]>
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 | ||
33db4125 JS |
132 | static void ovs_ct_get_labels(const struct nf_conn *ct, |
133 | struct ovs_key_ct_labels *labels) | |
c2ac6673 JS |
134 | { |
135 | struct nf_conn_labels *cl = ct ? nf_ct_labels_find(ct) : NULL; | |
136 | ||
137 | if (cl) { | |
23014011 | 138 | size_t len = sizeof(cl->bits); |
c2ac6673 | 139 | |
33db4125 JS |
140 | if (len > OVS_CT_LABELS_LEN) |
141 | len = OVS_CT_LABELS_LEN; | |
142 | else if (len < OVS_CT_LABELS_LEN) | |
143 | memset(labels, 0, OVS_CT_LABELS_LEN); | |
144 | memcpy(labels, cl->bits, len); | |
c2ac6673 | 145 | } else { |
33db4125 | 146 | memset(labels, 0, OVS_CT_LABELS_LEN); |
c2ac6673 JS |
147 | } |
148 | } | |
149 | ||
7f8a436e | 150 | static void __ovs_ct_update_key(struct sw_flow_key *key, u8 state, |
182e3042 JS |
151 | const struct nf_conntrack_zone *zone, |
152 | const struct nf_conn *ct) | |
7f8a436e JS |
153 | { |
154 | key->ct.state = state; | |
155 | key->ct.zone = zone->id; | |
0d5cdef8 | 156 | key->ct.mark = ovs_ct_get_mark(ct); |
33db4125 | 157 | ovs_ct_get_labels(ct, &key->ct.labels); |
7f8a436e JS |
158 | } |
159 | ||
05752523 JR |
160 | /* Update 'key' based on skb->nfct. If 'post_ct' is true, then OVS has |
161 | * previously sent the packet to conntrack via the ct action. If | |
162 | * 'keep_nat_flags' is true, the existing NAT flags retained, else they are | |
163 | * initialized from the connection status. | |
7f8a436e JS |
164 | */ |
165 | static void ovs_ct_update_key(const struct sk_buff *skb, | |
d110986c | 166 | const struct ovs_conntrack_info *info, |
05752523 JR |
167 | struct sw_flow_key *key, bool post_ct, |
168 | bool keep_nat_flags) | |
7f8a436e JS |
169 | { |
170 | const struct nf_conntrack_zone *zone = &nf_ct_zone_dflt; | |
171 | enum ip_conntrack_info ctinfo; | |
172 | struct nf_conn *ct; | |
173 | u8 state = 0; | |
174 | ||
175 | ct = nf_ct_get(skb, &ctinfo); | |
176 | if (ct) { | |
177 | state = ovs_ct_get_state(ctinfo); | |
9f13ded8 | 178 | /* All unconfirmed entries are NEW connections. */ |
4f0909ee JS |
179 | if (!nf_ct_is_confirmed(ct)) |
180 | state |= OVS_CS_F_NEW; | |
9f13ded8 JR |
181 | /* OVS persists the related flag for the duration of the |
182 | * connection. | |
183 | */ | |
7f8a436e JS |
184 | if (ct->master) |
185 | state |= OVS_CS_F_RELATED; | |
05752523 JR |
186 | if (keep_nat_flags) { |
187 | state |= key->ct.state & OVS_CS_F_NAT_MASK; | |
188 | } else { | |
189 | if (ct->status & IPS_SRC_NAT) | |
190 | state |= OVS_CS_F_SRC_NAT; | |
191 | if (ct->status & IPS_DST_NAT) | |
192 | state |= OVS_CS_F_DST_NAT; | |
193 | } | |
7f8a436e JS |
194 | zone = nf_ct_zone(ct); |
195 | } else if (post_ct) { | |
196 | state = OVS_CS_F_TRACKED | OVS_CS_F_INVALID; | |
d110986c JS |
197 | if (info) |
198 | zone = &info->zone; | |
7f8a436e | 199 | } |
182e3042 | 200 | __ovs_ct_update_key(key, state, zone, ct); |
7f8a436e JS |
201 | } |
202 | ||
9f13ded8 JR |
203 | /* This is called to initialize CT key fields possibly coming in from the local |
204 | * stack. | |
205 | */ | |
7f8a436e JS |
206 | void ovs_ct_fill_key(const struct sk_buff *skb, struct sw_flow_key *key) |
207 | { | |
05752523 | 208 | ovs_ct_update_key(skb, NULL, key, false, false); |
7f8a436e JS |
209 | } |
210 | ||
211 | int ovs_ct_put_key(const struct sw_flow_key *key, struct sk_buff *skb) | |
212 | { | |
fbccce59 | 213 | if (nla_put_u32(skb, OVS_KEY_ATTR_CT_STATE, key->ct.state)) |
7f8a436e JS |
214 | return -EMSGSIZE; |
215 | ||
216 | if (IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES) && | |
217 | nla_put_u16(skb, OVS_KEY_ATTR_CT_ZONE, key->ct.zone)) | |
218 | return -EMSGSIZE; | |
219 | ||
182e3042 JS |
220 | if (IS_ENABLED(CONFIG_NF_CONNTRACK_MARK) && |
221 | nla_put_u32(skb, OVS_KEY_ATTR_CT_MARK, key->ct.mark)) | |
222 | return -EMSGSIZE; | |
223 | ||
9723e6ab | 224 | if (IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS) && |
33db4125 JS |
225 | nla_put(skb, OVS_KEY_ATTR_CT_LABELS, sizeof(key->ct.labels), |
226 | &key->ct.labels)) | |
c2ac6673 JS |
227 | return -EMSGSIZE; |
228 | ||
182e3042 JS |
229 | return 0; |
230 | } | |
231 | ||
