |
219 | ||
220 | static bool tcphdr_ok(struct sk_buff *skb) | |
221 | { | |
222 | int th_ofs = skb_transport_offset(skb); | |
223 | int tcp_len; | |
224 | ||
225 | if (unlikely(!pskb_may_pull(skb, th_ofs + sizeof(struct tcphdr)))) | |
226 | return false; | |
227 | ||
228 | tcp_len = tcp_hdrlen(skb); | |
229 | if (unlikely(tcp_len < sizeof(struct tcphdr) || | |
230 | skb->len < th_ofs + tcp_len)) | |
231 | return false; | |
232 | ||
233 | return true; | |
234 | } | |
235 | ||
236 | static bool udphdr_ok(struct sk_buff *skb) | |
237 | { | |
238 | return pskb_may_pull(skb, skb_transport_offset(skb) + | |
239 | sizeof(struct udphdr)); | |
240 | } | |
241 | ||
a175a723 JS |
242 | static bool sctphdr_ok(struct sk_buff *skb) |
243 | { | |
244 | return pskb_may_pull(skb, skb_transport_offset(skb) + | |
245 | sizeof(struct sctphdr)); | |
246 | } | |
247 | ||
ccb1352e JG |
248 | static bool icmphdr_ok(struct sk_buff *skb) |
249 | { | |
250 | return pskb_may_pull(skb, skb_transport_offset(skb) + | |
251 | sizeof(struct icmphdr)); | |
252 | } | |
253 | ||
03f0d916 | 254 | static int parse_ipv6hdr(struct sk_buff *skb, struct sw_flow_key *key) |
ccb1352e JG |
255 | { |
256 | unsigned int nh_ofs = skb_network_offset(skb); | |
257 | unsigned int nh_len; | |
258 | int payload_ofs; | |
259 | struct ipv6hdr *nh; | |
260 | uint8_t nexthdr; | |
261 | __be16 frag_off; | |
262 | int err; | |
263 | ||
ccb1352e JG |
264 | err = check_header(skb, nh_ofs + sizeof(*nh)); |
265 | if (unlikely(err)) | |
266 | return err; | |
267 | ||
268 | nh = ipv6_hdr(skb); | |
269 | nexthdr = nh->nexthdr; | |
270 | payload_ofs = (u8 *)(nh + 1) - skb->data; | |
271 | ||
272 | key->ip.proto = NEXTHDR_NONE; | |
273 | key->ip.tos = ipv6_get_dsfield(nh); | |
274 | key->ip.ttl = nh->hop_limit; | |
275 | key->ipv6.label = *(__be32 *)nh & htonl(IPV6_FLOWINFO_FLOWLABEL); | |
276 | key->ipv6.addr.src = nh->saddr; | |
277 | key->ipv6.addr.dst = nh->daddr; | |
278 | ||
279 | payload_ofs = ipv6_skip_exthdr(skb, payload_ofs, &nexthdr, &frag_off); | |
ccb1352e JG |
280 | |
281 | if (frag_off) { | |
282 | if (frag_off & htons(~0x7)) | |
283 | key->ip.frag = OVS_FRAG_TYPE_LATER; | |
284 | else | |
285 | key->ip.frag = OVS_FRAG_TYPE_FIRST; | |
25ef1328 PS |
286 | } else { |
287 | key->ip.frag = OVS_FRAG_TYPE_NONE; | |
ccb1352e JG |
288 | } |
289 | ||
c30da497 SH |
290 | /* Delayed handling of error in ipv6_skip_exthdr() as it |
291 | * always sets frag_off to a valid value which may be | |
292 | * used to set key->ip.frag above. | |
293 | */ | |
294 | if (unlikely(payload_ofs < 0)) | |
295 | return -EPROTO; | |
296 | ||
ccb1352e JG |
297 | nh_len = payload_ofs - nh_ofs; |
298 | skb_set_transport_header(skb, nh_ofs + nh_len); | |
299 | key->ip.proto = nexthdr; | |
300 | return nh_len; | |
301 | } | |
302 | ||
303 | static bool icmp6hdr_ok(struct sk_buff *skb) | |
304 | { | |
305 | return pskb_may_pull(skb, skb_transport_offset(skb) + | |
306 | sizeof(struct icmp6hdr)); | |
