]>
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 | 48 | #include <net/ndisc.h> |
b2d0f5d5 | 49 | #include <net/nsh.h> |
ccb1352e | 50 | |
a581b96d | 51 | #include "conntrack.h" |
83c8df26 PS |
52 | #include "datapath.h" |
53 | #include "flow.h" | |
54 | #include "flow_netlink.h" | |
a581b96d | 55 | #include "vport.h" |
83c8df26 | 56 | |
e6445719 | 57 | u64 ovs_flow_used_time(unsigned long flow_jiffies) |
03f0d916 | 58 | { |
e6445719 PS |
59 | struct timespec cur_ts; |
60 | u64 cur_ms, idle_ms; | |
03f0d916 | 61 | |
e6445719 PS |
62 | ktime_get_ts(&cur_ts); |
63 | idle_ms = jiffies_to_msecs(jiffies - flow_jiffies); | |
64 | cur_ms = (u64)cur_ts.tv_sec * MSEC_PER_SEC + | |
65 | cur_ts.tv_nsec / NSEC_PER_MSEC; | |
03f0d916 | 66 | |
e6445719 | 67 | return cur_ms - idle_ms; |
5828cd9a AZ |
68 | } |
69 | ||
df23e9f6 | 70 | #define TCP_FLAGS_BE16(tp) (*(__be16 *)&tcp_flag_word(tp) & htons(0x0FFF)) |
03f0d916 | 71 | |
ad552007 | 72 | void ovs_flow_stats_update(struct sw_flow *flow, __be16 tcp_flags, |
12eb18f7 | 73 | const struct sk_buff *skb) |
03f0d916 | 74 | { |
e298e505 | 75 | struct flow_stats *stats; |
c4b2bf6b | 76 | unsigned 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 | 110 | __GFP_NOMEMALLOC, |
c57c054e | 111 | numa_node_id()); |
63e7959c JR |
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 | 120 | new_stats); |
c4b2bf6b | 121 | cpumask_set_cpu(cpu, &flow->cpu_used_mask); |
63e7959c JR |
122 | goto unlock; |
123 | } | |
124 | } | |
db74a333 | 125 | flow->stats_last_writer = cpu; |
63e7959c JR |
126 | } |
127 | } | |
128 | ||
e298e505 PS |
129 | stats->used = jiffies; |
130 | stats->packet_count++; | |
24cc59d1 | 131 | stats->byte_count += len; |
e298e505 | 132 | stats->tcp_flags |= tcp_flags; |
63e7959c | 133 | unlock: |
e298e505 PS |
134 | spin_unlock(&stats->lock); |
135 | } | |
136 | ||
86ec8dba JR |
137 | /* Must be called with rcu_read_lock or ovs_mutex. */ |
138 | void ovs_flow_stats_get(const struct sw_flow *flow, | |
139 | struct ovs_flow_stats *ovs_stats, | |
e298e505 PS |
140 | unsigned long *used, __be16 *tcp_flags) |
141 | { | |
db74a333 | 142 | int cpu; |
e298e505 PS |
143 | |
144 | *used = 0; | |
145 | *tcp_flags = 0; | |
146 | memset(ovs_stats, 0, sizeof(*ovs_stats)); | |
147 | ||
db74a333 | 148 | /* We open code this to make sure cpu 0 is always considered */ |
c4b2bf6b | 149 | for (cpu = 0; cpu < nr_cpu_ids; cpu = cpumask_next(cpu, &flow->cpu_used_mask)) { |
db74a333 | 150 | struct flow_stats *stats = rcu_dereference_ovsl(flow->stats[cpu]); |
63e7959c JR |
151 | |
152 | if (stats) { | |
153 | /* Local CPU may write on non-local stats, so we must | |
154 | * block bottom-halves here. | |
155 | */ | |
156 | spin_lock_bh(&stats->lock); | |
157 | if (!*used || time_after(stats->used, *used)) | |
158 | *used = stats->used; | |
159 | *tcp_flags |= stats->tcp_flags; | |
160 | ovs_stats->n_packets += stats->packet_count; | |
161 | ovs_stats->n_bytes += stats->byte_count; | |
162 | spin_unlock_bh(&stats->lock); | |
163 | } | |
e298e505 | 164 | } |
e298e505 PS |
165 | } |
166 | ||
86ec8dba | 167 | /* Called with ovs_mutex. */ |
e298e505 PS |
168 | void ovs_flow_stats_clear(struct sw_flow *flow) |
169 | { | |
db74a333 | 170 | int cpu; |
63e7959c | 171 | |
db74a333 | 172 | /* We open code this to make sure cpu 0 is always considered */ |
c4b2bf6b | 173 | for (cpu = 0; cpu < nr_cpu_ids; cpu = cpumask_next(cpu, &flow->cpu_used_mask)) { |
db74a333 | 174 | struct flow_stats *stats = ovsl_dereference(flow->stats[cpu]); |
63e7959c JR |
175 | |
176 | if (stats) { | |
177 | spin_lock_bh(&stats->lock); | |
178 | stats->used = 0; | |
179 | stats->packet_count = 0; | |
180 | stats->byte_count = 0; | |
181 | stats->tcp_flags = 0; | |
182 | spin_unlock_bh(&stats->lock); | |
183 | } | |
184 | } | |
03f0d916 AZ |
185 | } |
186 | ||
ccb1352e JG |