232 | static int ovs_ct_set_mark(struct sk_buff *skb, struct sw_flow_key *key, | |
233 | u32 ct_mark, u32 mask) | |
234 | { | |
0d5cdef8 | 235 | #if IS_ENABLED(CONFIG_NF_CONNTRACK_MARK) |
182e3042 JS |
236 | enum ip_conntrack_info ctinfo; |
237 | struct nf_conn *ct; | |
238 | u32 new_mark; | |
239 | ||
182e3042 JS |
240 | /* The connection could be invalid, in which case set_mark is no-op. */ |
241 | ct = nf_ct_get(skb, &ctinfo); | |
242 | if (!ct) | |
243 | return 0; | |
244 | ||
245 | new_mark = ct_mark | (ct->mark & ~(mask)); | |
246 | if (ct->mark != new_mark) { | |
247 | ct->mark = new_mark; | |
248 | nf_conntrack_event_cache(IPCT_MARK, ct); | |
249 | key->ct.mark = new_mark; | |
250 | } | |
251 | ||
7f8a436e | 252 | return 0; |
0d5cdef8 JS |
253 | #else |
254 | return -ENOTSUPP; | |
255 | #endif | |
7f8a436e JS |
256 | } |
257 | ||
33db4125 JS |
258 | static int ovs_ct_set_labels(struct sk_buff *skb, struct sw_flow_key *key, |
259 | const struct ovs_key_ct_labels *labels, | |
260 | const struct ovs_key_ct_labels *mask) | |
c2ac6673 JS |
261 | { |
262 | enum ip_conntrack_info ctinfo; | |
263 | struct nf_conn_labels *cl; | |
264 | struct nf_conn *ct; | |
265 | int err; | |
266 | ||
c2ac6673 JS |
267 | /* The connection could be invalid, in which case set_label is no-op.*/ |
268 | ct = nf_ct_get(skb, &ctinfo); | |
269 | if (!ct) | |
270 | return 0; | |
271 | ||
272 | cl = nf_ct_labels_find(ct); | |
273 | if (!cl) { | |
274 | nf_ct_labels_ext_add(ct); | |
275 | cl = nf_ct_labels_find(ct); | |
276 | } | |
23014011 | 277 | if (!cl || sizeof(cl->bits) < OVS_CT_LABELS_LEN) |
c2ac6673 JS |
278 | return -ENOSPC; |
279 | ||
33db4125 JS |
280 | err = nf_connlabels_replace(ct, (u32 *)labels, (u32 *)mask, |
281 | OVS_CT_LABELS_LEN / sizeof(u32)); | |
c2ac6673 JS |
282 | if (err) |
283 | return err; | |
284 | ||
33db4125 | 285 | ovs_ct_get_labels(ct, &key->ct.labels); |
c2ac6673 JS |
286 | return 0; |
287 | } | |
288 | ||
cae3a262 JS |
289 | /* 'skb' should already be pulled to nh_ofs. */ |
290 | static int ovs_ct_helper(struct sk_buff *skb, u16 proto) | |
291 | { | |
292 | const struct nf_conntrack_helper *helper; | |
293 | const struct nf_conn_help *help; | |
294 | enum ip_conntrack_info ctinfo; | |
295 | unsigned int protoff; | |
296 | struct nf_conn *ct; | |
05752523 | 297 | int err; |
cae3a262 JS |
298 | |
299 | ct = nf_ct_get(skb, &ctinfo); | |
300 | if (!ct || ctinfo == IP_CT_RELATED_REPLY) | |
301 | return NF_ACCEPT; | |
302 | ||
303 | help = nfct_help(ct); | |
304 | if (!help) | |
305 | return NF_ACCEPT; | |
306 | ||
307 | helper = rcu_dereference(help->helper); | |
308 | if (!helper) | |
309 | return NF_ACCEPT; | |
310 | ||
311 | switch (proto) { | |
312 | case NFPROTO_IPV4: | |
313 | protoff = ip_hdrlen(skb); | |
314 | break; | |
315 | case NFPROTO_IPV6: { | |
316 | u8 nexthdr = ipv6_hdr(skb)->nexthdr; | |
317 | __be16 frag_off; | |
cc570605 | 318 | int ofs; |
cae3a262 | 319 | |
cc570605 JS |
320 | ofs = ipv6_skip_exthdr(skb, sizeof(struct ipv6hdr), &nexthdr, |
321 | &frag_off); | |
322 | if (ofs < 0 || (frag_off & htons(~0x7)) != 0) { | |
cae3a262 JS |
323 | pr_debug("proto header not found\n"); |
324 | return NF_ACCEPT; | |
325 | } | |
cc570605 | 326 | protoff = ofs; |
cae3a262 JS |
327 | break; |
328 | } | |
329 | default: | |
330 | WARN_ONCE(1, "helper invoked on non-IP family!"); | |
331 | return NF_DROP; | |
332 | } | |
333 | ||
05752523 JR |
334 | err = helper->help(skb, protoff, ct, ctinfo); |
335 | if (err != NF_ACCEPT) | |
336 | return err; | |
337 | ||
338 | /* Adjust seqs after helper. This is needed due to some helpers (e.g., | |
339 | * FTP with NAT) adusting the TCP payload size when mangling IP | |
340 | * addresses and/or port numbers in the text-based control connection. | |
341 | */ | |
342 | if (test_bit(IPS_SEQ_ADJUST_BIT, &ct->status) && | |
343 | !nf_ct_seq_adjust(skb, ct, ctinfo, protoff)) | |
344 | return NF_DROP; | |
345 | return NF_ACCEPT; | |
cae3a262 JS |
346 | } |
347 | ||
74c16618 JS |
348 | /* Returns 0 on success, -EINPROGRESS if 'skb' is stolen, or other nonzero |
349 | * value if 'skb' is freed. | |
350 | */ | |
7f8a436e JS |
351 | static int handle_fragments(struct net *net, struct sw_flow_key *key, |