307 | } | |
308 | ||
018c1dda EG |
309 | /** |
310 | * Parse vlan tag from vlan header. | |
311 | * Returns ERROR on memory error. | |
312 | * Returns 0 if it encounters a non-vlan or incomplete packet. | |
313 | * Returns 1 after successfully parsing vlan tag. | |
314 | */ | |
df30f740 | 315 | static int parse_vlan_tag(struct sk_buff *skb, struct vlan_head *key_vh, |
316 | bool untag_vlan) | |
ccb1352e | 317 | { |
018c1dda | 318 | struct vlan_head *vh = (struct vlan_head *)skb->data; |
ccb1352e | 319 | |
018c1dda | 320 | if (likely(!eth_type_vlan(vh->tpid))) |
ccb1352e JG |
321 | return 0; |
322 | ||
018c1dda EG |
323 | if (unlikely(skb->len < sizeof(struct vlan_head) + sizeof(__be16))) |
324 | return 0; | |
325 | ||
326 | if (unlikely(!pskb_may_pull(skb, sizeof(struct vlan_head) + | |
327 | sizeof(__be16)))) | |
ccb1352e JG |
328 | return -ENOMEM; |
329 | ||
018c1dda EG |
330 | vh = (struct vlan_head *)skb->data; |
331 | key_vh->tci = vh->tci | htons(VLAN_TAG_PRESENT); | |
332 | key_vh->tpid = vh->tpid; | |
333 | ||
df30f740 | 334 | if (unlikely(untag_vlan)) { |
335 | int offset = skb->data - skb_mac_header(skb); | |
336 | u16 tci; | |
337 | int err; | |
338 | ||
339 | __skb_push(skb, offset); | |
340 | err = __skb_vlan_pop(skb, &tci); | |
341 | __skb_pull(skb, offset); | |
342 | if (err) | |
343 | return err; | |
344 | __vlan_hwaccel_put_tag(skb, key_vh->tpid, tci); | |
345 | } else { | |
346 | __skb_pull(skb, sizeof(struct vlan_head)); | |
347 | } | |
018c1dda EG |
348 | return 1; |
349 | } | |
350 | ||
5108bbad | 351 | static void clear_vlan(struct sw_flow_key *key) |
018c1dda | 352 | { |
018c1dda EG |
353 | key->eth.vlan.tci = 0; |
354 | key->eth.vlan.tpid = 0; | |
355 | key->eth.cvlan.tci = 0; | |
356 | key->eth.cvlan.tpid = 0; | |
5108bbad JB |
357 | } |
358 | ||
359 | static int parse_vlan(struct sk_buff *skb, struct sw_flow_key *key) | |
360 | { | |
361 | int res; | |
018c1dda | 362 | |
20ecf1e4 | 363 | if (skb_vlan_tag_present(skb)) { |
018c1dda EG |
364 | key->eth.vlan.tci = htons(skb->vlan_tci); |
365 | key->eth.vlan.tpid = skb->vlan_proto; | |
366 | } else { | |
367 | /* Parse outer vlan tag in the non-accelerated case. */ | |
df30f740 | 368 | res = parse_vlan_tag(skb, &key->eth.vlan, true); |
018c1dda EG |
369 | if (res <= 0) |
370 | return res; | |
371 | } | |
372 | ||
373 | /* Parse inner vlan tag. */ | |
df30f740 | 374 | res = parse_vlan_tag(skb, &key->eth.cvlan, false); |
018c1dda EG |
375 | if (res <= 0) |
376 | return res; | |
ccb1352e JG |
377 | |
378 | return 0; | |
379 | } | |
380 | ||
381 | static __be16 parse_ethertype(struct sk_buff *skb) | |
382 | { | |
383 | struct llc_snap_hdr { | |
384 | u8 dsap; /* Always 0xAA */ | |
385 | u8 ssap; /* Always 0xAA */ | |
386 | u8 ctrl; | |
387 | u8 oui[3]; | |
388 | __be16 ethertype; | |
389 | }; | |
390 | struct llc_snap_hdr *llc; | |
391 | __be16 proto; | |
392 | ||
393 | proto = *(__be16 *) skb->data; | |
394 | __skb_pull(skb, sizeof(__be16)); | |
