]>
Commit | Line | Data |
---|---|---|
7f8a436e JS |
1 | /* |
2 | * Copyright (c) 2015 Nicira, Inc. | |
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 | ||
14 | #include <linux/module.h> | |
15 | #include <linux/openvswitch.h> | |
05752523 JR |
16 | #include <linux/tcp.h> |
17 | #include <linux/udp.h> | |
18 | #include <linux/sctp.h> | |
7f8a436e JS |
19 | #include <net/ip.h> |
20 | #include <net/netfilter/nf_conntrack_core.h> | |
cae3a262 | 21 | #include <net/netfilter/nf_conntrack_helper.h> |
c2ac6673 | 22 | #include <net/netfilter/nf_conntrack_labels.h> |
05752523 | 23 | #include <net/netfilter/nf_conntrack_seqadj.h> |
7f8a436e JS |
24 | #include <net/netfilter/nf_conntrack_zones.h> |
25 | #include <net/netfilter/ipv6/nf_defrag_ipv6.h> | |
26 | ||
05752523 JR |
27 | #ifdef CONFIG_NF_NAT_NEEDED |
28 | #include <linux/netfilter/nf_nat.h> | |
29 | #include <net/netfilter/nf_nat_core.h> | |
30 | #include <net/netfilter/nf_nat_l3proto.h> | |
31 | #endif | |
32 | ||
7f8a436e JS |
33 | #include "datapath.h" |
34 | #include "conntrack.h" | |
35 | #include "flow.h" | |
36 | #include "flow_netlink.h" | |
37 | ||
38 | struct ovs_ct_len_tbl { | |
05752523 JR |
39 | int maxlen; |
40 | int minlen; | |
7f8a436e JS |
41 | }; |
42 | ||
182e3042 JS |
43 | /* Metadata mark for masked write to conntrack mark */ |
44 | struct md_mark { | |
45 | u32 value; | |
46 | u32 mask; | |
47 | }; | |
48 | ||
c2ac6673 | 49 | /* Metadata label for masked write to conntrack label. */ |
33db4125 JS |
50 | struct md_labels { |
51 | struct ovs_key_ct_labels value; | |
52 | struct ovs_key_ct_labels mask; | |
c2ac6673 JS |
53 | }; |
54 | ||
05752523 JR |
55 | enum ovs_ct_nat { |
56 | OVS_CT_NAT = 1 << 0, /* NAT for committed connections only. */ | |
57 | OVS_CT_SRC_NAT = 1 << 1, /* Source NAT for NEW connections. */ | |
58 | OVS_CT_DST_NAT = 1 << 2, /* Destination NAT for NEW connections. */ | |
59 | }; | |
60 | ||
7f8a436e JS |
61 | /* Conntrack action context for execution. */ |
62 | struct ovs_conntrack_info { | |
cae3a262 | 63 | struct nf_conntrack_helper *helper; |
7f8a436e JS |
64 | struct nf_conntrack_zone zone; |
65 | struct nf_conn *ct; | |
ab38a7b5 | 66 | u8 commit : 1; |
05752523 | 67 | u8 nat : 3; /* enum ovs_ct_nat */ |
7f8a436e | 68 | u16 family; |
182e3042 | 69 | struct md_mark mark; |
33db4125 | 70 | struct md_labels labels; |
05752523 JR |
71 | #ifdef CONFIG_NF_NAT_NEEDED |
72 | struct nf_nat_range range; /* Only present for SRC NAT and DST NAT. */ | |
73 | #endif | |
7f8a436e JS |
74 | }; |
75 | ||
2f3ab9f9 JS |
76 | static void __ovs_ct_free_action(struct ovs_conntrack_info *ct_info); |
77 | ||
7f8a436e JS |
78 | static u16 key_to_nfproto(const struct sw_flow_key *key) |
79 | { | |
80 | switch (ntohs(key->eth.type)) { | |
81 | case ETH_P_IP: | |
82 | return NFPROTO_IPV4; | |
83 | case ETH_P_IPV6: | |
84 | return NFPROTO_IPV6; | |
85 | default: | |
86 | return NFPROTO_UNSPEC; | |
87 | } | |
88 | } | |
89 | ||
90 | /* Map SKB connection state into the values used by flow definition. */ | |
91 | static u8 ovs_ct_get_state(enum ip_conntrack_info ctinfo) | |
92 | { | |
93 | u8 ct_state = OVS_CS_F_TRACKED; | |
94 | ||
95 | switch (ctinfo) { | |
96 | case IP_CT_ESTABLISHED_REPLY: | |
97 | case IP_CT_RELATED_REPLY: | |
7f8a436e JS |
98 | ct_state |= OVS_CS_F_REPLY_DIR; |
99 | break; | |
100 | default: | |
101 | break; | |
102 | } | |
103 | ||
104 | switch (ctinfo) { | |
105 | case IP_CT_ESTABLISHED: | |
106 | case IP_CT_ESTABLISHED_REPLY: | |
107 | ct_state |= OVS_CS_F_ESTABLISHED; | |
108 | break; | |
109 | case IP_CT_RELATED: | |
110 | case IP_CT_RELATED_REPLY: | |
111 | ct_state |= OVS_CS_F_RELATED; | |
112 | break; | |
113 | case IP_CT_NEW: | |
7f8a436e JS |
114 | ct_state |= OVS_CS_F_NEW; |
115 | break; | |
116 | default: | |
117 | break; | |
118 | } | |
119 | ||
120 | return ct_state; | |
121 | } | |
122 | ||
0d5cdef8 JS |
123 | static u32 ovs_ct_get_mark(const struct nf_conn *ct) |
124 | { | |
125 | #if IS_ENABLED(CONFIG_NF_CONNTRACK_MARK) | |
126 | return ct ? ct->mark : 0; | |
127 | #else | |
128 | return 0; | |
129 | #endif | |
130 | } | |
131 | ||
b87cec38 JR |
132 | /* Guard against conntrack labels max size shrinking below 128 bits. */ |
133 | #if NF_CT_LABELS_MAX_SIZE < 16 | |
134 | #error NF_CT_LABELS_MAX_SIZE must be at least 16 bytes | |
135 | #endif | |
136 | ||
33db4125 JS |
137 | static void ovs_ct_get_labels(const struct nf_conn *ct, |
138 | struct ovs_key_ct_labels *labels) | |
c2ac6673 JS |
139 | { |
140 | struct nf_conn_labels *cl = ct ? nf_ct_labels_find(ct) : NULL; | |
141 | ||
b87cec38 JR |
142 | if (cl) |
143 | memcpy(labels, cl->bits, OVS_CT_LABELS_LEN); | |
144 | else | |
33db4125 | 145 | memset(labels, 0, OVS_CT_LABELS_LEN); |
c2ac6673 JS |
146 | } |
147 | ||
7f8a436e | 148 | static void __ovs_ct_update_key(struct sw_flow_key *key, u8 state, |
182e3042 JS |
149 | const struct nf_conntrack_zone *zone, |
150 | const struct nf_conn *ct) | |
7f8a436e JS |
151 | { |
152 | key->ct.state = state; | |
153 | key->ct.zone = zone->id; | |
0d5cdef8 | 154 | key->ct.mark = ovs_ct_get_mark(ct); |
33db4125 | 155 | ovs_ct_get_labels(ct, &key->ct.labels); |
7f8a436e JS |
156 | } |
157 | ||
5e17da63 | 158 | /* Update 'key' based on skb->_nfct. If 'post_ct' is true, then OVS has |
05752523 JR |
159 | * previously sent the packet to conntrack via the ct action. If |
160 | * 'keep_nat_flags' is true, the existing NAT flags retained, else they are | |
161 | * initialized from the connection status. | |
7f8a436e JS |
162 | */ |
163 | static void ovs_ct_update_key(const struct sk_buff *skb, | |
d110986c | 164 | const struct ovs_conntrack_info *info, |
05752523 JR |
165 | struct sw_flow_key *key, bool post_ct, |
166 | bool keep_nat_flags) | |
7f8a436e JS |
167 | { |
168 | const struct nf_conntrack_zone *zone = &nf_ct_zone_dflt; | |
169 | enum ip_conntrack_info ctinfo; | |
170 | struct nf_conn *ct; | |
171 | u8 state = 0; | |
172 | ||
173 | ct = nf_ct_get(skb, &ctinfo); | |
174 | if (ct) { | |
175 | state = ovs_ct_get_state(ctinfo); | |
9f13ded8 | 176 | /* All unconfirmed entries are NEW connections. */ |
4f0909ee JS |
177 | if (!nf_ct_is_confirmed(ct)) |
178 | state |= OVS_CS_F_NEW; | |
9f13ded8 JR |
179 | /* OVS persists the related flag for the duration of the |
180 | * connection. | |
181 | */ | |
7f8a436e JS |
182 | if (ct->master) |
183 | state |= OVS_CS_F_RELATED; | |
05752523 JR |
184 | if (keep_nat_flags) { |
185 | state |= key->ct.state & OVS_CS_F_NAT_MASK; | |
186 | } else { | |
187 | if (ct->status & IPS_SRC_NAT) | |
188 | state |= OVS_CS_F_SRC_NAT; | |
189 | if (ct->status & IPS_DST_NAT) | |
190 | state |= OVS_CS_F_DST_NAT; | |
191 | } | |
7f8a436e JS |
192 | zone = nf_ct_zone(ct); |