352 | u16 zone, struct sk_buff *skb) | |
353 | { | |
354 | struct ovs_skb_cb ovs_cb = *OVS_CB(skb); | |
daaa7d64 | 355 | int err; |
7f8a436e JS |
356 | |
357 | if (key->eth.type == htons(ETH_P_IP)) { | |
358 | enum ip_defrag_users user = IP_DEFRAG_CONNTRACK_IN + zone; | |
7f8a436e JS |
359 | |
360 | memset(IPCB(skb), 0, sizeof(struct inet_skb_parm)); | |
19bcf9f2 | 361 | err = ip_defrag(net, skb, user); |
7f8a436e JS |
362 | if (err) |
363 | return err; | |
364 | ||
365 | ovs_cb.mru = IPCB(skb)->frag_max_size; | |
7f8a436e | 366 | #if IS_ENABLED(CONFIG_NF_DEFRAG_IPV6) |
74c16618 | 367 | } else if (key->eth.type == htons(ETH_P_IPV6)) { |
7f8a436e | 368 | enum ip6_defrag_users user = IP6_DEFRAG_CONNTRACK_IN + zone; |
7f8a436e JS |
369 | |
370 | memset(IP6CB(skb), 0, sizeof(struct inet6_skb_parm)); | |
daaa7d64 | 371 | err = nf_ct_frag6_gather(net, skb, user); |
f92a80a9 DDP |
372 | if (err) { |
373 | if (err != -EINPROGRESS) | |
374 | kfree_skb(skb); | |
daaa7d64 | 375 | return err; |
f92a80a9 | 376 | } |
7f8a436e | 377 | |
daaa7d64 | 378 | key->ip.proto = ipv6_hdr(skb)->nexthdr; |
7f8a436e | 379 | ovs_cb.mru = IP6CB(skb)->frag_max_size; |
7f8a436e JS |
380 | #endif |
381 | } else { | |
74c16618 | 382 | kfree_skb(skb); |
7f8a436e JS |
383 | return -EPFNOSUPPORT; |
384 | } | |
385 | ||
386 | key->ip.frag = OVS_FRAG_TYPE_NONE; | |
387 | skb_clear_hash(skb); | |
388 | skb->ignore_df = 1; | |
389 | *OVS_CB(skb) = ovs_cb; | |
390 | ||
391 | return 0; | |
392 | } | |
393 | ||
394 | static struct nf_conntrack_expect * | |
395 | ovs_ct_expect_find(struct net *net, const struct nf_conntrack_zone *zone, | |
396 | u16 proto, const struct sk_buff *skb) | |
397 | { | |
398 | struct nf_conntrack_tuple tuple; | |
399 | ||
a31f1adc | 400 | if (!nf_ct_get_tuplepr(skb, skb_network_offset(skb), proto, net, &tuple)) |
7f8a436e JS |
401 | return NULL; |
402 | return __nf_ct_expect_find(net, zone, &tuple); | |
403 | } | |
404 | ||
289f2253 JR |
405 | /* This replicates logic from nf_conntrack_core.c that is not exported. */ |
406 | static enum ip_conntrack_info | |
407 | ovs_ct_get_info(const struct nf_conntrack_tuple_hash *h) | |
408 | { | |
409 | const struct nf_conn *ct = nf_ct_tuplehash_to_ctrack(h); | |
410 | ||
411 | if (NF_CT_DIRECTION(h) == IP_CT_DIR_REPLY) | |
412 | return IP_CT_ESTABLISHED_REPLY; | |
413 | /* Once we've had two way comms, always ESTABLISHED. */ | |
414 | if (test_bit(IPS_SEEN_REPLY_BIT, &ct->status)) | |
415 | return IP_CT_ESTABLISHED; | |
416 | if (test_bit(IPS_EXPECTED_BIT, &ct->status)) | |
417 | return IP_CT_RELATED; | |
418 | return IP_CT_NEW; | |
419 | } | |
420 | ||
421 | /* Find an existing connection which this packet belongs to without | |
422 | * re-attributing statistics or modifying the connection state. This allows an | |
423 | * skb->nfct lost due to an upcall to be recovered during actions execution. | |
424 | * | |
425 | * Must be called with rcu_read_lock. | |
426 | * | |
427 | * On success, populates skb->nfct and skb->nfctinfo, and returns the | |
428 | * connection. Returns NULL if there is no existing entry. | |
429 | */ | |
430 | static struct nf_conn * | |
431 | ovs_ct_find_existing(struct net *net, const struct nf_conntrack_zone *zone, | |
432 | u8 l3num, struct sk_buff *skb) | |
433 | { | |
434 | struct nf_conntrack_l3proto *l3proto; | |
435 | struct nf_conntrack_l4proto *l4proto; | |
436 | struct nf_conntrack_tuple tuple; | |
437 | struct nf_conntrack_tuple_hash *h; | |
289f2253 JR |
438 | struct nf_conn *ct; |
439 | unsigned int dataoff; | |
440 | u8 protonum; | |
441 | ||
442 | l3proto = __nf_ct_l3proto_find(l3num); | |
289f2253 JR |
443 | if (l3proto->get_l4proto(skb, skb_network_offset(skb), &dataoff, |
444 | &protonum) <= 0) { | |
445 | pr_debug("ovs_ct_find_existing: Can't get protonum\n"); | |
446 | return NULL; | |
447 | } | |
448 | l4proto = __nf_ct_l4proto_find(l3num, protonum); | |
289f2253 JR |
449 | if (!nf_ct_get_tuple(skb, skb_network_offset(skb), dataoff, l3num, |
450 | protonum, net, &tuple, l3proto, l4proto)) { | |
451 | pr_debug("ovs_ct_find_existing: Can't get tuple\n"); | |
452 | return NULL; | |
453 | } | |