187 | static int check_header(struct sk_buff *skb, int len) |
188 | { | |
189 | if (unlikely(skb->len < len)) | |
190 | return -EINVAL; | |
191 | if (unlikely(!pskb_may_pull(skb, len))) | |
192 | return -ENOMEM; | |
193 | return 0; | |
194 | } | |
195 | ||
196 | static bool arphdr_ok(struct sk_buff *skb) | |
197 | { | |
198 | return pskb_may_pull(skb, skb_network_offset(skb) + | |
199 | sizeof(struct arp_eth_header)); | |
200 | } | |
201 | ||
202 | static int check_iphdr(struct sk_buff *skb) | |
203 | { | |
204 | unsigned int nh_ofs = skb_network_offset(skb); | |
205 | unsigned int ip_len; | |
206 | int err; | |
207 | ||
208 | err = check_header(skb, nh_ofs + sizeof(struct iphdr)); | |
209 | if (unlikely(err)) | |
210 | return err; | |
211 | ||
212 | ip_len = ip_hdrlen(skb); | |
213 | if (unlikely(ip_len < sizeof(struct iphdr) || | |
214 | skb->len < nh_ofs + ip_len)) | |
215 | return -EINVAL; | |
216 | ||
217 | skb_set_transport_header(skb, nh_ofs + ip_len); | |
218 | return 0; | |
219 | } | |
220 | ||
221 | static bool tcphdr_ok(struct sk_buff *skb) | |
222 | { | |
223 | int th_ofs = skb_transport_offset(skb); | |
224 | int tcp_len; | |
225 | ||
226 | if (unlikely(!pskb_may_pull(skb, th_ofs + sizeof(struct tcphdr)))) | |
227 | return false; | |
228 | ||
229 | tcp_len = tcp_hdrlen(skb); | |
230 | if (unlikely(tcp_len < sizeof(struct tcphdr) || | |
231 | skb->len < th_ofs + tcp_len)) | |
232 | return false; | |
233 | ||
234 | return true; | |
235 | } | |
236 | ||
237 | static bool udphdr_ok(struct sk_buff *skb) | |
238 | { | |
239 | return pskb_may_pull(skb, skb_transport_offset(skb) + | |
240 | sizeof(struct udphdr)); | |
241 | } | |
242 | ||
a175a723 JS |
243 | static bool sctphdr_ok(struct sk_buff *skb) |
244 | { | |
245 | return pskb_may_pull(skb, skb_transport_offset(skb) + | |
246 | sizeof(struct sctphdr)); | |
247 | } | |
248 | ||
ccb1352e JG |
249 | static bool icmphdr_ok(struct sk_buff *skb) |
250 | { | |
251 | return pskb_may_pull(skb, skb_transport_offset(skb) + | |
252 | sizeof(struct icmphdr)); | |
253 | } | |
254 | ||
03f0d916 | 255 | static int parse_ipv6hdr(struct sk_buff *skb, struct sw_flow_key *key) |
ccb1352e JG |
256 | { |
257 | unsigned int nh_ofs = skb_network_offset(skb); | |
258 | unsigned int nh_len; | |
259 | int payload_ofs; | |
260 | struct ipv6hdr *nh; | |
261 | uint8_t nexthdr; | |
262 | __be16 frag_off; | |
263 | int err; | |
264 | ||
ccb1352e JG |
265 | err = check_header(skb, nh_ofs + sizeof(*nh)); |
266 | if (unlikely(err)) | |
267 | return err; | |
268 | ||
269 | nh = ipv6_hdr(skb); | |
270 | nexthdr = nh->nexthdr; | |
271 | payload_ofs = (u8 *)(nh + 1) - skb->data; | |
272 | ||
273 | key->ip.proto = NEXTHDR_NONE; | |
274 | key->ip.tos = ipv6_get_dsfield(nh); | |
275 | key->ip.ttl = nh->hop_limit; | |
276 | key->ipv6.label = *(__be32 *)nh & htonl(IPV6_FLOWINFO_FLOWLABEL); | |
277 | key->ipv6.addr.src = nh->saddr; | |
278 | key->ipv6.addr.dst = nh->daddr; | |
279 | ||
280 | payload_ofs = ipv6_skip_exthdr(skb, payload_ofs, &nexthdr, &frag_off); | |
ccb1352e JG |
281 | |
282 | if (frag_off) { | |
283 | if (frag_off & htons(~0x7)) | |
284 | key->ip.frag = OVS_FRAG_TYPE_LATER; | |
285 | else | |
286 | key->ip.frag = OVS_FRAG_TYPE_FIRST; | |
25ef1328 PS |
287 | } else { |
288 | key->ip.frag = OVS_FRAG_TYPE_NONE; | |
ccb1352e JG |
289 | } |
290 | ||
c30da497 SH |
291 | /* Delayed handling of error in ipv6_skip_exthdr() as it |
292 | * always sets frag_off to a valid value which may be | |
293 | * used to set key->ip.frag above. | |
294 | */ | |
295 | if (unlikely(payload_ofs < 0)) | |
296 | return -EPROTO; | |
297 | ||
ccb1352e JG |
298 | nh_len = payload_ofs - nh_ofs; |
299 | skb_set_transport_header(skb, nh_ofs + nh_len); | |
300 | key->ip.proto = nexthdr; | |
301 | return nh_len; | |
302 | } | |
303 | ||
304 | static bool icmp6hdr_ok(struct sk_buff *skb) | |