193 | } else if (post_ct) { | |
194 | state = OVS_CS_F_TRACKED | OVS_CS_F_INVALID; | |
d110986c JS |
195 | if (info) |
196 | zone = &info->zone; | |
7f8a436e | 197 | } |
182e3042 | 198 | __ovs_ct_update_key(key, state, zone, ct); |
7f8a436e JS |
199 | } |
200 | ||
9f13ded8 JR |
201 | /* This is called to initialize CT key fields possibly coming in from the local |
202 | * stack. | |
203 | */ | |
7f8a436e JS |
204 | void ovs_ct_fill_key(const struct sk_buff *skb, struct sw_flow_key *key) |
205 | { | |
05752523 | 206 | ovs_ct_update_key(skb, NULL, key, false, false); |
7f8a436e JS |
207 | } |
208 | ||
209 | int ovs_ct_put_key(const struct sw_flow_key *key, struct sk_buff *skb) | |
210 | { | |
fbccce59 | 211 | if (nla_put_u32(skb, OVS_KEY_ATTR_CT_STATE, key->ct.state)) |
7f8a436e JS |
212 | return -EMSGSIZE; |
213 | ||
214 | if (IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES) && | |
215 | nla_put_u16(skb, OVS_KEY_ATTR_CT_ZONE, key->ct.zone)) | |
216 | return -EMSGSIZE; | |
217 | ||
182e3042 JS |
218 | if (IS_ENABLED(CONFIG_NF_CONNTRACK_MARK) && |
219 | nla_put_u32(skb, OVS_KEY_ATTR_CT_MARK, key->ct.mark)) | |
220 | return -EMSGSIZE; | |
221 | ||
9723e6ab | 222 | if (IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS) && |
33db4125 JS |
223 | nla_put(skb, OVS_KEY_ATTR_CT_LABELS, sizeof(key->ct.labels), |
224 | &key->ct.labels)) | |
c2ac6673 JS |
225 | return -EMSGSIZE; |
226 | ||
182e3042 JS |
227 | return 0; |
228 | } | |
229 | ||
6ffcea79 | 230 | static int ovs_ct_set_mark(struct nf_conn *ct, struct sw_flow_key *key, |
182e3042 JS |
231 | u32 ct_mark, u32 mask) |
232 | { | |
0d5cdef8 | 233 | #if IS_ENABLED(CONFIG_NF_CONNTRACK_MARK) |
182e3042 JS |
234 | u32 new_mark; |
235 | ||
182e3042 JS |
236 | new_mark = ct_mark | (ct->mark & ~(mask)); |
237 | if (ct->mark != new_mark) { | |
238 | ct->mark = new_mark; | |
193e3096 JR |
239 | if (nf_ct_is_confirmed(ct)) |
240 | nf_conntrack_event_cache(IPCT_MARK, ct); | |
182e3042 JS |
241 | key->ct.mark = new_mark; |
242 | } | |
243 | ||
7f8a436e | 244 | return 0; |
0d5cdef8 JS |
245 | #else |
246 | return -ENOTSUPP; | |
247 | #endif | |
7f8a436e JS |
248 | } |
249 | ||
6ffcea79 | 250 | static struct nf_conn_labels *ovs_ct_get_conn_labels(struct nf_conn *ct) |
c2ac6673 | 251 | { |
c2ac6673 | 252 | struct nf_conn_labels *cl; |
c2ac6673 JS |
253 | |
254 | cl = nf_ct_labels_find(ct); | |
255 | if (!cl) { | |
256 | nf_ct_labels_ext_add(ct); | |
257 | cl = nf_ct_labels_find(ct); | |
258 | } | |
6ffcea79 JR |
259 | |
260 | return cl; | |
261 | } | |
262 | ||
263 | /* Initialize labels for a new, yet to be committed conntrack entry. Note that | |
264 | * since the new connection is not yet confirmed, and thus no-one else has | |
265 | * access to it's labels, we simply write them over. Also, we refrain from | |
266 | * triggering events, as receiving change events before the create event would | |
267 | * be confusing. | |
268 | */ | |
269 | static int ovs_ct_init_labels(struct nf_conn *ct, struct sw_flow_key *key, | |
270 | const struct ovs_key_ct_labels *labels, | |
271 | const struct ovs_key_ct_labels *mask) | |
272 | { | |
273 | struct nf_conn_labels *cl; | |
274 | u32 *dst; | |
275 | int i; | |
276 | ||
277 | cl = ovs_ct_get_conn_labels(ct); | |
b87cec38 | 278 | if (!cl) |
c2ac6673 JS |
279 | return -ENOSPC; |
280 | ||
6ffcea79 JR |
281 | dst = (u32 *)cl->bits; |
282 | for (i = 0; i < OVS_CT_LABELS_LEN_32; i++) | |
283 | dst[i] = (dst[i] & ~mask->ct_labels_32[i]) | | |
284 | (labels->ct_labels_32[i] & mask->ct_labels_32[i]); | |
193e3096 | 285 | |
6ffcea79 JR |
286 | memcpy(&key->ct.labels, cl->bits, OVS_CT_LABELS_LEN); |
287 | ||
288 | return 0; | |
289 | } | |
290 | ||
291 | static int ovs_ct_set_labels(struct nf_conn *ct, struct sw_flow_key *key, | |
292 | const struct ovs_key_ct_labels *labels, | |
293 | const struct ovs_key_ct_labels *mask) | |
294 | { | |
295 | struct nf_conn_labels *cl; | |
296 | int err; | |
297 | ||
298 | cl = ovs_ct_get_conn_labels(ct); | |
299 | if (!cl) | |
300 | return -ENOSPC; | |
301 | ||
302 | err = nf_connlabels_replace(ct, labels->ct_labels_32, | |
303 | mask->ct_labels_32, | |
304 | OVS_CT_LABELS_LEN_32); | |
305 | if (err) | |
306 | return err; | |
307 | ||
308 | memcpy(&key->ct.labels, cl->bits, OVS_CT_LABELS_LEN); | |
c2ac6673 | 309 | |
c2ac6673 JS |
310 | return 0; |
311 | } | |
312 | ||
cae3a262 JS |
313 | /* 'skb' should already be pulled to nh_ofs. */ |
314 | static int ovs_ct_helper(struct sk_buff *skb, u16 proto) | |
315 | { | |
316 | const struct nf_conntrack_helper *helper; | |
317 | const struct nf_conn_help *help; | |
318 | enum ip_conntrack_info ctinfo; | |
319 | unsigned int protoff; | |
320 | struct nf_conn *ct; | |
05752523 | 321 | int err; |
cae3a262 JS |
322 | |
323 | ct = nf_ct_get(skb, &ctinfo); | |
324 | if (!ct || ctinfo == IP_CT_RELATED_REPLY) | |
325 | return NF_ACCEPT; | |
326 | ||
327 | help = nfct_help(ct); | |
328 | if (!help) | |
329 | return NF_ACCEPT; | |
330 | ||
331 | helper = rcu_dereference(help->helper); | |
332 | if (!helper) | |
333 | return NF_ACCEPT; | |
334 | ||
335 | switch (proto) { | |
336 | case NFPROTO_IPV4: | |
337 | protoff = ip_hdrlen(skb); | |
338 | break; | |
339 | case NFPROTO_IPV6: { | |
340 | u8 nexthdr = ipv6_hdr(skb)->nexthdr; | |
341 | __be16 frag_off; | |
cc570605 | 342 | int ofs; |
cae3a262 | 343 | |
cc570605 JS |
344 | ofs = ipv6_skip_exthdr(skb, sizeof(struct ipv6hdr), &nexthdr, |
345 | &frag_off); | |
346 | if (ofs < 0 || (frag_off & htons(~0x7)) != 0) { | |
cae3a262 JS |
347 | pr_debug("proto header not found\n"); |
348 | return NF_ACCEPT; | |
349 | } | |
cc570605 | 350 | protoff = ofs; |
cae3a262 JS |
351 | break; |
352 | } | |
353 | default: | |
354 | WARN_ONCE(1, "helper invoked on non-IP family!"); | |
355 | return NF_DROP; | |
356 | } | |
357 | ||
05752523 JR |
358 | err = helper->help(skb, protoff, ct, ctinfo); |
359 | if (err != NF_ACCEPT) | |
360 | return err; | |
361 | ||
362 | /* Adjust seqs after helper. This is needed due to some helpers (e.g., | |
363 | * FTP with NAT) adusting the TCP payload size when mangling IP | |
364 | * addresses and/or port numbers in the text-based control connection. | |
365 | */ | |
366 | if (test_bit(IPS_SEQ_ADJUST_BIT, &ct->status) && | |
367 | !nf_ct_seq_adjust(skb, ct, ctinfo, protoff)) | |
368 | return NF_DROP; | |
369 | return NF_ACCEPT; | |
cae3a262 JS |
370 | } |
371 | ||
74c16618 JS |
372 | /* Returns 0 on success, -EINPROGRESS if 'skb' is stolen, or other nonzero |
373 | * value if 'skb' is freed. | |
374 | */ | |
7f8a436e JS |
375 | static int handle_fragments(struct net *net, struct sw_flow_key *key, |
376 | u16 zone, struct sk_buff *skb) | |
377 | { | |
378 | struct ovs_skb_cb ovs_cb = *OVS_CB(skb); | |