454 | ||
455 | /* look for tuple match */ | |
456 | h = nf_conntrack_find_get(net, zone, &tuple); | |
457 | if (!h) | |
458 | return NULL; /* Not found. */ | |
459 | ||
460 | ct = nf_ct_tuplehash_to_ctrack(h); | |
461 | ||
289f2253 | 462 | skb->nfct = &ct->ct_general; |
bce91f8a | 463 | skb->nfctinfo = ovs_ct_get_info(h); |
289f2253 JR |
464 | return ct; |
465 | } | |
466 | ||
7f8a436e | 467 | /* Determine whether skb->nfct is equal to the result of conntrack lookup. */ |
289f2253 JR |
468 | static bool skb_nfct_cached(struct net *net, |
469 | const struct sw_flow_key *key, | |
470 | const struct ovs_conntrack_info *info, | |
471 | struct sk_buff *skb) | |
7f8a436e JS |
472 | { |
473 | enum ip_conntrack_info ctinfo; | |
474 | struct nf_conn *ct; | |
475 | ||
476 | ct = nf_ct_get(skb, &ctinfo); | |
289f2253 JR |
477 | /* If no ct, check if we have evidence that an existing conntrack entry |
478 | * might be found for this skb. This happens when we lose a skb->nfct | |
479 | * due to an upcall. If the connection was not confirmed, it is not | |
480 | * cached and needs to be run through conntrack again. | |
481 | */ | |
482 | if (!ct && key->ct.state & OVS_CS_F_TRACKED && | |
483 | !(key->ct.state & OVS_CS_F_INVALID) && | |
484 | key->ct.zone == info->zone.id) | |
485 | ct = ovs_ct_find_existing(net, &info->zone, info->family, skb); | |
7f8a436e JS |
486 | if (!ct) |
487 | return false; | |
488 | if (!net_eq(net, read_pnet(&ct->ct_net))) | |
489 | return false; | |
490 | if (!nf_ct_zone_equal_any(info->ct, nf_ct_zone(ct))) | |
491 | return false; | |
cae3a262 JS |
492 | if (info->helper) { |
493 | struct nf_conn_help *help; | |
494 | ||
495 | help = nf_ct_ext_find(ct, NF_CT_EXT_HELPER); | |
496 | if (help && rcu_access_pointer(help->helper) != info->helper) | |
497 | return false; | |
498 | } | |
7f8a436e JS |
499 | |
500 | return true; | |
501 | } | |
502 | ||
05752523 JR |
503 | #ifdef CONFIG_NF_NAT_NEEDED |
504 | /* Modelled after nf_nat_ipv[46]_fn(). | |
505 | * range is only used for new, uninitialized NAT state. | |
506 | * Returns either NF_ACCEPT or NF_DROP. | |
507 | */ | |
508 | static int ovs_ct_nat_execute(struct sk_buff *skb, struct nf_conn *ct, | |
509 | enum ip_conntrack_info ctinfo, | |
510 | const struct nf_nat_range *range, | |
511 | enum nf_nat_manip_type maniptype) | |
512 | { | |
513 | int hooknum, nh_off, err = NF_ACCEPT; | |
514 | ||
515 | nh_off = skb_network_offset(skb); | |
75f01a4c | 516 | skb_pull_rcsum(skb, nh_off); |
05752523 JR |
517 | |
518 | /* See HOOK2MANIP(). */ | |
519 | if (maniptype == NF_NAT_MANIP_SRC) | |
520 | hooknum = NF_INET_LOCAL_IN; /* Source NAT */ | |
521 | else | |
522 | hooknum = NF_INET_LOCAL_OUT; /* Destination NAT */ | |
523 | ||
524 | switch (ctinfo) { | |
525 | case IP_CT_RELATED: | |
526 | case IP_CT_RELATED_REPLY: | |
99b7248e AB |
527 | if (IS_ENABLED(CONFIG_NF_NAT_IPV4) && |
528 | skb->protocol == htons(ETH_P_IP) && | |
05752523 JR |
529 | ip_hdr(skb)->protocol == IPPROTO_ICMP) { |
530 | if (!nf_nat_icmp_reply_translation(skb, ct, ctinfo, | |
531 | hooknum)) | |
532 | err = NF_DROP; | |
533 | goto push; | |
99b7248e AB |
534 | } else if (IS_ENABLED(CONFIG_NF_NAT_IPV6) && |
535 | skb->protocol == htons(ETH_P_IPV6)) { | |
05752523 JR |
536 | __be16 frag_off; |
537 | u8 nexthdr = ipv6_hdr(skb)->nexthdr; | |
538 | int hdrlen = ipv6_skip_exthdr(skb, | |
539 | sizeof(struct ipv6hdr), | |
540 | &nexthdr, &frag_off); | |
541 | ||
542 | if (hdrlen >= 0 && nexthdr == IPPROTO_ICMPV6) { | |
543 | if (!nf_nat_icmpv6_reply_translation(skb, ct, | |
544 | ctinfo, | |
545 | hooknum, | |
546 | hdrlen)) | |
547 | err = NF_DROP; | |
548 | goto push; | |
549 | } | |
05752523 JR |
550 | } |
551 | /* Non-ICMP, fall thru to initialize if needed. */ | |
552 | case IP_CT_NEW: | |
553 | /* Seen it before? This can happen for loopback, retrans, | |
554 | * or local packets. | |
555 | */ | |
556 | if (!nf_nat_initialized(ct, maniptype)) { | |
557 | /* Initialize according to the NAT action. */ | |
558 | err = (range && range->flags & NF_NAT_RANGE_MAP_IPS) | |
559 | /* Action is set up to establish a new | |
560 | * mapping. | |
561 | */ | |