305 | { | |
306 | return pskb_may_pull(skb, skb_transport_offset(skb) + | |
307 | sizeof(struct icmp6hdr)); | |
308 | } | |
309 | ||
018c1dda EG |
310 | /** |
311 | * Parse vlan tag from vlan header. | |
312 | * Returns ERROR on memory error. | |
313 | * Returns 0 if it encounters a non-vlan or incomplete packet. | |
314 | * Returns 1 after successfully parsing vlan tag. | |
315 | */ | |
df30f740 | 316 | static int parse_vlan_tag(struct sk_buff *skb, struct vlan_head *key_vh, |
317 | bool untag_vlan) | |
ccb1352e | 318 | { |
018c1dda | 319 | struct vlan_head *vh = (struct vlan_head *)skb->data; |
ccb1352e | 320 | |
018c1dda | 321 | if (likely(!eth_type_vlan(vh->tpid))) |
ccb1352e JG |
322 | return 0; |
323 | ||
018c1dda EG |
324 | if (unlikely(skb->len < sizeof(struct vlan_head) + sizeof(__be16))) |
325 | return 0; | |
326 | ||
327 | if (unlikely(!pskb_may_pull(skb, sizeof(struct vlan_head) + | |
328 | sizeof(__be16)))) | |
ccb1352e JG |
329 | return -ENOMEM; |
330 | ||
018c1dda EG |
331 | vh = (struct vlan_head *)skb->data; |
332 | key_vh->tci = vh->tci | htons(VLAN_TAG_PRESENT); | |
333 | key_vh->tpid = vh->tpid; | |
334 | ||
df30f740 | 335 | if (unlikely(untag_vlan)) { |
336 | int offset = skb->data - skb_mac_header(skb); | |
337 | u16 tci; | |
338 | int err; | |
339 | ||
340 | __skb_push(skb, offset); | |
341 | err = __skb_vlan_pop(skb, &tci); | |
342 | __skb_pull(skb, offset); | |
343 | if (err) | |
344 | return err; | |
345 | __vlan_hwaccel_put_tag(skb, key_vh->tpid, tci); | |
346 | } else { | |
347 | __skb_pull(skb, sizeof(struct vlan_head)); | |
348 | } | |
018c1dda EG |
349 | return 1; |
350 | } | |
351 | ||
5108bbad | 352 | static void clear_vlan(struct sw_flow_key *key) |
018c1dda | 353 | { |
018c1dda EG |
354 | key->eth.vlan.tci = 0; |
355 | key->eth.vlan.tpid = 0; | |
356 | key->eth.cvlan.tci = 0; | |
357 | key->eth.cvlan.tpid = 0; | |
5108bbad JB |
358 | } |
359 | ||
360 | static int parse_vlan(struct sk_buff *skb, struct sw_flow_key *key) | |
361 | { | |
362 | int res; | |
018c1dda | 363 | |
20ecf1e4 | 364 | if (skb_vlan_tag_present(skb)) { |
018c1dda EG |
365 | key->eth.vlan.tci = htons(skb->vlan_tci); |
366 | key->eth.vlan.tpid = skb->vlan_proto; | |
367 | } else { | |
368 | /* Parse outer vlan tag in the non-accelerated case. */ | |
df30f740 | 369 | res = parse_vlan_tag(skb, &key->eth.vlan, true); |
018c1dda EG |
370 | if (res <= 0) |
371 | return res; | |
372 | } | |
373 | ||
374 | /* Parse inner vlan tag. */ | |
df30f740 | 375 | res = parse_vlan_tag(skb, &key->eth.cvlan, false); |
018c1dda EG |
376 | if (res <= 0) |
377 | return res; | |
ccb1352e JG |
378 | |
379 | return 0; | |
380 | } | |
381 | ||
382 | static __be16 parse_ethertype(struct sk_buff *skb) | |
383 | { | |
384 | struct llc_snap_hdr { | |
385 | u8 dsap; /* Always 0xAA */ | |
386 | u8 ssap; /* Always 0xAA */ | |
387 | u8 ctrl; | |
388 | u8 oui[3]; | |
389 | __be16 ethertype; | |
390 | }; | |
391 | struct llc_snap_hdr *llc; | |
392 | __be16 proto; | |
393 | ||
394 | proto = *(__be16 *) skb->data; | |
395 | __skb_pull(skb, sizeof(__be16)); | |
396 | ||
6713fc9b | 397 | if (eth_proto_is_802_3(proto)) |
ccb1352e JG |
398 | return proto; |
399 | ||
400 | if (skb->len < sizeof(struct llc_snap_hdr)) | |
401 | return htons(ETH_P_802_2); | |
402 | ||
403 | if (unlikely(!pskb_may_pull(skb, sizeof(struct llc_snap_hdr)))) | |
404 | return htons(0); | |
405 | ||
406 | llc = (struct llc_snap_hdr *) skb->data; | |
407 | if (llc->dsap != LLC_SAP_SNAP || | |
408 | llc->ssap != LLC_SAP_SNAP || | |
409 | (llc->oui[0] | llc->oui[1] | llc->oui[2]) != 0) | |
410 | return htons(ETH_P_802_2); | |
411 | ||
412 | __skb_pull(skb, sizeof(struct llc_snap_hdr)); | |
17b682a0 | 413 | |
6713fc9b | 414 | if (eth_proto_is_802_3(llc->ethertype)) |
17b682a0 RL |
415 | return llc->ethertype; |
416 | ||