daaa7d64 | 379 | int err; |
7f8a436e JS |
380 | |
381 | if (key->eth.type == htons(ETH_P_IP)) { | |
382 | enum ip_defrag_users user = IP_DEFRAG_CONNTRACK_IN + zone; | |
7f8a436e JS |
383 | |
384 | memset(IPCB(skb), 0, sizeof(struct inet_skb_parm)); | |
19bcf9f2 | 385 | err = ip_defrag(net, skb, user); |
7f8a436e JS |
386 | if (err) |
387 | return err; | |
388 | ||
389 | ovs_cb.mru = IPCB(skb)->frag_max_size; | |
7f8a436e | 390 | #if IS_ENABLED(CONFIG_NF_DEFRAG_IPV6) |
74c16618 | 391 | } else if (key->eth.type == htons(ETH_P_IPV6)) { |
7f8a436e | 392 | enum ip6_defrag_users user = IP6_DEFRAG_CONNTRACK_IN + zone; |
7f8a436e | 393 | |
49e261a8 | 394 | skb_orphan(skb); |
7f8a436e | 395 | memset(IP6CB(skb), 0, sizeof(struct inet6_skb_parm)); |
daaa7d64 | 396 | err = nf_ct_frag6_gather(net, skb, user); |
f92a80a9 DDP |
397 | if (err) { |
398 | if (err != -EINPROGRESS) | |
399 | kfree_skb(skb); | |
daaa7d64 | 400 | return err; |
f92a80a9 | 401 | } |
7f8a436e | 402 | |
daaa7d64 | 403 | key->ip.proto = ipv6_hdr(skb)->nexthdr; |
7f8a436e | 404 | ovs_cb.mru = IP6CB(skb)->frag_max_size; |
7f8a436e JS |
405 | #endif |
406 | } else { | |
74c16618 | 407 | kfree_skb(skb); |
7f8a436e JS |
408 | return -EPFNOSUPPORT; |
409 | } | |
410 | ||
411 | key->ip.frag = OVS_FRAG_TYPE_NONE; | |
412 | skb_clear_hash(skb); | |
413 | skb->ignore_df = 1; | |
414 | *OVS_CB(skb) = ovs_cb; | |
415 | ||
416 | return 0; | |
417 | } | |
418 | ||
419 | static struct nf_conntrack_expect * | |
420 | ovs_ct_expect_find(struct net *net, const struct nf_conntrack_zone *zone, | |
421 | u16 proto, const struct sk_buff *skb) | |
422 | { | |
423 | struct nf_conntrack_tuple tuple; | |
424 | ||
a31f1adc | 425 | if (!nf_ct_get_tuplepr(skb, skb_network_offset(skb), proto, net, &tuple)) |
7f8a436e JS |
426 | return NULL; |
427 | return __nf_ct_expect_find(net, zone, &tuple); | |
428 | } | |
429 | ||
289f2253 JR |
430 | /* This replicates logic from nf_conntrack_core.c that is not exported. */ |
431 | static enum ip_conntrack_info | |
432 | ovs_ct_get_info(const struct nf_conntrack_tuple_hash *h) | |
433 | { | |
434 | const struct nf_conn *ct = nf_ct_tuplehash_to_ctrack(h); | |
435 | ||
436 | if (NF_CT_DIRECTION(h) == IP_CT_DIR_REPLY) | |
437 | return IP_CT_ESTABLISHED_REPLY; | |
438 | /* Once we've had two way comms, always ESTABLISHED. */ | |
439 | if (test_bit(IPS_SEEN_REPLY_BIT, &ct->status)) | |
440 | return IP_CT_ESTABLISHED; | |
441 | if (test_bit(IPS_EXPECTED_BIT, &ct->status)) | |
442 | return IP_CT_RELATED; | |
443 | return IP_CT_NEW; | |
444 | } | |
445 | ||
446 | /* Find an existing connection which this packet belongs to without | |
447 | * re-attributing statistics or modifying the connection state. This allows an | |
5e17da63 | 448 | * skb->_nfct lost due to an upcall to be recovered during actions execution. |
289f2253 JR |
449 | * |
450 | * Must be called with rcu_read_lock. | |
451 | * | |
5e17da63 JR |
452 | * On success, populates skb->_nfct and returns the connection. Returns NULL |
453 | * if there is no existing entry. | |
289f2253 JR |
454 | */ |
455 | static struct nf_conn * | |
456 | ovs_ct_find_existing(struct net *net, const struct nf_conntrack_zone *zone, | |
9ff464db | 457 | u8 l3num, struct sk_buff *skb, bool natted) |
289f2253 JR |
458 | { |
459 | struct nf_conntrack_l3proto *l3proto; | |
460 | struct nf_conntrack_l4proto *l4proto; | |
461 | struct nf_conntrack_tuple tuple; | |
462 | struct nf_conntrack_tuple_hash *h; | |
289f2253 JR |
463 | struct nf_conn *ct; |
464 | unsigned int dataoff; | |
465 | u8 protonum; | |
466 | ||
467 | l3proto = __nf_ct_l3proto_find(l3num); | |
289f2253 JR |
468 | if (l3proto->get_l4proto(skb, skb_network_offset(skb), &dataoff, |
469 | &protonum) <= 0) { | |
470 | pr_debug("ovs_ct_find_existing: Can't get protonum\n"); | |
471 | return NULL; | |
472 | } | |
473 | l4proto = __nf_ct_l4proto_find(l3num, protonum); | |
289f2253 JR |
474 | if (!nf_ct_get_tuple(skb, skb_network_offset(skb), dataoff, l3num, |
475 | protonum, net, &tuple, l3proto, l4proto)) { | |
476 | pr_debug("ovs_ct_find_existing: Can't get tuple\n"); | |
477 | return NULL; | |
478 | } | |
479 | ||
9ff464db JR |
480 | /* Must invert the tuple if skb has been transformed by NAT. */ |
481 | if (natted) { | |
482 | struct nf_conntrack_tuple inverse; | |
483 | ||
484 | if (!nf_ct_invert_tuple(&inverse, &tuple, l3proto, l4proto)) { | |
485 | pr_debug("ovs_ct_find_existing: Inversion failed!\n"); | |
486 | return NULL; | |
487 | } | |
488 | tuple = inverse; | |
489 | } | |
490 | ||
289f2253 JR |
491 | /* look for tuple match */ |
492 | h = nf_conntrack_find_get(net, zone, &tuple); | |
493 | if (!h) | |
494 | return NULL; /* Not found. */ | |
495 | ||
496 | ct = nf_ct_tuplehash_to_ctrack(h); | |
497 | ||
9ff464db JR |
498 | /* Inverted packet tuple matches the reverse direction conntrack tuple, |
499 | * select the other tuplehash to get the right 'ctinfo' bits for this | |
500 | * packet. | |
501 | */ | |
502 | if (natted) | |
503 | h = &ct->tuplehash[!h->tuple.dst.dir]; | |
504 | ||
c74454fa | 505 | nf_ct_set(skb, ct, ovs_ct_get_info(h)); |
289f2253 JR |
506 | return ct; |
507 | } | |
508 | ||
5e17da63 | 509 | /* Determine whether skb->_nfct is equal to the result of conntrack lookup. */ |
289f2253 JR |
510 | static bool skb_nfct_cached(struct net *net, |
511 | const struct sw_flow_key *key, | |
512 | const struct ovs_conntrack_info *info, | |
513 | struct sk_buff *skb) | |
7f8a436e JS |
514 | { |
515 | enum ip_conntrack_info ctinfo; | |
516 | struct nf_conn *ct; | |
517 | ||
518 | ct = nf_ct_get(skb, &ctinfo); | |
289f2253 | 519 | /* If no ct, check if we have evidence that an existing conntrack entry |
5e17da63 | 520 | * might be found for this skb. This happens when we lose a skb->_nfct |
289f2253 JR |
521 | * due to an upcall. If the connection was not confirmed, it is not |
522 | * cached and needs to be run through conntrack again. | |
523 | */ | |
524 | if (!ct && key->ct.state & OVS_CS_F_TRACKED && | |
525 | !(key->ct.state & OVS_CS_F_INVALID) && | |
526 | key->ct.zone == info->zone.id) | |
9ff464db JR |
527 | ct = ovs_ct_find_existing(net, &info->zone, info->family, skb, |
528 | !!(key->ct.state | |
529 | & OVS_CS_F_NAT_MASK)); | |
7f8a436e JS |
530 | if (!ct) |
531 | return false; | |
532 | if (!net_eq(net, read_pnet(&ct->ct_net))) | |
533 | return false; | |
534 | if (!nf_ct_zone_equal_any(info->ct, nf_ct_zone(ct))) | |
535 | return false; | |
cae3a262 JS |
536 | if (info->helper) { |
537 | struct nf_conn_help *help; | |
538 | ||
539 | help = nf_ct_ext_find(ct, NF_CT_EXT_HELPER); | |
540 | if (help && rcu_access_pointer(help->helper) != info->helper) | |
541 | return false; | |