395 | ||
6713fc9b | 396 | if (eth_proto_is_802_3(proto)) |
ccb1352e JG |
397 | return proto; |
398 | ||
399 | if (skb->len < sizeof(struct llc_snap_hdr)) | |
400 | return htons(ETH_P_802_2); | |
401 | ||
402 | if (unlikely(!pskb_may_pull(skb, sizeof(struct llc_snap_hdr)))) | |
403 | return htons(0); | |
404 | ||
405 | llc = (struct llc_snap_hdr *) skb->data; | |
406 | if (llc->dsap != LLC_SAP_SNAP || | |
407 | llc->ssap != LLC_SAP_SNAP || | |
408 | (llc->oui[0] | llc->oui[1] | llc->oui[2]) != 0) | |
409 | return htons(ETH_P_802_2); | |
410 | ||
411 | __skb_pull(skb, sizeof(struct llc_snap_hdr)); | |
17b682a0 | 412 | |
6713fc9b | 413 | if (eth_proto_is_802_3(llc->ethertype)) |
17b682a0 RL |
414 | return llc->ethertype; |
415 | ||
416 | return htons(ETH_P_802_2); | |
ccb1352e JG |
417 | } |
418 | ||
419 | static int parse_icmpv6(struct sk_buff *skb, struct sw_flow_key *key, | |
03f0d916 | 420 | int nh_len) |
ccb1352e JG |
421 | { |
422 | struct icmp6hdr *icmp = icmp6_hdr(skb); | |
ccb1352e JG |
423 | |
424 | /* The ICMPv6 type and code fields use the 16-bit transport port | |
425 | * fields, so we need to store them in 16-bit network byte order. | |
426 | */ | |
1139e241 JR |
427 | key->tp.src = htons(icmp->icmp6_type); |
428 | key->tp.dst = htons(icmp->icmp6_code); | |
25ef1328 | 429 | memset(&key->ipv6.nd, 0, sizeof(key->ipv6.nd)); |
ccb1352e JG |
430 | |
431 | if (icmp->icmp6_code == 0 && | |
432 | (icmp->icmp6_type == NDISC_NEIGHBOUR_SOLICITATION || | |
433 | icmp->icmp6_type == NDISC_NEIGHBOUR_ADVERTISEMENT)) { | |
434 | int icmp_len = skb->len - skb_transport_offset(skb); | |
435 | struct nd_msg *nd; | |
436 | int offset; | |
437 | ||
ccb1352e JG |
438 | /* In order to process neighbor discovery options, we need the |
439 | * entire packet. | |
440 | */ | |
441 | if (unlikely(icmp_len < sizeof(*nd))) | |
03f0d916 AZ |
442 | return 0; |
443 | ||
444 | if (unlikely(skb_linearize(skb))) | |
445 | return -ENOMEM; | |
ccb1352e JG |
446 | |
447 | nd = (struct nd_msg *)skb_transport_header(skb); | |
448 | key->ipv6.nd.target = nd->target; | |
ccb1352e JG |
449 | |
450 | icmp_len -= sizeof(*nd); | |
451 | offset = 0; | |
452 | while (icmp_len >= 8) { | |
453 | struct nd_opt_hdr *nd_opt = | |
454 | (struct nd_opt_hdr *)(nd->opt + offset); | |
455 | int opt_len = nd_opt->nd_opt_len * 8; | |
456 | ||
457 | if (unlikely(!opt_len || opt_len > icmp_len)) | |
03f0d916 | 458 | return 0; |
ccb1352e JG |
459 | |
460 | /* Store the link layer address if the appropriate | |
461 | * option is provided. It is considered an error if | |
462 | * the same link layer option is specified twice. | |
463 | */ | |
464 | if (nd_opt->nd_opt_type == ND_OPT_SOURCE_LL_ADDR | |
465 | && opt_len == 8) { | |
466 | if (unlikely(!is_zero_ether_addr(key->ipv6.nd.sll))) | |
467 | goto invalid; | |
8c63ff09 JP |
468 | ether_addr_copy(key->ipv6.nd.sll, |
469 | &nd->opt[offset+sizeof(*nd_opt)]); | |
ccb1352e JG |
470 | } else if (nd_opt->nd_opt_type == ND_OPT_TARGET_LL_ADDR |