417 | return htons(ETH_P_802_2); | |
ccb1352e JG |
418 | } |
419 | ||
420 | static int parse_icmpv6(struct sk_buff *skb, struct sw_flow_key *key, | |
03f0d916 | 421 | int nh_len) |
ccb1352e JG |
422 | { |
423 | struct icmp6hdr *icmp = icmp6_hdr(skb); | |
ccb1352e JG |
424 | |
425 | /* The ICMPv6 type and code fields use the 16-bit transport port | |
426 | * fields, so we need to store them in 16-bit network byte order. | |
427 | */ | |
1139e241 JR |
428 | key->tp.src = htons(icmp->icmp6_type); |
429 | key->tp.dst = htons(icmp->icmp6_code); | |
25ef1328 | 430 | memset(&key->ipv6.nd, 0, sizeof(key->ipv6.nd)); |
ccb1352e JG |
431 | |
432 | if (icmp->icmp6_code == 0 && | |
433 | (icmp->icmp6_type == NDISC_NEIGHBOUR_SOLICITATION || | |
434 | icmp->icmp6_type == NDISC_NEIGHBOUR_ADVERTISEMENT)) { | |
435 | int icmp_len = skb->len - skb_transport_offset(skb); | |
436 | struct nd_msg *nd; | |
437 | int offset; | |
438 | ||
ccb1352e JG |
439 | /* In order to process neighbor discovery options, we need the |
440 | * entire packet. | |
441 | */ | |
442 | if (unlikely(icmp_len < sizeof(*nd))) | |
03f0d916 AZ |
443 | return 0; |
444 | ||
445 | if (unlikely(skb_linearize(skb))) | |
446 | return -ENOMEM; | |
ccb1352e JG |
447 | |
448 | nd = (struct nd_msg *)skb_transport_header(skb); | |
449 | key->ipv6.nd.target = nd->target; | |
ccb1352e JG |
450 | |
451 | icmp_len -= sizeof(*nd); | |
452 | offset = 0; | |
453 | while (icmp_len >= 8) { | |
454 | struct nd_opt_hdr *nd_opt = | |
455 | (struct nd_opt_hdr *)(nd->opt + offset); | |
456 | int opt_len = nd_opt->nd_opt_len * 8; | |
457 | ||
458 | if (unlikely(!opt_len || opt_len > icmp_len)) | |
03f0d916 | 459 | return 0; |
ccb1352e JG |
460 | |
461 | /* Store the link layer address if the appropriate | |
462 | * option is provided. It is considered an error if | |
463 | * the same link layer option is specified twice. | |
464 | */ | |
465 | if (nd_opt->nd_opt_type == ND_OPT_SOURCE_LL_ADDR | |
466 | && opt_len == 8) { | |
467 | if (unlikely(!is_zero_ether_addr(key->ipv6.nd.sll))) | |
468 | goto invalid; | |
8c63ff09 JP |
469 | ether_addr_copy(key->ipv6.nd.sll, |
470 | &nd->opt[offset+sizeof(*nd_opt)]); | |
ccb1352e JG |
471 | } else if (nd_opt->nd_opt_type == ND_OPT_TARGET_LL_ADDR |
472 | && opt_len == 8) { | |
473 | if (unlikely(!is_zero_ether_addr(key->ipv6.nd.tll))) | |
474 | goto invalid; | |
8c63ff09 JP |
475 | ether_addr_copy(key->ipv6.nd.tll, |
476 | &nd->opt[offset+sizeof(*nd_opt)]); | |
ccb1352e JG |
477 | } |
478 | ||
479 | icmp_len -= opt_len; | |
480 | offset += opt_len; | |
481 | } | |
482 | } | |
483 | ||
03f0d916 | 484 | return 0; |
ccb1352e JG |
485 | |
486 | invalid: | |
487 | memset(&key->ipv6.nd.target, 0, sizeof(key->ipv6.nd.target)); | |
488 | memset(key->ipv6.nd.sll, 0, sizeof(key->ipv6.nd.sll)); | |
489 | memset(key->ipv6.nd.tll, 0, sizeof(key->ipv6.nd.tll)); | |
490 | ||
03f0d916 | 491 | return 0; |
ccb1352e JG |
492 | } |
493 | ||
b2d0f5d5 YY |
494 | static int parse_nsh(struct sk_buff *skb, struct sw_flow_key *key) |
495 | { | |
496 | struct nshhdr *nh; | |
497 | unsigned int nh_ofs = skb_network_offset(skb); | |
498 | u8 version, length; | |
499 | int err; | |
500 | ||
501 | err = check_header(skb, nh_ofs + NSH_BASE_HDR_LEN); | |
502 | if (unlikely(err)) | |
503 | return err; | |
504 | ||
505 | nh = nsh_hdr(skb); | |
506 | version = nsh_get_ver(nh); | |
507 | length = nsh_hdr_len(nh); | |
508 | ||
509 | if (version != 0) | |
510 | return -EINVAL; | |
511 | ||
512 | err = check_header(skb, nh_ofs + length); | |
513 | if (unlikely(err)) | |
514 | return err; | |
515 | ||
516 | nh = nsh_hdr(skb); | |
517 | key->nsh.base.flags = nsh_get_flags(nh); | |
518 | key->nsh.base.ttl = nsh_get_ttl(nh); | |
519 | key->nsh.base.mdtype = nh->mdtype; | |
520 | key->nsh.base.np = nh->np; | |
521 | key->nsh.base.path_hdr = nh->path_hdr; | |