542 | } | |
7f8a436e JS |
543 | |
544 | return true; | |
545 | } | |
546 | ||
05752523 JR |
547 | #ifdef CONFIG_NF_NAT_NEEDED |
548 | /* Modelled after nf_nat_ipv[46]_fn(). | |
549 | * range is only used for new, uninitialized NAT state. | |
550 | * Returns either NF_ACCEPT or NF_DROP. | |
551 | */ | |
552 | static int ovs_ct_nat_execute(struct sk_buff *skb, struct nf_conn *ct, | |
553 | enum ip_conntrack_info ctinfo, | |
554 | const struct nf_nat_range *range, | |
555 | enum nf_nat_manip_type maniptype) | |
556 | { | |
557 | int hooknum, nh_off, err = NF_ACCEPT; | |
558 | ||
559 | nh_off = skb_network_offset(skb); | |
75f01a4c | 560 | skb_pull_rcsum(skb, nh_off); |
05752523 JR |
561 | |
562 | /* See HOOK2MANIP(). */ | |
563 | if (maniptype == NF_NAT_MANIP_SRC) | |
564 | hooknum = NF_INET_LOCAL_IN; /* Source NAT */ | |
565 | else | |
566 | hooknum = NF_INET_LOCAL_OUT; /* Destination NAT */ | |
567 | ||
568 | switch (ctinfo) { | |
569 | case IP_CT_RELATED: | |
570 | case IP_CT_RELATED_REPLY: | |
99b7248e AB |
571 | if (IS_ENABLED(CONFIG_NF_NAT_IPV4) && |
572 | skb->protocol == htons(ETH_P_IP) && | |
05752523 JR |
573 | ip_hdr(skb)->protocol == IPPROTO_ICMP) { |
574 | if (!nf_nat_icmp_reply_translation(skb, ct, ctinfo, | |
575 | hooknum)) | |
576 | err = NF_DROP; | |
577 | goto push; | |
99b7248e AB |
578 | } else if (IS_ENABLED(CONFIG_NF_NAT_IPV6) && |
579 | skb->protocol == htons(ETH_P_IPV6)) { | |
05752523 JR |
580 | __be16 frag_off; |
581 | u8 nexthdr = ipv6_hdr(skb)->nexthdr; | |
582 | int hdrlen = ipv6_skip_exthdr(skb, | |
583 | sizeof(struct ipv6hdr), | |
584 | &nexthdr, &frag_off); | |
585 | ||
586 | if (hdrlen >= 0 && nexthdr == IPPROTO_ICMPV6) { | |
587 | if (!nf_nat_icmpv6_reply_translation(skb, ct, | |
588 | ctinfo, | |
589 | hooknum, | |
590 | hdrlen)) | |
591 | err = NF_DROP; | |
592 | goto push; | |
593 | } | |
05752523 JR |
594 | } |
595 | /* Non-ICMP, fall thru to initialize if needed. */ | |
596 | case IP_CT_NEW: | |
597 | /* Seen it before? This can happen for loopback, retrans, | |
598 | * or local packets. | |
599 | */ | |
600 | if (!nf_nat_initialized(ct, maniptype)) { | |
601 | /* Initialize according to the NAT action. */ | |
602 | err = (range && range->flags & NF_NAT_RANGE_MAP_IPS) | |
603 | /* Action is set up to establish a new | |
604 | * mapping. | |
605 | */ | |
606 | ? nf_nat_setup_info(ct, range, maniptype) | |
607 | : nf_nat_alloc_null_binding(ct, hooknum); | |
608 | if (err != NF_ACCEPT) | |
609 | goto push; | |
610 | } | |
611 | break; | |
612 | ||
613 | case IP_CT_ESTABLISHED: | |
614 | case IP_CT_ESTABLISHED_REPLY: | |
615 | break; | |
616 | ||
617 | default: | |
618 | err = NF_DROP; | |
619 | goto push; | |
620 | } | |
621 | ||
622 | err = nf_nat_packet(ct, ctinfo, hooknum, skb); | |
623 | push: | |
624 | skb_push(skb, nh_off); | |
75f01a4c | 625 | skb_postpush_rcsum(skb, skb->data, nh_off); |
05752523 JR |
626 | |
627 | return err; | |
628 | } | |
629 | ||
630 | static void ovs_nat_update_key(struct sw_flow_key *key, | |
631 | const struct sk_buff *skb, | |
632 | enum nf_nat_manip_type maniptype) | |
633 | { | |
634 | if (maniptype == NF_NAT_MANIP_SRC) { | |
635 | __be16 src; | |
636 | ||
637 | key->ct.state |= OVS_CS_F_SRC_NAT; | |
638 | if (key->eth.type == htons(ETH_P_IP)) | |
639 | key->ipv4.addr.src = ip_hdr(skb)->saddr; | |
640 | else if (key->eth.type == htons(ETH_P_IPV6)) | |
641 | memcpy(&key->ipv6.addr.src, &ipv6_hdr(skb)->saddr, | |
642 | sizeof(key->ipv6.addr.src)); | |
643 | else | |
644 | return; | |
645 | ||
646 | if (key->ip.proto == IPPROTO_UDP) | |
647 | src = udp_hdr(skb)->source; | |
648 | else if (key->ip.proto == IPPROTO_TCP) | |
649 | src = tcp_hdr(skb)->source; | |
650 | else if (key->ip.proto == IPPROTO_SCTP) | |
651 | src = sctp_hdr(skb)->source; | |
652 | else | |
653 | return; | |
654 | ||
655 | key->tp.src = src; | |
656 | } else { | |
657 | __be16 dst; | |
658 | ||
659 | key->ct.state |= OVS_CS_F_DST_NAT; | |
660 | if (key->eth.type == htons(ETH_P_IP)) | |
661 | key->ipv4.addr.dst = ip_hdr(skb)->daddr; | |
662 | else if (key->eth.type == htons(ETH_P_IPV6)) | |
663 | memcpy(&key->ipv6.addr.dst, &ipv6_hdr(skb)->daddr, | |
664 | sizeof(key->ipv6.addr.dst)); | |
665 | else | |
666 | return; | |
667 | ||
668 | if (key->ip.proto == IPPROTO_UDP) | |
669 | dst = udp_hdr(skb)->dest; | |
670 | else if (key->ip.proto == IPPROTO_TCP) | |
671 | dst = tcp_hdr(skb)->dest; | |
672 | else if (key->ip.proto == IPPROTO_SCTP) | |
673 | dst = sctp_hdr(skb)->dest; | |
674 | else | |
675 | return; | |
676 | ||
677 | key->tp.dst = dst; | |
678 | } | |
679 | } | |
680 | ||
681 | /* Returns NF_DROP if the packet should be dropped, NF_ACCEPT otherwise. */ | |
682 | static int ovs_ct_nat(struct net *net, struct sw_flow_key *key, | |
683 | const struct ovs_conntrack_info *info, | |
684 | struct sk_buff *skb, struct nf_conn *ct, | |
685 | enum ip_conntrack_info ctinfo) | |
686 | { | |
687 | enum nf_nat_manip_type maniptype; | |
688 | int err; | |
689 | ||
690 | if (nf_ct_is_untracked(ct)) { | |
691 | /* A NAT action may only be performed on tracked packets. */ | |
692 | return NF_ACCEPT; | |
693 | } | |
694 | ||
695 | /* Add NAT extension if not confirmed yet. */ | |
696 | if (!nf_ct_is_confirmed(ct) && !nf_ct_nat_ext_add(ct)) | |
697 | return NF_ACCEPT; /* Can't NAT. */ | |
698 | ||
699 | /* Determine NAT type. | |
700 | * Check if the NAT type can be deduced from the tracked connection. | |
5745b0be JR |
701 | * Make sure new expected connections (IP_CT_RELATED) are NATted only |
702 | * when committing. | |
05752523 JR |
703 | */ |
704 | if (info->nat & OVS_CT_NAT && ctinfo != IP_CT_NEW && | |
705 | ct->status & IPS_NAT_MASK && | |
5745b0be | 706 | (ctinfo != IP_CT_RELATED || info->commit)) { |
05752523 JR |
707 | /* NAT an established or related connection like before. */ |
708 | if (CTINFO2DIR(ctinfo) == IP_CT_DIR_REPLY) | |
709 | /* This is the REPLY direction for a connection | |
710 | * for which NAT was applied in the forward | |
711 | * direction. Do the reverse NAT. | |
712 | */ | |
713 | maniptype = ct->status & IPS_SRC_NAT | |
714 | ? NF_NAT_MANIP_DST : NF_NAT_MANIP_SRC; | |
715 | else | |
716 | maniptype = ct->status & IPS_SRC_NAT | |
717 | ? NF_NAT_MANIP_SRC : NF_NAT_MANIP_DST; | |
718 | } else if (info->nat & OVS_CT_SRC_NAT) { | |
719 | maniptype = NF_NAT_MANIP_SRC; | |
720 | } else if (info->nat & OVS_CT_DST_NAT) { | |
721 | maniptype = NF_NAT_MANIP_DST; | |
722 | } else { | |
723 | return NF_ACCEPT; /* Connection is not NATed. */ | |
724 | } | |
725 | err = ovs_ct_nat_execute(skb, ct, ctinfo, &info->range, maniptype); | |
726 | ||
727 | /* Mark NAT done if successful and update the flow key. */ | |
728 | if (err == NF_ACCEPT) | |