471 | && opt_len == 8) { | |
472 | if (unlikely(!is_zero_ether_addr(key->ipv6.nd.tll))) | |
473 | goto invalid; | |
8c63ff09 JP |
474 | ether_addr_copy(key->ipv6.nd.tll, |
475 | &nd->opt[offset+sizeof(*nd_opt)]); | |
ccb1352e JG |
476 | } |
477 | ||
478 | icmp_len -= opt_len; | |
479 | offset += opt_len; | |
480 | } | |
481 | } | |
482 | ||
03f0d916 | 483 | return 0; |
ccb1352e JG |
484 | |
485 | invalid: | |
486 | memset(&key->ipv6.nd.target, 0, sizeof(key->ipv6.nd.target)); | |
487 | memset(key->ipv6.nd.sll, 0, sizeof(key->ipv6.nd.sll)); | |
488 | memset(key->ipv6.nd.tll, 0, sizeof(key->ipv6.nd.tll)); | |
489 | ||
03f0d916 | 490 | return 0; |
ccb1352e JG |
491 | } |
492 | ||
493 | /** | |
83c8df26 | 494 | * key_extract - extracts a flow key from an Ethernet frame. |
ccb1352e JG |
495 | * @skb: sk_buff that contains the frame, with skb->data pointing to the |
496 | * Ethernet header | |
ccb1352e | 497 | * @key: output flow key |
ccb1352e JG |
498 | * |
499 | * The caller must ensure that skb->len >= ETH_HLEN. | |
500 | * | |
501 | * Returns 0 if successful, otherwise a negative errno value. | |
502 | * | |
5108bbad | 503 | * Initializes @skb header fields as follows: |
ccb1352e | 504 | * |
5108bbad | 505 | * - skb->mac_header: the L2 header. |
ccb1352e | 506 | * |
5108bbad JB |
507 | * - skb->network_header: just past the L2 header, or just past the |
508 | * VLAN header, to the first byte of the L2 payload. | |
ccb1352e | 509 | * |
34d94f21 | 510 | * - skb->transport_header: If key->eth.type is ETH_P_IP or ETH_P_IPV6 |
ccb1352e JG |
511 | * on output, then just past the IP header, if one is present and |
512 | * of a correct length, otherwise the same as skb->network_header. | |
34d94f21 | 513 | * For other key->eth.type values it is left untouched. |
5108bbad JB |
514 | * |
515 | * - skb->protocol: the type of the data starting at skb->network_header. | |
516 | * Equals to key->eth.type. | |
ccb1352e | 517 | */ |
83c8df26 | 518 | static int key_extract(struct sk_buff *skb, struct sw_flow_key *key) |
ccb1352e | 519 | { |
03f0d916 | 520 | int error; |
ccb1352e JG |
521 | struct ethhdr *eth; |
522 | ||
f5796684 JG |
523 | /* Flags are always used as part of stats */ |
524 | key->tp.flags = 0; | |
525 | ||
ccb1352e JG |
526 | skb_reset_mac_header(skb); |
527 | ||
5108bbad JB |
528 | /* Link layer. */ |
529 | clear_vlan(key); | |
530 | if (key->mac_proto == MAC_PROTO_NONE) { | |
531 | if (unlikely(eth_type_vlan(skb->protocol))) | |
532 | return -EINVAL; | |
ccb1352e | 533 | |
5108bbad JB |
534 | skb_reset_network_header(skb); |
535 | } else { | |
536 | eth = eth_hdr(skb); | |
537 | ether_addr_copy(key->eth.src, eth->h_source); | |
538 | ether_addr_copy(key->eth.dst, eth->h_dest); | |
ccb1352e | 539 | |
5108bbad JB |
540 | __skb_pull(skb, 2 * ETH_ALEN); |
541 | /* We are going to push all headers that we pull, so no need to | |
542 | * update skb->csum here. | |
543 | */ | |
ccb1352e | 544 | |
5108bbad JB |