522 | switch (key->nsh.base.mdtype) { | |
523 | case NSH_M_TYPE1: | |
524 | if (length != NSH_M_TYPE1_LEN) | |
525 | return -EINVAL; | |
526 | memcpy(key->nsh.context, nh->md1.context, | |
527 | sizeof(nh->md1)); | |
528 | break; | |
529 | case NSH_M_TYPE2: | |
530 | memset(key->nsh.context, 0, | |
531 | sizeof(nh->md1)); | |
532 | break; | |
533 | default: | |
534 | return -EINVAL; | |
535 | } | |
536 | ||
537 | return 0; | |
538 | } | |
539 | ||
ccb1352e | 540 | /** |
83c8df26 | 541 | * key_extract - extracts a flow key from an Ethernet frame. |
ccb1352e JG |
542 | * @skb: sk_buff that contains the frame, with skb->data pointing to the |
543 | * Ethernet header | |
ccb1352e | 544 | * @key: output flow key |
ccb1352e JG |
545 | * |
546 | * The caller must ensure that skb->len >= ETH_HLEN. | |
547 | * | |
548 | * Returns 0 if successful, otherwise a negative errno value. | |
549 | * | |
5108bbad | 550 | * Initializes @skb header fields as follows: |
ccb1352e | 551 | * |
5108bbad | 552 | * - skb->mac_header: the L2 header. |
ccb1352e | 553 | * |
5108bbad JB |
554 | * - skb->network_header: just past the L2 header, or just past the |
555 | * VLAN header, to the first byte of the L2 payload. | |
ccb1352e | 556 | * |
34d94f21 | 557 | * - skb->transport_header: If key->eth.type is ETH_P_IP or ETH_P_IPV6 |
ccb1352e JG |
558 | * on output, then just past the IP header, if one is present and |
559 | * of a correct length, otherwise the same as skb->network_header. | |
34d94f21 | 560 | * For other key->eth.type values it is left untouched. |
5108bbad JB |
561 | * |
562 | * - skb->protocol: the type of the data starting at skb->network_header. | |
563 | * Equals to key->eth.type. | |
ccb1352e | 564 | */ |
83c8df26 | 565 | static int key_extract(struct sk_buff *skb, struct sw_flow_key *key) |
ccb1352e | 566 | { |
03f0d916 | 567 | int error; |
ccb1352e JG |
568 | struct ethhdr *eth; |
569 | ||
f5796684 JG |
570 | /* Flags are always used as part of stats */ |
571 | key->tp.flags = 0; | |
572 | ||
ccb1352e JG |
573 | skb_reset_mac_header(skb); |
574 | ||
5108bbad JB |
575 | /* Link layer. */ |
576 | clear_vlan(key); | |
6f56f618 | 577 | if (ovs_key_mac_proto(key) == MAC_PROTO_NONE) { |
5108bbad JB |
578 | if (unlikely(eth_type_vlan(skb->protocol))) |
579 | return -EINVAL; | |
ccb1352e | 580 | |
5108bbad | 581 | skb_reset_network_header(skb); |
c48e7473 | 582 | key->eth.type = skb->protocol; |
5108bbad JB |
583 | } else { |
584 | eth = eth_hdr(skb); | |
585 | ether_addr_copy(key->eth.src, eth->h_source); | |
586 | ether_addr_copy(key->eth.dst, eth->h_dest); | |
ccb1352e | 587 | |
5108bbad JB |
588 | __skb_pull(skb, 2 * ETH_ALEN); |
589 | /* We are going to push all headers that we pull, so no need to | |
590 | * update skb->csum here. | |
591 | */ | |
ccb1352e | 592 | |
5108bbad JB |
593 | if (unlikely(parse_vlan(skb, key))) |
594 | return -ENOMEM; | |
595 | ||
c48e7473 EG |
596 | key->eth.type = parse_ethertype(skb); |
597 | if (unlikely(key->eth.type == htons(0))) | |
5108bbad | 598 | return -ENOMEM; |
ccb1352e | 599 | |
c48e7473 EG |
600 | /* Multiple tagged packets need to retain TPID to satisfy |
601 | * skb_vlan_pop(), which will later shift the ethertype into | |
602 | * skb->protocol. | |
603 | */ | |
604 | if (key->eth.cvlan.tci & htons(VLAN_TAG_PRESENT)) | |
605 | skb->protocol = key->eth.cvlan.tpid; | |
606 | else | |
607 | skb->protocol = key->eth.type; | |
608 | ||
5108bbad JB |
609 | skb_reset_network_header(skb); |
610 | __skb_push(skb, skb->data - skb_mac_header(skb)); | |
611 | } | |
25cd9ba0 | 612 | skb_reset_mac_len(skb); |
ccb1352e JG |
613 | |
614 | /* Network layer. */ | |
615 | if (key->eth.type == htons(ETH_P_IP)) { | |
616 | struct iphdr *nh; | |
617 | __be16 offset; | |
618 | ||
ccb1352e JG |
619 | error = check_iphdr(skb); |
620 | if (unlikely(error)) { | |