562 | ? nf_nat_setup_info(ct, range, maniptype) | |
563 | : nf_nat_alloc_null_binding(ct, hooknum); | |
564 | if (err != NF_ACCEPT) | |
565 | goto push; | |
566 | } | |
567 | break; | |
568 | ||
569 | case IP_CT_ESTABLISHED: | |
570 | case IP_CT_ESTABLISHED_REPLY: | |
571 | break; | |
572 | ||
573 | default: | |
574 | err = NF_DROP; | |
575 | goto push; | |
576 | } | |
577 | ||
578 | err = nf_nat_packet(ct, ctinfo, hooknum, skb); | |
579 | push: | |
580 | skb_push(skb, nh_off); | |
75f01a4c | 581 | skb_postpush_rcsum(skb, skb->data, nh_off); |
05752523 JR |
582 | |
583 | return err; | |
584 | } | |
585 | ||
586 | static void ovs_nat_update_key(struct sw_flow_key *key, | |
587 | const struct sk_buff *skb, | |
588 | enum nf_nat_manip_type maniptype) | |
589 | { | |
590 | if (maniptype == NF_NAT_MANIP_SRC) { | |
591 | __be16 src; | |
592 | ||
593 | key->ct.state |= OVS_CS_F_SRC_NAT; | |
594 | if (key->eth.type == htons(ETH_P_IP)) | |
595 | key->ipv4.addr.src = ip_hdr(skb)->saddr; | |
596 | else if (key->eth.type == htons(ETH_P_IPV6)) | |
597 | memcpy(&key->ipv6.addr.src, &ipv6_hdr(skb)->saddr, | |
598 | sizeof(key->ipv6.addr.src)); | |
599 | else | |
600 | return; | |
601 | ||
602 | if (key->ip.proto == IPPROTO_UDP) | |
603 | src = udp_hdr(skb)->source; | |
604 | else if (key->ip.proto == IPPROTO_TCP) | |
605 | src = tcp_hdr(skb)->source; | |
606 | else if (key->ip.proto == IPPROTO_SCTP) | |
607 | src = sctp_hdr(skb)->source; | |
608 | else | |
609 | return; | |
610 | ||
611 | key->tp.src = src; | |
612 | } else { | |
613 | __be16 dst; | |
614 | ||
615 | key->ct.state |= OVS_CS_F_DST_NAT; | |
616 | if (key->eth.type == htons(ETH_P_IP)) | |
617 | key->ipv4.addr.dst = ip_hdr(skb)->daddr; | |
618 | else if (key->eth.type == htons(ETH_P_IPV6)) | |
619 | memcpy(&key->ipv6.addr.dst, &ipv6_hdr(skb)->daddr, | |
620 | sizeof(key->ipv6.addr.dst)); | |
621 | else | |
622 | return; | |
623 | ||
624 | if (key->ip.proto == IPPROTO_UDP) | |
625 | dst = udp_hdr(skb)->dest; | |
626 | else if (key->ip.proto == IPPROTO_TCP) | |
627 | dst = tcp_hdr(skb)->dest; | |
628 | else if (key->ip.proto == IPPROTO_SCTP) | |
629 | dst = sctp_hdr(skb)->dest; | |
630 | else | |
631 | return; | |
632 | ||
633 | key->tp.dst = dst; | |
634 | } | |
635 | } | |
636 | ||
637 | /* Returns NF_DROP if the packet should be dropped, NF_ACCEPT otherwise. */ | |
638 | static int ovs_ct_nat(struct net *net, struct sw_flow_key *key, | |
639 | const struct ovs_conntrack_info *info, | |
640 | struct sk_buff *skb, struct nf_conn *ct, | |
641 | enum ip_conntrack_info ctinfo) | |
642 | { | |
643 | enum nf_nat_manip_type maniptype; | |
644 | int err; | |
645 | ||
646 | if (nf_ct_is_untracked(ct)) { | |
647 | /* A NAT action may only be performed on tracked packets. */ | |
648 | return NF_ACCEPT; | |
649 | } | |
650 | ||
651 | /* Add NAT extension if not confirmed yet. */ | |
652 | if (!nf_ct_is_confirmed(ct) && !nf_ct_nat_ext_add(ct)) | |
653 | return NF_ACCEPT; /* Can't NAT. */ | |
654 | ||
655 | /* Determine NAT type. | |
656 | * Check if the NAT type can be deduced from the tracked connection. | |
5745b0be JR |
657 | * Make sure new expected connections (IP_CT_RELATED) are NATted only |
658 | * when committing. | |
05752523 JR |
659 | */ |
660 | if (info->nat & OVS_CT_NAT && ctinfo != IP_CT_NEW && | |
661 | ct->status & IPS_NAT_MASK && | |
5745b0be | 662 | (ctinfo != IP_CT_RELATED || info->commit)) { |
05752523 JR |
663 | /* NAT an established or related connection like before. */ |
664 | if (CTINFO2DIR(ctinfo) == IP_CT_DIR_REPLY) | |
665 | /* This is the REPLY direction for a connection | |
666 | * for which NAT was applied in the forward | |
667 | * direction. Do the reverse NAT. | |
668 | */ | |
669 | maniptype = ct->status & IPS_SRC_NAT | |
670 | ? NF_NAT_MANIP_DST : NF_NAT_MANIP_SRC; | |
671 | else | |
672 | maniptype = ct->status & IPS_SRC_NAT | |
673 | ? NF_NAT_MANIP_SRC : NF_NAT_MANIP_DST; | |
674 | } else if (info->nat & OVS_CT_SRC_NAT) { | |
675 | maniptype = NF_NAT_MANIP_SRC; | |
676 | } else if (info->nat & OVS_CT_DST_NAT) { | |
677 | maniptype = NF_NAT_MANIP_DST; | |
678 | } else { | |
679 | return NF_ACCEPT; /* Connection is not NATed. */ | |