545 | if (unlikely(parse_vlan(skb, key))) |
546 | return -ENOMEM; | |
547 | ||
548 | skb->protocol = parse_ethertype(skb); | |
549 | if (unlikely(skb->protocol == htons(0))) | |
550 | return -ENOMEM; | |
ccb1352e | 551 | |
5108bbad JB |
552 | skb_reset_network_header(skb); |
553 | __skb_push(skb, skb->data - skb_mac_header(skb)); | |
554 | } | |
25cd9ba0 | 555 | skb_reset_mac_len(skb); |
5108bbad | 556 | key->eth.type = skb->protocol; |
ccb1352e JG |
557 | |
558 | /* Network layer. */ | |
559 | if (key->eth.type == htons(ETH_P_IP)) { | |
560 | struct iphdr *nh; | |
561 | __be16 offset; | |
562 | ||
ccb1352e JG |
563 | error = check_iphdr(skb); |
564 | if (unlikely(error)) { | |
07148121 JG |
565 | memset(&key->ip, 0, sizeof(key->ip)); |
566 | memset(&key->ipv4, 0, sizeof(key->ipv4)); | |
ccb1352e JG |
567 | if (error == -EINVAL) { |
568 | skb->transport_header = skb->network_header; | |
569 | error = 0; | |
570 | } | |
03f0d916 | 571 | return error; |
ccb1352e JG |
572 | } |
573 | ||
574 | nh = ip_hdr(skb); | |
575 | key->ipv4.addr.src = nh->saddr; | |
576 | key->ipv4.addr.dst = nh->daddr; | |
577 | ||
578 | key->ip.proto = nh->protocol; | |
579 | key->ip.tos = nh->tos; | |
580 | key->ip.ttl = nh->ttl; | |
581 | ||
582 | offset = nh->frag_off & htons(IP_OFFSET); | |
583 | if (offset) { | |
584 | key->ip.frag = OVS_FRAG_TYPE_LATER; | |
03f0d916 | 585 | return 0; |
ccb1352e JG |
586 | } |
587 | if (nh->frag_off & htons(IP_MF) || | |
07148121 | 588 | skb_shinfo(skb)->gso_type & SKB_GSO_UDP) |
ccb1352e | 589 | key->ip.frag = OVS_FRAG_TYPE_FIRST; |
07148121 JG |
590 | else |
591 | key->ip.frag = OVS_FRAG_TYPE_NONE; | |
ccb1352e JG |
592 | |
593 | /* Transport layer. */ | |
594 | if (key->ip.proto == IPPROTO_TCP) { | |
ccb1352e JG |
595 | if (tcphdr_ok(skb)) { |
596 | struct tcphdr *tcp = tcp_hdr(skb); | |
1139e241 JR |
597 | key->tp.src = tcp->source; |
598 | key->tp.dst = tcp->dest; | |
599 | key->tp.flags = TCP_FLAGS_BE16(tcp); | |
07148121 JG |
600 | } else { |
601 | memset(&key->tp, 0, sizeof(key->tp)); | |
ccb1352e | 602 | } |
07148121 | 603 | |
ccb1352e | 604 | } else if (key->ip.proto == IPPROTO_UDP) { |
ccb1352e JG |
605 | if (udphdr_ok(skb)) { |
606 | struct udphdr *udp = udp_hdr(skb); | |
1139e241 JR |
607 | key->tp.src = udp->source; |
608 | key->tp.dst = udp->dest; | |
07148121 JG |
609 | } else { |
610 | memset(&key->tp, 0, sizeof(key->tp)); | |
ccb1352e | 611 | } |
a175a723 JS |
612 | } else if (key->ip.proto == IPPROTO_SCTP) { |
613 | if (sctphdr_ok(skb)) { | |
614 | struct sctphdr *sctp = sctp_hdr(skb); | |
1139e241 JR |
615 | key->tp.src = sctp->source; |
616 | key->tp.dst = sctp->dest; | |
07148121 JG |
617 | } else { |
618 | memset(&key->tp, 0, sizeof(key->tp)); | |
a175a723 | 619 | } |
ccb1352e | 620 | } else if (key->ip.proto == IPPROTO_ICMP) { |
ccb1352e JG |
621 | if (icmphdr_ok(skb)) { |
622 | struct icmphdr *icmp = icmp_hdr(skb); | |
623 | /* The ICMP type and code fields use the 16-bit | |