07148121 JG |
621 | memset(&key->ip, 0, sizeof(key->ip)); |
622 | memset(&key->ipv4, 0, sizeof(key->ipv4)); | |
ccb1352e JG |
623 | if (error == -EINVAL) { |
624 | skb->transport_header = skb->network_header; | |
625 | error = 0; | |
626 | } | |
03f0d916 | 627 | return error; |
ccb1352e JG |
628 | } |
629 | ||
630 | nh = ip_hdr(skb); | |
631 | key->ipv4.addr.src = nh->saddr; | |
632 | key->ipv4.addr.dst = nh->daddr; | |
633 | ||
634 | key->ip.proto = nh->protocol; | |
635 | key->ip.tos = nh->tos; | |
636 | key->ip.ttl = nh->ttl; | |
637 | ||
638 | offset = nh->frag_off & htons(IP_OFFSET); | |
639 | if (offset) { | |
640 | key->ip.frag = OVS_FRAG_TYPE_LATER; | |
03f0d916 | 641 | return 0; |
ccb1352e | 642 | } |
0c19f846 WB |
643 | if (nh->frag_off & htons(IP_MF) || |
644 | skb_shinfo(skb)->gso_type & SKB_GSO_UDP) | |
ccb1352e | 645 | key->ip.frag = OVS_FRAG_TYPE_FIRST; |
07148121 JG |
646 | else |
647 | key->ip.frag = OVS_FRAG_TYPE_NONE; | |
ccb1352e JG |
648 | |
649 | /* Transport layer. */ | |
650 | if (key->ip.proto == IPPROTO_TCP) { | |
ccb1352e JG |
651 | if (tcphdr_ok(skb)) { |
652 | struct tcphdr *tcp = tcp_hdr(skb); | |
1139e241 JR |
653 | key->tp.src = tcp->source; |
654 | key->tp.dst = tcp->dest; | |
655 | key->tp.flags = TCP_FLAGS_BE16(tcp); | |
07148121 JG |
656 | } else { |
657 | memset(&key->tp, 0, sizeof(key->tp)); | |
ccb1352e | 658 | } |
07148121 | 659 | |
ccb1352e | 660 | } else if (key->ip.proto == IPPROTO_UDP) { |
ccb1352e JG |
661 | if (udphdr_ok(skb)) { |
662 | struct udphdr *udp = udp_hdr(skb); | |
1139e241 JR |
663 | key->tp.src = udp->source; |
664 | key->tp.dst = udp->dest; | |
07148121 JG |
665 | } else { |
666 | memset(&key->tp, 0, sizeof(key->tp)); | |
ccb1352e | 667 | } |
a175a723 JS |
668 | } else if (key->ip.proto == IPPROTO_SCTP) { |
669 | if (sctphdr_ok(skb)) { | |
670 | struct sctphdr *sctp = sctp_hdr(skb); | |
1139e241 JR |
671 | key->tp.src = sctp->source; |
672 | key->tp.dst = sctp->dest; | |
07148121 JG |
673 | } else { |
674 | memset(&key->tp, 0, sizeof(key->tp)); | |
a175a723 | 675 | } |
ccb1352e | 676 | } else if (key->ip.proto == IPPROTO_ICMP) { |
ccb1352e JG |
677 | if (icmphdr_ok(skb)) { |
678 | struct icmphdr *icmp = icmp_hdr(skb); | |
679 | /* The ICMP type and code fields use the 16-bit | |
680 | * transport port fields, so we need to store | |
681 | * them in 16-bit network byte order. */ | |
1139e241 JR |
682 | key->tp.src = htons(icmp->type); |
683 | key->tp.dst = htons(icmp->code); | |
07148121 JG |
684 | } else { |
685 | memset(&key->tp, 0, sizeof(key->tp)); | |
ccb1352e JG |
686 | } |
687 | } | |
688 | ||
07148121 JG |
689 | } else if (key->eth.type == htons(ETH_P_ARP) || |
690 | key->eth.type == htons(ETH_P_RARP)) { | |
ccb1352e | 691 | struct arp_eth_header *arp; |
389f4894 | 692 | bool arp_available = arphdr_ok(skb); |
ccb1352e JG |
693 | |
694 | arp = (struct arp_eth_header *)skb_network_header(skb); | |
695 | ||
389f4894 | 696 | if (arp_available && |
07148121 JG |
697 | arp->ar_hrd == htons(ARPHRD_ETHER) && |
698 | arp->ar_pro == htons(ETH_P_IP) && | |
699 | arp->ar_hln == ETH_ALEN && | |
700 | arp->ar_pln == 4) { | |
ccb1352e JG |
701 | |
702 | /* We only match on the lower 8 bits of the opcode. */ | |
703 | if (ntohs(arp->ar_op) <= 0xff) | |
704 | key->ip.proto = ntohs(arp->ar_op); | |
07148121 JG |
705 | else |
706 | key->ip.proto = 0; | |
707 | ||
d04d3829 MM |
708 | memcpy(&key->ipv4.addr.src, arp->ar_sip, sizeof(key->ipv4.addr.src)); |
709 | memcpy(&key->ipv4.addr.dst, arp->ar_tip, sizeof(key->ipv4.addr.dst)); | |
8c63ff09 JP |
710 | ether_addr_copy(key->ipv4.arp.sha, arp->ar_sha); |
711 | ether_addr_copy(key->ipv4.arp.tha, arp->ar_tha); | |
07148121 JG |
712 | } else { |
713 | memset(&key->ip, 0, sizeof(key->ip)); | |
714 | memset(&key->ipv4, 0, sizeof(key->ipv4)); | |
ccb1352e | 715 | } |
25cd9ba0 SH |