680 | } | |
681 | err = ovs_ct_nat_execute(skb, ct, ctinfo, &info->range, maniptype); | |
682 | ||
683 | /* Mark NAT done if successful and update the flow key. */ | |
684 | if (err == NF_ACCEPT) | |
685 | ovs_nat_update_key(key, skb, maniptype); | |
686 | ||
687 | return err; | |
688 | } | |
689 | #else /* !CONFIG_NF_NAT_NEEDED */ | |
690 | static int ovs_ct_nat(struct net *net, struct sw_flow_key *key, | |
691 | const struct ovs_conntrack_info *info, | |
692 | struct sk_buff *skb, struct nf_conn *ct, | |
693 | enum ip_conntrack_info ctinfo) | |
694 | { | |
695 | return NF_ACCEPT; | |
696 | } | |
697 | #endif | |
698 | ||
9f13ded8 | 699 | /* Pass 'skb' through conntrack in 'net', using zone configured in 'info', if |
394e910e JR |
700 | * not done already. Update key with new CT state after passing the packet |
701 | * through conntrack. | |
9f13ded8 JR |
702 | * Note that if the packet is deemed invalid by conntrack, skb->nfct will be |
703 | * set to NULL and 0 will be returned. | |
704 | */ | |
4f0909ee | 705 | static int __ovs_ct_lookup(struct net *net, struct sw_flow_key *key, |
7f8a436e JS |
706 | const struct ovs_conntrack_info *info, |
707 | struct sk_buff *skb) | |
708 | { | |
709 | /* If we are recirculating packets to match on conntrack fields and | |
710 | * committing with a separate conntrack action, then we don't need to | |
711 | * actually run the packet through conntrack twice unless it's for a | |
712 | * different zone. | |
713 | */ | |
28b6e0c1 JR |
714 | bool cached = skb_nfct_cached(net, key, info, skb); |
715 | enum ip_conntrack_info ctinfo; | |
716 | struct nf_conn *ct; | |
717 | ||
718 | if (!cached) { | |
7f8a436e | 719 | struct nf_conn *tmpl = info->ct; |
5b6b9293 | 720 | int err; |
7f8a436e JS |
721 | |
722 | /* Associate skb with specified zone. */ | |
723 | if (tmpl) { | |
724 | if (skb->nfct) | |
725 | nf_conntrack_put(skb->nfct); | |
726 | nf_conntrack_get(&tmpl->ct_general); | |
727 | skb->nfct = &tmpl->ct_general; | |
728 | skb->nfctinfo = IP_CT_NEW; | |
729 | } | |
730 | ||
08733a0c PNA |
731 | err = nf_conntrack_in(net, info->family, |
732 | NF_INET_PRE_ROUTING, skb); | |
5b6b9293 | 733 | if (err != NF_ACCEPT) |
7f8a436e | 734 | return -ENOENT; |
cae3a262 | 735 | |
05752523 JR |
736 | /* Clear CT state NAT flags to mark that we have not yet done |
737 | * NAT after the nf_conntrack_in() call. We can actually clear | |
738 | * the whole state, as it will be re-initialized below. | |
739 | */ | |
740 | key->ct.state = 0; | |
741 | ||
742 | /* Update the key, but keep the NAT flags. */ | |
743 | ovs_ct_update_key(skb, info, key, true, true); | |
28b6e0c1 | 744 | } |
394e910e | 745 | |
28b6e0c1 | 746 | ct = nf_ct_get(skb, &ctinfo); |
05752523 JR |
747 | if (ct) { |
748 | /* Packets starting a new connection must be NATted before the | |
749 | * helper, so that the helper knows about the NAT. We enforce | |
750 | * this by delaying both NAT and helper calls for unconfirmed | |
751 | * connections until the committing CT action. For later | |
752 | * packets NAT and Helper may be called in either order. | |
753 | * | |
754 | * NAT will be done only if the CT action has NAT, and only | |
755 | * once per packet (per zone), as guarded by the NAT bits in | |
756 | * the key->ct.state. | |
757 | */ | |
758 | if (info->nat && !(key->ct.state & OVS_CS_F_NAT_MASK) && | |
759 | (nf_ct_is_confirmed(ct) || info->commit) && | |
760 | ovs_ct_nat(net, key, info, skb, ct, ctinfo) != NF_ACCEPT) { | |
761 | return -EINVAL; | |
762 | } | |
763 | ||
16ec3d4f JS |
764 | /* Userspace may decide to perform a ct lookup without a helper |
765 | * specified followed by a (recirculate and) commit with one. | |
766 | * Therefore, for unconfirmed connections which we will commit, | |
767 | * we need to attach the helper here. | |
768 | */ | |
769 | if (!nf_ct_is_confirmed(ct) && info->commit && | |
770 | info->helper && !nfct_help(ct)) { | |
771 | int err = __nf_ct_try_assign_helper(ct, info->ct, | |
772 | GFP_ATOMIC); | |
773 | if (err) | |
774 | return err; | |
775 | } | |
776 | ||
05752523 JR |
777 | /* Call the helper only if: |
778 | * - nf_conntrack_in() was executed above ("!cached") for a | |