729 | ovs_nat_update_key(key, skb, maniptype); | |
730 | ||
731 | return err; | |
732 | } | |
733 | #else /* !CONFIG_NF_NAT_NEEDED */ | |
734 | static int ovs_ct_nat(struct net *net, struct sw_flow_key *key, | |
735 | const struct ovs_conntrack_info *info, | |
736 | struct sk_buff *skb, struct nf_conn *ct, | |
737 | enum ip_conntrack_info ctinfo) | |
738 | { | |
739 | return NF_ACCEPT; | |
740 | } | |
741 | #endif | |
742 | ||
9f13ded8 | 743 | /* Pass 'skb' through conntrack in 'net', using zone configured in 'info', if |
394e910e JR |
744 | * not done already. Update key with new CT state after passing the packet |
745 | * through conntrack. | |
5e17da63 | 746 | * Note that if the packet is deemed invalid by conntrack, skb->_nfct will be |
9f13ded8 JR |
747 | * set to NULL and 0 will be returned. |
748 | */ | |
4f0909ee | 749 | static int __ovs_ct_lookup(struct net *net, struct sw_flow_key *key, |
7f8a436e JS |
750 | const struct ovs_conntrack_info *info, |
751 | struct sk_buff *skb) | |
752 | { | |
753 | /* If we are recirculating packets to match on conntrack fields and | |
754 | * committing with a separate conntrack action, then we don't need to | |
755 | * actually run the packet through conntrack twice unless it's for a | |
756 | * different zone. | |
757 | */ | |
28b6e0c1 JR |
758 | bool cached = skb_nfct_cached(net, key, info, skb); |
759 | enum ip_conntrack_info ctinfo; | |
760 | struct nf_conn *ct; | |
761 | ||
762 | if (!cached) { | |
7f8a436e | 763 | struct nf_conn *tmpl = info->ct; |
5b6b9293 | 764 | int err; |
7f8a436e JS |
765 | |
766 | /* Associate skb with specified zone. */ | |
767 | if (tmpl) { | |
cb9c6836 FW |
768 | if (skb_nfct(skb)) |
769 | nf_conntrack_put(skb_nfct(skb)); | |
7f8a436e | 770 | nf_conntrack_get(&tmpl->ct_general); |
c74454fa | 771 | nf_ct_set(skb, tmpl, IP_CT_NEW); |
7f8a436e JS |
772 | } |
773 | ||
08733a0c PNA |
774 | err = nf_conntrack_in(net, info->family, |
775 | NF_INET_PRE_ROUTING, skb); | |
5b6b9293 | 776 | if (err != NF_ACCEPT) |
7f8a436e | 777 | return -ENOENT; |
cae3a262 | 778 | |
05752523 JR |
779 | /* Clear CT state NAT flags to mark that we have not yet done |
780 | * NAT after the nf_conntrack_in() call. We can actually clear | |
781 | * the whole state, as it will be re-initialized below. | |
782 | */ | |
783 | key->ct.state = 0; | |
784 | ||
785 | /* Update the key, but keep the NAT flags. */ | |
786 | ovs_ct_update_key(skb, info, key, true, true); | |
28b6e0c1 | 787 | } |
394e910e | 788 | |
28b6e0c1 | 789 | ct = nf_ct_get(skb, &ctinfo); |
05752523 JR |
790 | if (ct) { |
791 | /* Packets starting a new connection must be NATted before the | |
792 | * helper, so that the helper knows about the NAT. We enforce | |
793 | * this by delaying both NAT and helper calls for unconfirmed | |
794 | * connections until the committing CT action. For later | |
795 | * packets NAT and Helper may be called in either order. | |
796 | * | |
797 | * NAT will be done only if the CT action has NAT, and only | |
798 | * once per packet (per zone), as guarded by the NAT bits in | |
799 | * the key->ct.state. | |
800 | */ | |
801 | if (info->nat && !(key->ct.state & OVS_CS_F_NAT_MASK) && | |
802 | (nf_ct_is_confirmed(ct) || info->commit) && | |
803 | ovs_ct_nat(net, key, info, skb, ct, ctinfo) != NF_ACCEPT) { | |
804 | return -EINVAL; | |
805 | } | |
806 | ||
16ec3d4f JS |
807 | /* Userspace may decide to perform a ct lookup without a helper |
808 | * specified followed by a (recirculate and) commit with one. | |
809 | * Therefore, for unconfirmed connections which we will commit, | |
810 | * we need to attach the helper here. | |
811 | */ | |
812 | if (!nf_ct_is_confirmed(ct) && info->commit && | |
813 | info->helper && !nfct_help(ct)) { | |
814 | int err = __nf_ct_try_assign_helper(ct, info->ct, | |
815 | GFP_ATOMIC); | |
816 | if (err) | |
817 | return err; | |
818 | } | |
819 | ||
05752523 JR |
820 | /* Call the helper only if: |
821 | * - nf_conntrack_in() was executed above ("!cached") for a | |
822 | * confirmed connection, or | |
823 | * - When committing an unconfirmed connection. | |
824 | */ | |
825 | if ((nf_ct_is_confirmed(ct) ? !cached : info->commit) && | |
826 | ovs_ct_helper(skb, info->family) != NF_ACCEPT) { | |
827 | return -EINVAL; | |
828 | } | |
7f8a436e JS |
829 | } |
830 | ||
831 | return 0; | |
832 | } | |
833 | ||
834 | /* Lookup connection and read fields into key. */ | |
835 | static int ovs_ct_lookup(struct net *net, struct sw_flow_key *key, | |
836 | const struct ovs_conntrack_info *info, | |
837 | struct sk_buff *skb) | |
838 | { | |
839 | struct nf_conntrack_expect *exp; | |
840 | ||
9f13ded8 JR |
841 | /* If we pass an expected packet through nf_conntrack_in() the |
842 | * expectation is typically removed, but the packet could still be | |
843 | * lost in upcall processing. To prevent this from happening we | |
844 | * perform an explicit expectation lookup. Expected connections are | |
845 | * always new, and will be passed through conntrack only when they are | |
846 | * committed, as it is OK to remove the expectation at that time. | |
847 | */ | |
7f8a436e JS |
848 | exp = ovs_ct_expect_find(net, &info->zone, info->family, skb); |
849 | if (exp) { | |
850 | u8 state; | |
851 | ||
05752523 JR |
852 | /* NOTE: New connections are NATted and Helped only when |
853 | * committed, so we are not calling into NAT here. | |
854 | */ | |
7f8a436e | 855 | state = OVS_CS_F_TRACKED | OVS_CS_F_NEW | OVS_CS_F_RELATED; |
182e3042 | 856 | __ovs_ct_update_key(key, state, &info->zone, exp->master); |
d913d3a7 SG |
857 | } else { |
858 | struct nf_conn *ct; | |
859 | int err; | |
860 | ||
861 | err = __ovs_ct_lookup(net, key, info, skb); | |
862 | if (err) | |
863 | return err; | |
864 | ||
cb9c6836 | 865 | ct = (struct nf_conn *)skb_nfct(skb); |
d913d3a7 SG |
866 | if (ct) |
867 | nf_ct_deliver_cached_events(ct); | |
868 | } | |
7f8a436e JS |
869 | |
870 | return 0; | |
871 | } | |
872 | ||
33db4125 | 873 | static bool labels_nonzero(const struct ovs_key_ct_labels *labels) |
c2ac6673 JS |
874 | { |
875 | size_t i; | |
876 | ||
cb80d58f JR |
877 | for (i = 0; i < OVS_CT_LABELS_LEN_32; i++) |
878 | if (labels->ct_labels_32[i]) | |
c2ac6673 JS |
879 | return true; |
880 | ||
881 | return false; | |
882 | } | |
883 | ||
7d904c7b JR |
884 | /* Lookup connection and confirm if unconfirmed. */ |
885 | static int ovs_ct_commit(struct net *net, struct sw_flow_key *key, | |
886 | const struct ovs_conntrack_info *info, | |
887 | struct sk_buff *skb) | |
888 | { | |
6ffcea79 JR |
889 | enum ip_conntrack_info ctinfo; |
890 | struct nf_conn *ct; | |
7d904c7b JR |
891 | int err; |
892 | ||
893 | err = __ovs_ct_lookup(net, key, info, skb); | |