716 | } else if (eth_p_mpls(key->eth.type)) { |
717 | size_t stack_len = MPLS_HLEN; | |
718 | ||
f7d49bce | 719 | skb_set_inner_network_header(skb, skb->mac_len); |
25cd9ba0 SH |
720 | while (1) { |
721 | __be32 lse; | |
722 | ||
723 | error = check_header(skb, skb->mac_len + stack_len); | |
724 | if (unlikely(error)) | |
725 | return 0; | |
726 | ||
f7d49bce | 727 | memcpy(&lse, skb_inner_network_header(skb), MPLS_HLEN); |
25cd9ba0 SH |
728 | |
729 | if (stack_len == MPLS_HLEN) | |
730 | memcpy(&key->mpls.top_lse, &lse, MPLS_HLEN); | |
731 | ||
f7d49bce | 732 | skb_set_inner_network_header(skb, skb->mac_len + stack_len); |
25cd9ba0 SH |
733 | if (lse & htonl(MPLS_LS_S_MASK)) |
734 | break; | |
735 | ||
736 | stack_len += MPLS_HLEN; | |
737 | } | |
ccb1352e JG |
738 | } else if (key->eth.type == htons(ETH_P_IPV6)) { |
739 | int nh_len; /* IPv6 Header + Extensions */ | |
740 | ||
03f0d916 | 741 | nh_len = parse_ipv6hdr(skb, key); |
ccb1352e | 742 | if (unlikely(nh_len < 0)) { |
c30da497 SH |
743 | switch (nh_len) { |
744 | case -EINVAL: | |
745 | memset(&key->ip, 0, sizeof(key->ip)); | |
746 | memset(&key->ipv6.addr, 0, sizeof(key->ipv6.addr)); | |
747 | /* fall-through */ | |
748 | case -EPROTO: | |
ccb1352e | 749 | skb->transport_header = skb->network_header; |
03f0d916 | 750 | error = 0; |
c30da497 SH |
751 | break; |
752 | default: | |
ccb1352e | 753 | error = nh_len; |
03f0d916 AZ |
754 | } |
755 | return error; | |
ccb1352e JG |
756 | } |
757 | ||
758 | if (key->ip.frag == OVS_FRAG_TYPE_LATER) | |
03f0d916 | 759 | return 0; |
0c19f846 WB |
760 | if (skb_shinfo(skb)->gso_type & SKB_GSO_UDP) |
761 | key->ip.frag = OVS_FRAG_TYPE_FIRST; | |
762 | ||
ccb1352e JG |
763 | /* Transport layer. */ |
764 | if (key->ip.proto == NEXTHDR_TCP) { | |
ccb1352e JG |
765 | if (tcphdr_ok(skb)) { |
766 | struct tcphdr *tcp = tcp_hdr(skb); | |
1139e241 JR |
767 | key->tp.src = tcp->source; |
768 | key->tp.dst = tcp->dest; | |
769 | key->tp.flags = TCP_FLAGS_BE16(tcp); | |
07148121 JG |
770 | } else { |
771 | memset(&key->tp, 0, sizeof(key->tp)); | |
ccb1352e JG |
772 | } |
773 | } else if (key->ip.proto == NEXTHDR_UDP) { | |
ccb1352e JG |
774 | if (udphdr_ok(skb)) { |
775 | struct udphdr *udp = udp_hdr(skb); | |
1139e241 JR |
776 | key->tp.src = udp->source; |
777 | key->tp.dst = udp->dest; | |
07148121 JG |
778 | } else { |
779 | memset(&key->tp, 0, sizeof(key->tp)); | |
ccb1352e | 780 | } |
a175a723 JS |
781 | } else if (key->ip.proto == NEXTHDR_SCTP) { |
782 | if (sctphdr_ok(skb)) { | |
783 | struct sctphdr *sctp = sctp_hdr(skb); | |
1139e241 JR |
784 | key->tp.src = sctp->source; |
785 | key->tp.dst = sctp->dest; | |
07148121 JG |
786 | } else { |
787 | memset(&key->tp, 0, sizeof(key->tp)); | |
a175a723 | 788 | } |
ccb1352e | 789 | } else if (key->ip.proto == NEXTHDR_ICMP) { |
ccb1352e | 790 | if (icmp6hdr_ok(skb)) { |
03f0d916 AZ |
791 | error = parse_icmpv6(skb, key, nh_len); |
792 | if (error) | |
793 | return error; | |
07148121 JG |
794 | } else { |
795 | memset(&key->tp, 0, sizeof(key->tp)); | |
ccb1352e JG |
796 | } |
797 | } | |
b2d0f5d5 YY |
798 | } else if (key->eth.type == htons(ETH_P_NSH)) { |
799 | error = parse_nsh(skb, key); | |
800 | if (error) | |
801 | return error; | |
ccb1352e | 802 | } |
03f0d916 | 803 | return 0; |
ccb1352e | 804 | } |
83c8df26 | 805 | |
971427f3 AZ |
806 | int ovs_flow_key_update(struct sk_buff *skb, struct sw_flow_key *key) |
807 | { | |
6f56f618 YHW |
808 | int res; |
809 | ||
810 | res = key_extract(skb, key); | |
811 | if (!res) | |
812 | key->mac_proto &= ~SW_FLOW_KEY_INVALID; | |
813 | ||
814 | return res; | |
971427f3 AZ |
815 | } |
816 | ||
5108bbad JB |
817 | static int key_extract_mac_proto(struct sk_buff *skb) |
818 | { | |
819 | switch (skb->dev->type) { | |
820 | case ARPHRD_ETHER: | |
821 | return MAC_PROTO_ETHERNET; | |