779 | * confirmed connection, or | |
780 | * - When committing an unconfirmed connection. | |
781 | */ | |
782 | if ((nf_ct_is_confirmed(ct) ? !cached : info->commit) && | |
783 | ovs_ct_helper(skb, info->family) != NF_ACCEPT) { | |
784 | return -EINVAL; | |
785 | } | |
7f8a436e JS |
786 | } |
787 | ||
788 | return 0; | |
789 | } | |
790 | ||
791 | /* Lookup connection and read fields into key. */ | |
792 | static int ovs_ct_lookup(struct net *net, struct sw_flow_key *key, | |
793 | const struct ovs_conntrack_info *info, | |
794 | struct sk_buff *skb) | |
795 | { | |
796 | struct nf_conntrack_expect *exp; | |
797 | ||
9f13ded8 JR |
798 | /* If we pass an expected packet through nf_conntrack_in() the |
799 | * expectation is typically removed, but the packet could still be | |
800 | * lost in upcall processing. To prevent this from happening we | |
801 | * perform an explicit expectation lookup. Expected connections are | |
802 | * always new, and will be passed through conntrack only when they are | |
803 | * committed, as it is OK to remove the expectation at that time. | |
804 | */ | |
7f8a436e JS |
805 | exp = ovs_ct_expect_find(net, &info->zone, info->family, skb); |
806 | if (exp) { | |
807 | u8 state; | |
808 | ||
05752523 JR |
809 | /* NOTE: New connections are NATted and Helped only when |
810 | * committed, so we are not calling into NAT here. | |
811 | */ | |
7f8a436e | 812 | state = OVS_CS_F_TRACKED | OVS_CS_F_NEW | OVS_CS_F_RELATED; |
182e3042 | 813 | __ovs_ct_update_key(key, state, &info->zone, exp->master); |
d913d3a7 SG |
814 | } else { |
815 | struct nf_conn *ct; | |
816 | int err; | |
817 | ||
818 | err = __ovs_ct_lookup(net, key, info, skb); | |
819 | if (err) | |
820 | return err; | |
821 | ||
822 | ct = (struct nf_conn *)skb->nfct; | |
823 | if (ct) | |
824 | nf_ct_deliver_cached_events(ct); | |
825 | } | |
7f8a436e JS |
826 | |
827 | return 0; | |
828 | } | |
829 | ||
33db4125 | 830 | static bool labels_nonzero(const struct ovs_key_ct_labels *labels) |
c2ac6673 JS |
831 | { |
832 | size_t i; | |
833 | ||
33db4125 JS |
834 | for (i = 0; i < sizeof(*labels); i++) |
835 | if (labels->ct_labels[i]) | |
c2ac6673 JS |
836 | return true; |
837 | ||
838 | return false; | |
839 | } | |
840 | ||
7d904c7b JR |
841 | /* Lookup connection and confirm if unconfirmed. */ |
842 | static int ovs_ct_commit(struct net *net, struct sw_flow_key *key, | |
843 | const struct ovs_conntrack_info *info, | |
844 | struct sk_buff *skb) | |
845 | { | |
846 | int err; | |
847 | ||
848 | err = __ovs_ct_lookup(net, key, info, skb); | |
849 | if (err) | |
850 | return err; | |
851 | ||
852 | /* Apply changes before confirming the connection so that the initial | |
853 | * conntrack NEW netlink event carries the values given in the CT | |
854 | * action. | |
855 | */ | |
856 | if (info->mark.mask) { | |
857 | err = ovs_ct_set_mark(skb, key, info->mark.value, | |
858 | info->mark.mask); | |
859 | if (err) | |
860 | return err; | |
861 | } | |
862 | if (labels_nonzero(&info->labels.mask)) { | |
863 | err = ovs_ct_set_labels(skb, key, &info->labels.value, | |
864 | &info->labels.mask); | |
865 | if (err) | |
866 | return err; | |
867 | } | |
868 | /* This will take care of sending queued events even if the connection | |
869 | * is already confirmed. | |
870 | */ | |
871 | if (nf_conntrack_confirm(skb) != NF_ACCEPT) | |
872 | return -EINVAL; | |
873 | ||
874 | return 0; | |
875 | } | |
876 | ||
74c16618 JS |
877 | /* Returns 0 on success, -EINPROGRESS if 'skb' is stolen, or other nonzero |
878 | * value if 'skb' is freed. | |
879 | */ | |
7f8a436e JS |
880 | int ovs_ct_execute(struct net *net, struct sk_buff *skb, |
881 | struct sw_flow_key *key, | |
882 | const struct ovs_conntrack_info *info) | |
883 | { | |
884 | int nh_ofs; | |
885 | int err; | |
886 | ||
887 | /* The conntrack module expects to be working at L3. */ | |
888 | nh_ofs = skb_network_offset(skb); | |
75f01a4c | 889 | skb_pull_rcsum(skb, nh_ofs); |
7f8a436e JS |
890 | |
891 | if (key->ip.frag != OVS_FRAG_TYPE_NONE) { | |
892 | err = handle_fragments(net, key, info->zone.id, skb); | |
893 | if (err) | |