894 | if (err) | |
895 | return err; | |
896 | ||
6ffcea79 JR |
897 | /* The connection could be invalid, in which case this is a no-op.*/ |
898 | ct = nf_ct_get(skb, &ctinfo); | |
899 | if (!ct) | |
900 | return 0; | |
901 | ||
7d904c7b JR |
902 | /* Apply changes before confirming the connection so that the initial |
903 | * conntrack NEW netlink event carries the values given in the CT | |
904 | * action. | |
905 | */ | |
906 | if (info->mark.mask) { | |
6ffcea79 | 907 | err = ovs_ct_set_mark(ct, key, info->mark.value, |
7d904c7b JR |
908 | info->mark.mask); |
909 | if (err) | |
910 | return err; | |
911 | } | |
912 | if (labels_nonzero(&info->labels.mask)) { | |
6ffcea79 JR |
913 | if (!nf_ct_is_confirmed(ct)) |
914 | err = ovs_ct_init_labels(ct, key, &info->labels.value, | |
915 | &info->labels.mask); | |
916 | else | |
917 | err = ovs_ct_set_labels(ct, key, &info->labels.value, | |
918 | &info->labels.mask); | |
7d904c7b JR |
919 | if (err) |
920 | return err; | |
921 | } | |
922 | /* This will take care of sending queued events even if the connection | |
923 | * is already confirmed. | |
924 | */ | |
925 | if (nf_conntrack_confirm(skb) != NF_ACCEPT) | |
926 | return -EINVAL; | |
927 | ||
928 | return 0; | |
929 | } | |
930 | ||
74c16618 JS |
931 | /* Returns 0 on success, -EINPROGRESS if 'skb' is stolen, or other nonzero |
932 | * value if 'skb' is freed. | |
933 | */ | |
7f8a436e JS |
934 | int ovs_ct_execute(struct net *net, struct sk_buff *skb, |
935 | struct sw_flow_key *key, | |
936 | const struct ovs_conntrack_info *info) | |
937 | { | |
938 | int nh_ofs; | |
939 | int err; | |
940 | ||
941 | /* The conntrack module expects to be working at L3. */ | |
942 | nh_ofs = skb_network_offset(skb); | |
75f01a4c | 943 | skb_pull_rcsum(skb, nh_ofs); |
7f8a436e JS |
944 | |
945 | if (key->ip.frag != OVS_FRAG_TYPE_NONE) { | |
946 | err = handle_fragments(net, key, info->zone.id, skb); | |
947 | if (err) | |
948 | return err; | |
949 | } | |
950 | ||
ab38a7b5 | 951 | if (info->commit) |
7d904c7b | 952 | err = ovs_ct_commit(net, key, info, skb); |
7f8a436e JS |
953 | else |
954 | err = ovs_ct_lookup(net, key, info, skb); | |
955 | ||
956 | skb_push(skb, nh_ofs); | |
75f01a4c | 957 | skb_postpush_rcsum(skb, skb->data, nh_ofs); |
74c16618 JS |
958 | if (err) |
959 | kfree_skb(skb); | |
7f8a436e JS |
960 | return err; |
961 | } | |
962 | ||
cae3a262 JS |
963 | static int ovs_ct_add_helper(struct ovs_conntrack_info *info, const char *name, |
964 | const struct sw_flow_key *key, bool log) | |
965 | { | |
966 | struct nf_conntrack_helper *helper; | |
967 | struct nf_conn_help *help; | |
968 | ||
969 | helper = nf_conntrack_helper_try_module_get(name, info->family, | |
970 | key->ip.proto); | |
971 | if (!helper) { | |
972 | OVS_NLERR(log, "Unknown helper \"%s\"", name); | |
973 | return -EINVAL; | |
974 | } | |
975 | ||
976 | help = nf_ct_helper_ext_add(info->ct, helper, GFP_KERNEL); | |
977 | if (!help) { | |
978 | module_put(helper->me); | |
979 | return -ENOMEM; | |
980 | } | |
981 | ||
982 | rcu_assign_pointer(help->helper, helper); | |
983 | info->helper = helper; | |
984 | return 0; | |
985 | } | |
986 | ||
05752523 JR |
987 | #ifdef CONFIG_NF_NAT_NEEDED |
988 | static int parse_nat(const struct nlattr *attr, | |
989 | struct ovs_conntrack_info *info, bool log) | |
990 | { | |
991 | struct nlattr *a; | |
992 | int rem; | |
993 | bool have_ip_max = false; | |
994 | bool have_proto_max = false; | |
995 | bool ip_vers = (info->family == NFPROTO_IPV6); | |
996 | ||
997 | nla_for_each_nested(a, attr, rem) { | |
998 | static const int ovs_nat_attr_lens[OVS_NAT_ATTR_MAX + 1][2] = { | |
999 | [OVS_NAT_ATTR_SRC] = {0, 0}, | |
1000 | [OVS_NAT_ATTR_DST] = {0, 0}, | |
1001 | [OVS_NAT_ATTR_IP_MIN] = {sizeof(struct in_addr), | |
1002 | sizeof(struct in6_addr)}, | |
1003 | [OVS_NAT_ATTR_IP_MAX] = {sizeof(struct in_addr), | |
1004 | sizeof(struct in6_addr)}, | |
1005 | [OVS_NAT_ATTR_PROTO_MIN] = {sizeof(u16), sizeof(u16)}, | |
1006 | [OVS_NAT_ATTR_PROTO_MAX] = {sizeof(u16), sizeof(u16)}, | |
1007 | [OVS_NAT_ATTR_PERSISTENT] = {0, 0}, | |
1008 | [OVS_NAT_ATTR_PROTO_HASH] = {0, 0}, | |
1009 | [OVS_NAT_ATTR_PROTO_RANDOM] = {0, 0}, | |
1010 | }; | |
1011 | int type = nla_type(a); | |
1012 | ||
1013 | if (type > OVS_NAT_ATTR_MAX) { | |
1014 | OVS_NLERR(log, | |
1015 | "Unknown NAT attribute (type=%d, max=%d).\n", | |
1016 | type, OVS_NAT_ATTR_MAX); | |
1017 | return -EINVAL; | |
1018 | } | |
1019 | ||
1020 | if (nla_len(a) != ovs_nat_attr_lens[type][ip_vers]) { | |
1021 | OVS_NLERR(log, | |
1022 | "NAT attribute type %d has unexpected length (%d != %d).\n", | |
1023 | type, nla_len(a), | |
1024 | ovs_nat_attr_lens[type][ip_vers]); | |
1025 | return -EINVAL; | |
1026 | } | |
1027 | ||
1028 | switch (type) { | |
1029 | case OVS_NAT_ATTR_SRC: | |
1030 | case OVS_NAT_ATTR_DST: | |
1031 | if (info->nat) { | |
1032 | OVS_NLERR(log, | |
1033 | "Only one type of NAT may be specified.\n" | |
1034 | ); | |
1035 | return -ERANGE; | |
1036 | } | |
1037 | info->nat |= OVS_CT_NAT; | |
1038 | info->nat |= ((type == OVS_NAT_ATTR_SRC) | |
1039 | ? OVS_CT_SRC_NAT : OVS_CT_DST_NAT); | |
1040 | break; | |
1041 | ||
1042 | case OVS_NAT_ATTR_IP_MIN: | |
ac71b46e HY |
1043 | nla_memcpy(&info->range.min_addr, a, |
1044 | sizeof(info->range.min_addr)); | |
05752523 JR |
1045 | info->range.flags |= NF_NAT_RANGE_MAP_IPS; |
1046 | break; | |
1047 | ||
1048 | case OVS_NAT_ATTR_IP_MAX: | |
1049 | have_ip_max = true; | |
1050 | nla_memcpy(&info->range.max_addr, a, | |
1051 | sizeof(info->range.max_addr)); | |
1052 | info->range.flags |= NF_NAT_RANGE_MAP_IPS; | |
1053 | break; | |
1054 | ||
1055 | case OVS_NAT_ATTR_PROTO_MIN: | |
1056 | info->range.min_proto.all = htons(nla_get_u16(a)); | |
1057 | info->range.flags |= NF_NAT_RANGE_PROTO_SPECIFIED; | |
1058 | break; | |
1059 | ||
1060 | case OVS_NAT_ATTR_PROTO_MAX: | |
1061 | have_proto_max = true; | |
1062 | info->range.max_proto.all = htons(nla_get_u16(a)); | |
1063 | info->range.flags |= NF_NAT_RANGE_PROTO_SPECIFIED; | |
1064 | break; | |
1065 | ||
1066 | case OVS_NAT_ATTR_PERSISTENT: | |
1067 | info->range.flags |= NF_NAT_RANGE_PERSISTENT; | |
1068 | break; | |
1069 | ||
1070 | case OVS_NAT_ATTR_PROTO_HASH: | |
1071 | info->range.flags |= NF_NAT_RANGE_PROTO_RANDOM; | |
1072 | break; | |
1073 | ||
1074 | case OVS_NAT_ATTR_PROTO_RANDOM: | |
1075 | info->range.flags |= NF_NAT_RANGE_PROTO_RANDOM_FULLY; | |
1076 | break; | |
1077 | ||
1078 | default: | |
1079 | OVS_NLERR(log, "Unknown nat attribute (%d).\n", type); | |
1080 | return -EINVAL; | |
1081 | } | |
1082 | } | |
1083 | ||
1084 | if (rem > 0) { | |
1085 | OVS_NLERR(log, "NAT attribute has %d unknown bytes.\n", rem); | |