822 | case ARPHRD_NONE: | |
823 | if (skb->protocol == htons(ETH_P_TEB)) | |
824 | return MAC_PROTO_ETHERNET; | |
825 | return MAC_PROTO_NONE; | |
826 | } | |
827 | WARN_ON_ONCE(1); | |
828 | return -EINVAL; | |
829 | } | |
830 | ||
1d8fff90 | 831 | int ovs_flow_key_extract(const struct ip_tunnel_info *tun_info, |
8c8b1b83 | 832 | struct sk_buff *skb, struct sw_flow_key *key) |
83c8df26 | 833 | { |
9dd7f890 | 834 | int res, err; |
5108bbad | 835 | |
83c8df26 | 836 | /* Extract metadata from packet. */ |
f5796684 | 837 | if (tun_info) { |
00a93bab | 838 | key->tun_proto = ip_tunnel_info_af(tun_info); |
1d8fff90 | 839 | memcpy(&key->tun_key, &tun_info->key, sizeof(key->tun_key)); |
f5796684 | 840 | |
4c222798 | 841 | if (tun_info->options_len) { |
f5796684 JG |
842 | BUILD_BUG_ON((1 << (sizeof(tun_info->options_len) * |
843 | 8)) - 1 | |
844 | > sizeof(key->tun_opts)); | |
4c222798 PS |
845 | |
846 | ip_tunnel_info_opts_get(TUN_METADATA_OPTS(key, tun_info->options_len), | |
847 | tun_info); | |
f5796684 JG |
848 | key->tun_opts_len = tun_info->options_len; |
849 | } else { | |
850 | key->tun_opts_len = 0; | |
851 | } | |
852 | } else { | |
00a93bab | 853 | key->tun_proto = 0; |
f5796684 | 854 | key->tun_opts_len = 0; |
07148121 | 855 | memset(&key->tun_key, 0, sizeof(key->tun_key)); |
f5796684 | 856 | } |
83c8df26 PS |
857 | |
858 | key->phy.priority = skb->priority; | |
859 | key->phy.in_port = OVS_CB(skb)->input_vport->port_no; | |
860 | key->phy.skb_mark = skb->mark; | |
07148121 | 861 | key->ovs_flow_hash = 0; |
5108bbad JB |
862 | res = key_extract_mac_proto(skb); |
863 | if (res < 0) | |
864 | return res; | |
865 | key->mac_proto = res; | |
07148121 JG |
866 | key->recirc_id = 0; |
867 | ||
9dd7f890 JR |
868 | err = key_extract(skb, key); |
869 | if (!err) | |
870 | ovs_ct_fill_key(skb, key); /* Must be after key_extract(). */ | |
871 | return err; | |
83c8df26 PS |
872 | } |
873 | ||
c2ac6673 | 874 | int ovs_flow_key_extract_userspace(struct net *net, const struct nlattr *attr, |
83c8df26 | 875 | struct sk_buff *skb, |
05da5898 | 876 | struct sw_flow_key *key, bool log) |
83c8df26 | 877 | { |
9dd7f890 JR |
878 | const struct nlattr *a[OVS_KEY_ATTR_MAX + 1]; |
879 | u64 attrs = 0; | |
83c8df26 PS |
880 | int err; |
881 | ||
9dd7f890 JR |
882 | err = parse_flow_nlattrs(attr, a, &attrs, log); |
883 | if (err) | |
884 | return -EINVAL; | |
885 | ||
83c8df26 | 886 | /* Extract metadata from netlink attributes. */ |
9dd7f890 | 887 | err = ovs_nla_get_flow_metadata(net, a, attrs, key, log); |
83c8df26 PS |
888 | if (err) |
889 | return err; | |
890 | ||
df30f740 | 891 | /* key_extract assumes that skb->protocol is set-up for |
892 | * layer 3 packets which is the case for other callers, | |
893 | * in particular packets received from the network stack. | |
894 | * Here the correct value can be set from the metadata | |
895 | * extracted above. | |
896 | * For L2 packet key eth type would be zero. skb protocol | |
897 | * would be set to correct value later during key-extact. | |
898 | */ | |
5108bbad | 899 | |
df30f740 | 900 | skb->protocol = key->eth.type; |
9dd7f890 JR |
901 | err = key_extract(skb, key); |
902 | if (err) | |
903 | return err; | |
904 | ||
905 | /* Check that we have conntrack original direction tuple metadata only | |
906 | * for packets for which it makes sense. Otherwise the key may be | |
907 | * corrupted due to overlapping key fields. | |
908 | */ | |
909 | if (attrs & (1 << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV4) && | |
910 | key->eth.type != htons(ETH_P_IP)) | |
911 | return -EINVAL; | |
912 | if (attrs & (1 << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6) && | |
913 | (key->eth.type != htons(ETH_P_IPV6) || | |
914 | sw_flow_key_is_nd(key))) | |
915 | return -EINVAL; | |
916 | ||
917 | return 0; | |
83c8df26 | 918 | } |