624 | * transport port fields, so we need to store | |
625 | * them in 16-bit network byte order. */ | |
1139e241 JR |
626 | key->tp.src = htons(icmp->type); |
627 | key->tp.dst = htons(icmp->code); | |
07148121 JG |
628 | } else { |
629 | memset(&key->tp, 0, sizeof(key->tp)); | |
ccb1352e JG |
630 | } |
631 | } | |
632 | ||
07148121 JG |
633 | } else if (key->eth.type == htons(ETH_P_ARP) || |
634 | key->eth.type == htons(ETH_P_RARP)) { | |
ccb1352e | 635 | struct arp_eth_header *arp; |
389f4894 | 636 | bool arp_available = arphdr_ok(skb); |
ccb1352e JG |
637 | |
638 | arp = (struct arp_eth_header *)skb_network_header(skb); | |
639 | ||
389f4894 | 640 | if (arp_available && |
07148121 JG |
641 | arp->ar_hrd == htons(ARPHRD_ETHER) && |
642 | arp->ar_pro == htons(ETH_P_IP) && | |
643 | arp->ar_hln == ETH_ALEN && | |
644 | arp->ar_pln == 4) { | |
ccb1352e JG |
645 | |
646 | /* We only match on the lower 8 bits of the opcode. */ | |
647 | if (ntohs(arp->ar_op) <= 0xff) | |
648 | key->ip.proto = ntohs(arp->ar_op); | |
07148121 JG |
649 | else |
650 | key->ip.proto = 0; | |
651 | ||
d04d3829 MM |
652 | memcpy(&key->ipv4.addr.src, arp->ar_sip, sizeof(key->ipv4.addr.src)); |
653 | memcpy(&key->ipv4.addr.dst, arp->ar_tip, sizeof(key->ipv4.addr.dst)); | |
8c63ff09 JP |
654 | ether_addr_copy(key->ipv4.arp.sha, arp->ar_sha); |
655 | ether_addr_copy(key->ipv4.arp.tha, arp->ar_tha); | |
07148121 JG |
656 | } else { |
657 | memset(&key->ip, 0, sizeof(key->ip)); | |
658 | memset(&key->ipv4, 0, sizeof(key->ipv4)); | |
ccb1352e | 659 | } |
25cd9ba0 SH |
660 | } else if (eth_p_mpls(key->eth.type)) { |
661 | size_t stack_len = MPLS_HLEN; | |
662 | ||
f7d49bce | 663 | skb_set_inner_network_header(skb, skb->mac_len); |
25cd9ba0 SH |
664 | while (1) { |
665 | __be32 lse; | |
666 | ||
667 | error = check_header(skb, skb->mac_len + stack_len); | |
668 | if (unlikely(error)) | |
669 | return 0; | |
670 | ||
f7d49bce | 671 | memcpy(&lse, skb_inner_network_header(skb), MPLS_HLEN); |
25cd9ba0 SH |
672 | |
673 | if (stack_len == MPLS_HLEN) | |
674 | memcpy(&key->mpls.top_lse, &lse, MPLS_HLEN); | |
675 | ||
f7d49bce | 676 | skb_set_inner_network_header(skb, skb->mac_len + stack_len); |
25cd9ba0 SH |
677 | if (lse & htonl(MPLS_LS_S_MASK)) |
678 | break; | |
679 | ||
680 | stack_len += MPLS_HLEN; | |
681 | } | |
ccb1352e JG |
682 | } else if (key->eth.type == htons(ETH_P_IPV6)) { |
683 | int nh_len; /* IPv6 Header + Extensions */ | |
684 | ||
03f0d916 | 685 | nh_len = parse_ipv6hdr(skb, key); |
ccb1352e | 686 | if (unlikely(nh_len < 0)) { |
c30da497 SH |
687 | switch (nh_len) { |
688 | case -EINVAL: | |
689 | memset(&key->ip, 0, sizeof(key->ip)); | |
690 | memset(&key->ipv6.addr, 0, sizeof(key->ipv6.addr)); | |
691 | /* fall-through */ | |
692 | case -EPROTO: | |
ccb1352e | 693 | skb->transport_header = skb->network_header; |
03f0d916 | 694 | error = 0; |
c30da497 SH |
695 | break; |
696 | default: | |
ccb1352e | 697 | error = nh_len; |
03f0d916 AZ |