1086 | return -EINVAL; | |
1087 | } | |
1088 | if (!info->nat) { | |
1089 | /* Do not allow flags if no type is given. */ | |
1090 | if (info->range.flags) { | |
1091 | OVS_NLERR(log, | |
1092 | "NAT flags may be given only when NAT range (SRC or DST) is also specified.\n" | |
1093 | ); | |
1094 | return -EINVAL; | |
1095 | } | |
1096 | info->nat = OVS_CT_NAT; /* NAT existing connections. */ | |
1097 | } else if (!info->commit) { | |
1098 | OVS_NLERR(log, | |
1099 | "NAT attributes may be specified only when CT COMMIT flag is also specified.\n" | |
1100 | ); | |
1101 | return -EINVAL; | |
1102 | } | |
1103 | /* Allow missing IP_MAX. */ | |
1104 | if (info->range.flags & NF_NAT_RANGE_MAP_IPS && !have_ip_max) { | |
1105 | memcpy(&info->range.max_addr, &info->range.min_addr, | |
1106 | sizeof(info->range.max_addr)); | |
1107 | } | |
1108 | /* Allow missing PROTO_MAX. */ | |
1109 | if (info->range.flags & NF_NAT_RANGE_PROTO_SPECIFIED && | |
1110 | !have_proto_max) { | |
1111 | info->range.max_proto.all = info->range.min_proto.all; | |
1112 | } | |
1113 | return 0; | |
1114 | } | |
1115 | #endif | |
1116 | ||
7f8a436e | 1117 | static const struct ovs_ct_len_tbl ovs_ct_attr_lens[OVS_CT_ATTR_MAX + 1] = { |
ab38a7b5 | 1118 | [OVS_CT_ATTR_COMMIT] = { .minlen = 0, .maxlen = 0 }, |
7f8a436e JS |
1119 | [OVS_CT_ATTR_ZONE] = { .minlen = sizeof(u16), |
1120 | .maxlen = sizeof(u16) }, | |
182e3042 JS |
1121 | [OVS_CT_ATTR_MARK] = { .minlen = sizeof(struct md_mark), |
1122 | .maxlen = sizeof(struct md_mark) }, | |
33db4125 JS |
1123 | [OVS_CT_ATTR_LABELS] = { .minlen = sizeof(struct md_labels), |
1124 | .maxlen = sizeof(struct md_labels) }, | |
cae3a262 | 1125 | [OVS_CT_ATTR_HELPER] = { .minlen = 1, |
05752523 JR |
1126 | .maxlen = NF_CT_HELPER_NAME_LEN }, |
1127 | #ifdef CONFIG_NF_NAT_NEEDED | |
1128 | /* NAT length is checked when parsing the nested attributes. */ | |
1129 | [OVS_CT_ATTR_NAT] = { .minlen = 0, .maxlen = INT_MAX }, | |
1130 | #endif | |
7f8a436e JS |
1131 | }; |
1132 | ||
1133 | static int parse_ct(const struct nlattr *attr, struct ovs_conntrack_info *info, | |
cae3a262 | 1134 | const char **helper, bool log) |
7f8a436e JS |
1135 | { |
1136 | struct nlattr *a; | |
1137 | int rem; | |
1138 | ||
1139 | nla_for_each_nested(a, attr, rem) { | |
1140 | int type = nla_type(a); | |
1141 | int maxlen = ovs_ct_attr_lens[type].maxlen; | |
1142 | int minlen = ovs_ct_attr_lens[type].minlen; | |
1143 | ||
1144 | if (type > OVS_CT_ATTR_MAX) { | |
1145 | OVS_NLERR(log, | |
1146 | "Unknown conntrack attr (type=%d, max=%d)", | |
1147 | type, OVS_CT_ATTR_MAX); | |
1148 | return -EINVAL; | |
1149 | } | |
1150 | if (nla_len(a) < minlen || nla_len(a) > maxlen) { | |
1151 | OVS_NLERR(log, | |
1152 | "Conntrack attr type has unexpected length (type=%d, length=%d, expected=%d)", | |
1153 | type, nla_len(a), maxlen); | |
1154 | return -EINVAL; | |
1155 | } | |
1156 | ||
1157 | switch (type) { | |
ab38a7b5 JS |
1158 | case OVS_CT_ATTR_COMMIT: |
1159 | info->commit = true; | |
7f8a436e JS |
1160 | break; |
1161 | #ifdef CONFIG_NF_CONNTRACK_ZONES | |
1162 | case OVS_CT_ATTR_ZONE: | |
1163 | info->zone.id = nla_get_u16(a); | |
1164 | break; | |
182e3042 JS |
1165 | #endif |
1166 | #ifdef CONFIG_NF_CONNTRACK_MARK | |
1167 | case OVS_CT_ATTR_MARK: { | |
1168 | struct md_mark *mark = nla_data(a); | |
1169 | ||
e754ec69 JS |
1170 | if (!mark->mask) { |
1171 | OVS_NLERR(log, "ct_mark mask cannot be 0"); | |
1172 | return -EINVAL; | |
1173 | } | |
182e3042 JS |
1174 | info->mark = *mark; |
1175 | break; | |
1176 | } | |
c2ac6673 JS |
1177 | #endif |
1178 | #ifdef CONFIG_NF_CONNTRACK_LABELS | |
33db4125 JS |
1179 | case OVS_CT_ATTR_LABELS: { |
1180 | struct md_labels *labels = nla_data(a); | |
c2ac6673 | 1181 | |
e754ec69 JS |
1182 | if (!labels_nonzero(&labels->mask)) { |
1183 | OVS_NLERR(log, "ct_labels mask cannot be 0"); | |
1184 | return -EINVAL; | |
1185 | } | |
33db4125 | 1186 | info->labels = *labels; |
c2ac6673 JS |
1187 | break; |
1188 | } | |
7f8a436e | 1189 | #endif |
cae3a262 JS |
1190 | case OVS_CT_ATTR_HELPER: |
1191 | *helper = nla_data(a); | |
1192 | if (!memchr(*helper, '\0', nla_len(a))) { | |
1193 | OVS_NLERR(log, "Invalid conntrack helper"); | |
1194 | return -EINVAL; | |
1195 | } | |
1196 | break; | |
05752523 JR |
1197 | #ifdef CONFIG_NF_NAT_NEEDED |
1198 | case OVS_CT_ATTR_NAT: { | |
1199 | int err = parse_nat(a, info, log); | |
1200 | ||
1201 | if (err) | |
1202 | return err; | |
1203 | break; | |
1204 | } | |
1205 | #endif | |
7f8a436e JS |
1206 | default: |
1207 | OVS_NLERR(log, "Unknown conntrack attr (%d)", | |
1208 | type); | |
1209 | return -EINVAL; | |
1210 | } | |
1211 | } | |
1212 | ||
7d904c7b JR |
1213 | #ifdef CONFIG_NF_CONNTRACK_MARK |
1214 | if (!info->commit && info->mark.mask) { | |
1215 | OVS_NLERR(log, | |
1216 | "Setting conntrack mark requires 'commit' flag."); | |
1217 | return -EINVAL; | |
1218 | } | |
1219 | #endif | |
1220 | #ifdef CONFIG_NF_CONNTRACK_LABELS | |
1221 | if (!info->commit && labels_nonzero(&info->labels.mask)) { | |
1222 | OVS_NLERR(log, | |
1223 | "Setting conntrack labels requires 'commit' flag."); | |
1224 | return -EINVAL; | |
1225 | } | |
1226 | #endif | |
7f8a436e JS |
1227 | if (rem > 0) { |
1228 | OVS_NLERR(log, "Conntrack attr has %d unknown bytes", rem); | |
1229 | return -EINVAL; | |
1230 | } | |
1231 | ||
1232 | return 0; | |
1233 | } | |
1234 | ||
c2ac6673 | 1235 | bool ovs_ct_verify(struct net *net, enum ovs_key_attr attr) |
7f8a436e JS |
1236 | { |
1237 | if (attr == OVS_KEY_ATTR_CT_STATE) | |
1238 | return true; | |
1239 | if (IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES) && | |
1240 | attr == OVS_KEY_ATTR_CT_ZONE) | |
1241 | return true; | |
182e3042 JS |
1242 | if (IS_ENABLED(CONFIG_NF_CONNTRACK_MARK) && |
1243 | attr == OVS_KEY_ATTR_CT_MARK) | |
1244 | return true; | |
c2ac6673 | 1245 | if (IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS) && |
33db4125 | 1246 | attr == OVS_KEY_ATTR_CT_LABELS) { |
c2ac6673 JS |
1247 | struct ovs_net *ovs_net = net_generic(net, ovs_net_id); |
1248 | ||
1249 | return ovs_net->xt_label; | |
1250 | } | |
7f8a436e JS |
1251 | |
1252 | return false; | |
1253 | } | |
1254 | ||
1255 | int ovs_ct_copy_action(struct net *net, const struct nlattr *attr, | |
1256 | const struct sw_flow_key *key, | |
1257 | struct sw_flow_actions **sfa, bool log) | |
1258 | { | |
1259 | struct ovs_conntrack_info ct_info; | |
cae3a262 | 1260 | const char *helper = NULL; |
7f8a436e JS |
1261 | u16 family; |
1262 | int err; | |
1263 | ||
1264 | family = key_to_nfproto(key); | |
1265 | if (family == NFPROTO_UNSPEC) { | |
1266 | OVS_NLERR(log, "ct family unspecified"); | |
1267 | return -EINVAL; | |
1268 | } | |
1269 | ||