698 | } |
699 | return error; | |
ccb1352e JG |
700 | } |
701 | ||
702 | if (key->ip.frag == OVS_FRAG_TYPE_LATER) | |
03f0d916 | 703 | return 0; |
ccb1352e JG |
704 | if (skb_shinfo(skb)->gso_type & SKB_GSO_UDP) |
705 | key->ip.frag = OVS_FRAG_TYPE_FIRST; | |
706 | ||
707 | /* Transport layer. */ | |
708 | if (key->ip.proto == NEXTHDR_TCP) { | |
ccb1352e JG |
709 | if (tcphdr_ok(skb)) { |
710 | struct tcphdr *tcp = tcp_hdr(skb); | |
1139e241 JR |
711 | key->tp.src = tcp->source; |
712 | key->tp.dst = tcp->dest; | |
713 | key->tp.flags = TCP_FLAGS_BE16(tcp); | |
07148121 JG |
714 | } else { |
715 | memset(&key->tp, 0, sizeof(key->tp)); | |
ccb1352e JG |
716 | } |
717 | } else if (key->ip.proto == NEXTHDR_UDP) { | |
ccb1352e JG |
718 | if (udphdr_ok(skb)) { |
719 | struct udphdr *udp = udp_hdr(skb); | |
1139e241 JR |
720 | key->tp.src = udp->source; |
721 | key->tp.dst = udp->dest; | |
07148121 JG |
722 | } else { |
723 | memset(&key->tp, 0, sizeof(key->tp)); | |
ccb1352e | 724 | } |
a175a723 JS |
725 | } else if (key->ip.proto == NEXTHDR_SCTP) { |
726 | if (sctphdr_ok(skb)) { | |
727 | struct sctphdr *sctp = sctp_hdr(skb); | |
1139e241 JR |
728 | key->tp.src = sctp->source; |
729 | key->tp.dst = sctp->dest; | |
07148121 JG |
730 | } else { |
731 | memset(&key->tp, 0, sizeof(key->tp)); | |
a175a723 | 732 | } |
ccb1352e | 733 | } else if (key->ip.proto == NEXTHDR_ICMP) { |
ccb1352e | 734 | if (icmp6hdr_ok(skb)) { |
03f0d916 AZ |
735 | error = parse_icmpv6(skb, key, nh_len); |
736 | if (error) | |
737 | return error; | |
07148121 JG |
738 | } else { |
739 | memset(&key->tp, 0, sizeof(key->tp)); | |
ccb1352e JG |
740 | } |
741 | } | |
742 | } | |
03f0d916 | 743 | return 0; |
ccb1352e | 744 | } |
83c8df26 | 745 | |
971427f3 AZ |
746 | int ovs_flow_key_update(struct sk_buff *skb, struct sw_flow_key *key) |
747 | { | |
748 | return key_extract(skb, key); | |
749 | } | |
750 | ||
5108bbad JB |
751 | static int key_extract_mac_proto(struct sk_buff *skb) |
752 | { | |
753 | switch (skb->dev->type) { | |
754 | case ARPHRD_ETHER: | |
755 | return MAC_PROTO_ETHERNET; | |
756 | case ARPHRD_NONE: | |
757 | if (skb->protocol == htons(ETH_P_TEB)) | |
758 | return MAC_PROTO_ETHERNET; | |
759 | return MAC_PROTO_NONE; | |
760 | } | |
761 | WARN_ON_ONCE(1); | |
762 | return -EINVAL; | |
763 | } | |
764 | ||
1d8fff90 | 765 | int ovs_flow_key_extract(const struct ip_tunnel_info *tun_info, |
8c8b1b83 | 766 | struct sk_buff *skb, struct sw_flow_key *key) |
83c8df26 | 767 | { |
9dd7f890 | 768 | int res, err; |
5108bbad | 769 | |
83c8df26 | 770 | /* Extract metadata from packet. */ |
f5796684 | 771 | if (tun_info) { |
00a93bab | 772 | key->tun_proto = ip_tunnel_info_af(tun_info); |
1d8fff90 | 773 | memcpy(&key->tun_key, &tun_info->key, sizeof(key->tun_key)); |
f5796684 | 774 | |
4c222798 | 775 | if (tun_info->options_len) { |
f5796684 JG |
776 | BUILD_BUG_ON((1 << (sizeof(tun_info->options_len) * |
777 | 8)) - 1 | |
778 | > sizeof(key->tun_opts)); | |
4c222798 PS |
779 | |
780 | ip_tunnel_info_opts_get(TUN_METADATA_OPTS(key, tun_info->options_len), | |
781 | tun_info); | |
f5796684 JG |
782 | key->tun_opts_len = tun_info->options_len; |
783 | } else { | |
784 | key->tun_opts_len = 0; | |
785 | } | |
786 | } else { | |
00a93bab | 787 | key->tun_proto = 0; |
f5796684 | 788 | key->tun_opts_len = 0; |
07148121 | 789 | memset(&key->tun_key, 0, sizeof(key->tun_key)); |
f5796684 | 790 | } |
83c8df26 PS |
791 | |
792 | key->phy.priority = skb->priority; | |
793 | key->phy.in_port = OVS_CB(skb)->input_vport->port_no; | |
794 | key->phy.skb_mark = skb->mark; | |
07148121 | 795 | key->ovs_flow_hash = 0; |
5108bbad JB |
796 | res = key_extract_mac_proto(skb); |
797 | if (res < 0) | |
798 | return res; | |
799 | key->mac_proto = res; | |
07148121 JG |
800 | key->recirc_id = 0; |
801 | ||
9dd7f890 JR |
802 | err = key_extract(skb, key); |
803 | if (!err) | |
804 | ovs_ct_fill_key(skb, key); /* Must be after key_extract(). */ | |
805 | return err; | |
83c8df26 PS |
806 | } |
807 | ||
c2ac6673 | 808 | int ovs_flow_key_extract_userspace(struct net *net, const struct nlattr *attr, |
83c8df26 | 809 | struct sk_buff *skb, |
05da5898 | 810 | struct sw_flow_key *key, bool log) |
83c8df26 | 811 | { |
9dd7f890 JR |
812 | const struct nlattr *a[OVS_KEY_ATTR_MAX + 1]; |
813 | u64 attrs = 0; | |
83c8df26 PS |
814 | int err; |
815 | ||
9dd7f890 JR |
816 | err = parse_flow_nlattrs(attr, a, &attrs, log); |
817 | if (err) | |
818 | return -EINVAL; | |
819 | ||
83c8df26 | 820 | /* Extract metadata from netlink attributes. */ |
9dd7f890 | 821 | err = ovs_nla_get_flow_metadata(net, a, attrs, key, log); |
83c8df26 PS |
822 | if (err) |
823 | return err; | |
824 | ||
df30f740 | 825 | /* key_extract assumes that skb->protocol is set-up for |
826 | * layer 3 packets which is the case for other callers, | |
827 | * in particular packets received from the network stack. | |
828 | * Here the correct value can be set from the metadata | |
829 | * extracted above. | |
830 | * For L2 packet key eth type would be zero. skb protocol | |
831 | * would be set to correct value later during key-extact. | |
832 | */ | |
5108bbad | 833 | |
df30f740 | 834 | skb->protocol = key->eth.type; |
9dd7f890 JR |
835 | err = key_extract(skb, key); |
836 | if (err) | |
837 | return err; | |
838 | ||
839 | /* Check that we have conntrack original direction tuple metadata only | |
840 | * for packets for which it makes sense. Otherwise the key may be | |
841 | * corrupted due to overlapping key fields. | |
842 | */ | |
843 | if (attrs & (1 << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV4) && | |
844 | key->eth.type != htons(ETH_P_IP)) | |
845 | return -EINVAL; | |
846 | if (attrs & (1 << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6) && | |
847 | (key->eth.type != htons(ETH_P_IPV6) || | |
848 | sw_flow_key_is_nd(key))) | |
849 | return -EINVAL; | |
850 | ||
851 | return 0; | |
83c8df26 | 852 | } |