1270 | memset(&ct_info, 0, sizeof(ct_info)); | |
1271 | ct_info.family = family; | |
1272 | ||
1273 | nf_ct_zone_init(&ct_info.zone, NF_CT_DEFAULT_ZONE_ID, | |
1274 | NF_CT_DEFAULT_ZONE_DIR, 0); | |
1275 | ||
cae3a262 | 1276 | err = parse_ct(attr, &ct_info, &helper, log); |
7f8a436e JS |
1277 | if (err) |
1278 | return err; | |
1279 | ||
1280 | /* Set up template for tracking connections in specific zones. */ | |
1281 | ct_info.ct = nf_ct_tmpl_alloc(net, &ct_info.zone, GFP_KERNEL); | |
1282 | if (!ct_info.ct) { | |
1283 | OVS_NLERR(log, "Failed to allocate conntrack template"); | |
1284 | return -ENOMEM; | |
1285 | } | |
90c7afc9 JS |
1286 | |
1287 | __set_bit(IPS_CONFIRMED_BIT, &ct_info.ct->status); | |
1288 | nf_conntrack_get(&ct_info.ct->ct_general); | |
1289 | ||
cae3a262 JS |
1290 | if (helper) { |
1291 | err = ovs_ct_add_helper(&ct_info, helper, key, log); | |
1292 | if (err) | |
1293 | goto err_free_ct; | |
1294 | } | |
7f8a436e JS |
1295 | |
1296 | err = ovs_nla_add_action(sfa, OVS_ACTION_ATTR_CT, &ct_info, | |
1297 | sizeof(ct_info), log); | |
1298 | if (err) | |
1299 | goto err_free_ct; | |
1300 | ||
7f8a436e JS |
1301 | return 0; |
1302 | err_free_ct: | |
2f3ab9f9 | 1303 | __ovs_ct_free_action(&ct_info); |
7f8a436e JS |
1304 | return err; |
1305 | } | |
1306 | ||
05752523 JR |
1307 | #ifdef CONFIG_NF_NAT_NEEDED |
1308 | static bool ovs_ct_nat_to_attr(const struct ovs_conntrack_info *info, | |
1309 | struct sk_buff *skb) | |
1310 | { | |
1311 | struct nlattr *start; | |
1312 | ||
1313 | start = nla_nest_start(skb, OVS_CT_ATTR_NAT); | |
1314 | if (!start) | |
1315 | return false; | |
1316 | ||
1317 | if (info->nat & OVS_CT_SRC_NAT) { | |
1318 | if (nla_put_flag(skb, OVS_NAT_ATTR_SRC)) | |
1319 | return false; | |
1320 | } else if (info->nat & OVS_CT_DST_NAT) { | |
1321 | if (nla_put_flag(skb, OVS_NAT_ATTR_DST)) | |
1322 | return false; | |
1323 | } else { | |
1324 | goto out; | |
1325 | } | |
1326 | ||
1327 | if (info->range.flags & NF_NAT_RANGE_MAP_IPS) { | |
99b7248e AB |
1328 | if (IS_ENABLED(CONFIG_NF_NAT_IPV4) && |
1329 | info->family == NFPROTO_IPV4) { | |
05752523 JR |
1330 | if (nla_put_in_addr(skb, OVS_NAT_ATTR_IP_MIN, |
1331 | info->range.min_addr.ip) || | |
1332 | (info->range.max_addr.ip | |
1333 | != info->range.min_addr.ip && | |
1334 | (nla_put_in_addr(skb, OVS_NAT_ATTR_IP_MAX, | |
1335 | info->range.max_addr.ip)))) | |
1336 | return false; | |
99b7248e AB |
1337 | } else if (IS_ENABLED(CONFIG_NF_NAT_IPV6) && |
1338 | info->family == NFPROTO_IPV6) { | |
05752523 JR |
1339 | if (nla_put_in6_addr(skb, OVS_NAT_ATTR_IP_MIN, |
1340 | &info->range.min_addr.in6) || | |
1341 | (memcmp(&info->range.max_addr.in6, | |
1342 | &info->range.min_addr.in6, | |
1343 | sizeof(info->range.max_addr.in6)) && | |
1344 | (nla_put_in6_addr(skb, OVS_NAT_ATTR_IP_MAX, | |
1345 | &info->range.max_addr.in6)))) | |
1346 | return false; | |
05752523 JR |
1347 | } else { |
1348 | return false; | |
1349 | } | |
1350 | } | |
1351 | if (info->range.flags & NF_NAT_RANGE_PROTO_SPECIFIED && | |
1352 | (nla_put_u16(skb, OVS_NAT_ATTR_PROTO_MIN, | |
1353 | ntohs(info->range.min_proto.all)) || | |
1354 | (info->range.max_proto.all != info->range.min_proto.all && | |
1355 | nla_put_u16(skb, OVS_NAT_ATTR_PROTO_MAX, | |
1356 | ntohs(info->range.max_proto.all))))) | |
1357 | return false; | |
1358 | ||
1359 | if (info->range.flags & NF_NAT_RANGE_PERSISTENT && | |
1360 | nla_put_flag(skb, OVS_NAT_ATTR_PERSISTENT)) | |
1361 | return false; | |
1362 | if (info->range.flags & NF_NAT_RANGE_PROTO_RANDOM && | |
1363 | nla_put_flag(skb, OVS_NAT_ATTR_PROTO_HASH)) | |
1364 | return false; | |
1365 | if (info->range.flags & NF_NAT_RANGE_PROTO_RANDOM_FULLY && | |
1366 | nla_put_flag(skb, OVS_NAT_ATTR_PROTO_RANDOM)) | |
1367 | return false; | |
1368 | out: | |
1369 | nla_nest_end(skb, start); | |
1370 | ||
1371 | return true; | |
1372 | } | |
1373 | #endif | |
1374 | ||
7f8a436e JS |
1375 | int ovs_ct_action_to_attr(const struct ovs_conntrack_info *ct_info, |
1376 | struct sk_buff *skb) | |
1377 | { | |
1378 | struct nlattr *start; | |
1379 | ||
1380 | start = nla_nest_start(skb, OVS_ACTION_ATTR_CT); | |
1381 | if (!start) | |
1382 | return -EMSGSIZE; | |
1383 | ||
ab38a7b5 | 1384 | if (ct_info->commit && nla_put_flag(skb, OVS_CT_ATTR_COMMIT)) |
7f8a436e JS |
1385 | return -EMSGSIZE; |
1386 | if (IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES) && | |
1387 | nla_put_u16(skb, OVS_CT_ATTR_ZONE, ct_info->zone.id)) | |
1388 | return -EMSGSIZE; | |
e754ec69 | 1389 | if (IS_ENABLED(CONFIG_NF_CONNTRACK_MARK) && ct_info->mark.mask && |
182e3042 JS |
1390 | nla_put(skb, OVS_CT_ATTR_MARK, sizeof(ct_info->mark), |
1391 | &ct_info->mark)) | |
1392 | return -EMSGSIZE; | |
c2ac6673 | 1393 | if (IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS) && |
e754ec69 | 1394 | labels_nonzero(&ct_info->labels.mask) && |
33db4125 JS |
1395 | nla_put(skb, OVS_CT_ATTR_LABELS, sizeof(ct_info->labels), |
1396 | &ct_info->labels)) | |
c2ac6673 | 1397 | return -EMSGSIZE; |
cae3a262 JS |
1398 | if (ct_info->helper) { |
1399 | if (nla_put_string(skb, OVS_CT_ATTR_HELPER, | |
1400 | ct_info->helper->name)) | |
1401 | return -EMSGSIZE; | |
1402 | } | |
05752523 JR |
1403 | #ifdef CONFIG_NF_NAT_NEEDED |
1404 | if (ct_info->nat && !ovs_ct_nat_to_attr(ct_info, skb)) | |
1405 | return -EMSGSIZE; | |
1406 | #endif | |
7f8a436e JS |
1407 | nla_nest_end(skb, start); |
1408 | ||
1409 | return 0; | |
1410 | } | |
1411 | ||
1412 | void ovs_ct_free_action(const struct nlattr *a) | |
1413 | { | |
1414 | struct ovs_conntrack_info *ct_info = nla_data(a); | |
1415 | ||
2f3ab9f9 JS |
1416 | __ovs_ct_free_action(ct_info); |
1417 | } | |
1418 | ||
1419 | static void __ovs_ct_free_action(struct ovs_conntrack_info *ct_info) | |
1420 | { | |
cae3a262 JS |
1421 | if (ct_info->helper) |
1422 | module_put(ct_info->helper->me); | |
7f8a436e | 1423 | if (ct_info->ct) |
76644232 | 1424 | nf_ct_tmpl_free(ct_info->ct); |
7f8a436e | 1425 | } |
c2ac6673 JS |
1426 | |
1427 | void ovs_ct_init(struct net *net) | |
1428 | { | |
33db4125 | 1429 | unsigned int n_bits = sizeof(struct ovs_key_ct_labels) * BITS_PER_BYTE; |
c2ac6673 JS |
1430 | struct ovs_net *ovs_net = net_generic(net, ovs_net_id); |
1431 | ||
adff6c65 | 1432 | if (nf_connlabels_get(net, n_bits - 1)) { |
c2ac6673 JS |
1433 | ovs_net->xt_label = false; |
1434 | OVS_NLERR(true, "Failed to set connlabel length"); | |
1435 | } else { | |
1436 | ovs_net->xt_label = true; | |
1437 | } | |
1438 | } | |
1439 | ||
1440 | void ovs_ct_exit(struct net *net) | |
1441 | { | |
1442 | struct ovs_net *ovs_net = net_generic(net, ovs_net_id); | |
1443 | ||
1444 | if (ovs_net->xt_label) | |
1445 | nf_connlabels_put(net); | |
1446 | } |