894 | return err; | |
895 | } | |
896 | ||
ab38a7b5 | 897 | if (info->commit) |
7d904c7b | 898 | err = ovs_ct_commit(net, key, info, skb); |
7f8a436e JS |
899 | else |
900 | err = ovs_ct_lookup(net, key, info, skb); | |
901 | ||
902 | skb_push(skb, nh_ofs); | |
75f01a4c | 903 | skb_postpush_rcsum(skb, skb->data, nh_ofs); |
74c16618 JS |
904 | if (err) |
905 | kfree_skb(skb); | |
7f8a436e JS |
906 | return err; |
907 | } | |
908 | ||
cae3a262 JS |
909 | static int ovs_ct_add_helper(struct ovs_conntrack_info *info, const char *name, |
910 | const struct sw_flow_key *key, bool log) | |
911 | { | |
912 | struct nf_conntrack_helper *helper; | |
913 | struct nf_conn_help *help; | |
914 | ||
915 | helper = nf_conntrack_helper_try_module_get(name, info->family, | |
916 | key->ip.proto); | |
917 | if (!helper) { | |
918 | OVS_NLERR(log, "Unknown helper \"%s\"", name); | |
919 | return -EINVAL; | |
920 | } | |
921 | ||
922 | help = nf_ct_helper_ext_add(info->ct, helper, GFP_KERNEL); | |
923 | if (!help) { | |
924 | module_put(helper->me); | |
925 | return -ENOMEM; | |
926 | } | |
927 | ||
928 | rcu_assign_pointer(help->helper, helper); | |
929 | info->helper = helper; | |
930 | return 0; | |
931 | } | |
932 | ||
05752523 JR |
933 | #ifdef CONFIG_NF_NAT_NEEDED |
934 | static int parse_nat(const struct nlattr *attr, | |
935 | struct ovs_conntrack_info *info, bool log) | |
936 | { | |
937 | struct nlattr *a; | |
938 | int rem; | |
939 | bool have_ip_max = false; | |
940 | bool have_proto_max = false; | |
941 | bool ip_vers = (info->family == NFPROTO_IPV6); | |
942 | ||
943 | nla_for_each_nested(a, attr, rem) { | |
944 | static const int ovs_nat_attr_lens[OVS_NAT_ATTR_MAX + 1][2] = { | |
945 | [OVS_NAT_ATTR_SRC] = {0, 0}, | |
946 | [OVS_NAT_ATTR_DST] = {0, 0}, | |
947 | [OVS_NAT_ATTR_IP_MIN] = {sizeof(struct in_addr), | |
948 | sizeof(struct in6_addr)}, | |
949 | [OVS_NAT_ATTR_IP_MAX] = {sizeof(struct in_addr), | |
950 | sizeof(struct in6_addr)}, | |
951 | [OVS_NAT_ATTR_PROTO_MIN] = {sizeof(u16), sizeof(u16)}, | |
952 | [OVS_NAT_ATTR_PROTO_MAX] = {sizeof(u16), sizeof(u16)}, | |
953 | [OVS_NAT_ATTR_PERSISTENT] = {0, 0}, | |
954 | [OVS_NAT_ATTR_PROTO_HASH] = {0, 0}, | |
955 | [OVS_NAT_ATTR_PROTO_RANDOM] = {0, 0}, | |
956 | }; | |
957 | int type = nla_type(a); | |
958 | ||
959 | if (type > OVS_NAT_ATTR_MAX) { | |
960 | OVS_NLERR(log, | |
961 | "Unknown NAT attribute (type=%d, max=%d).\n", | |
962 | type, OVS_NAT_ATTR_MAX); | |
963 | return -EINVAL; | |
964 | } | |
965 | ||
966 | if (nla_len(a) != ovs_nat_attr_lens[type][ip_vers]) { | |
967 | OVS_NLERR(log, | |
968 | "NAT attribute type %d has unexpected length (%d != %d).\n", | |
969 | type, nla_len(a), | |
970 | ovs_nat_attr_lens[type][ip_vers]); | |
971 | return -EINVAL; | |
972 | } | |
973 | ||
974 | switch (type) { | |
975 | case OVS_NAT_ATTR_SRC: | |
976 | case OVS_NAT_ATTR_DST: | |
977 | if (info->nat) { | |
978 | OVS_NLERR(log, | |
979 | "Only one type of NAT may be specified.\n" | |
980 | ); | |
981 | return -ERANGE; | |
982 | } | |
983 | info->nat |= OVS_CT_NAT; | |
984 | info->nat |= ((type == OVS_NAT_ATTR_SRC) | |
985 | ? OVS_CT_SRC_NAT : OVS_CT_DST_NAT); | |
986 | break; | |
987 | ||
988 | case OVS_NAT_ATTR_IP_MIN: | |
ac71b46e HY |
989 | nla_memcpy(&info->range.min_addr, a, |
990 | sizeof(info->range.min_addr)); | |
05752523 JR |
991 | info->range.flags |= NF_NAT_RANGE_MAP_IPS; |
992 | break; | |
993 | ||
994 | case OVS_NAT_ATTR_IP_MAX: | |
995 | have_ip_max = true; | |
996 | nla_memcpy(&info->range.max_addr, a, | |
997 | sizeof(info->range.max_addr)); | |
998 | info->range.flags |= NF_NAT_RANGE_MAP_IPS; | |
999 | break; | |
1000 | ||
1001 | case OVS_NAT_ATTR_PROTO_MIN: | |
1002 | info->range.min_proto.all = htons(nla_get_u16(a)); | |
1003 | info->range.flags |= NF_NAT_RANGE_PROTO_SPECIFIED; | |
1004 | break; | |
1005 | ||
1006 | case OVS_NAT_ATTR_PROTO_MAX: | |
1007 | have_proto_max = true; | |
1008 | info->range.max_proto.all = htons(nla_get_u16(a)); | |
1009 | info->range.flags |= NF_NAT_RANGE_PROTO_SPECIFIED; | |
1010 | break; | |
1011 | ||
1012 | case OVS_NAT_ATTR_PERSISTENT: | |
1013 | info->range.flags |= NF_NAT_RANGE_PERSISTENT; | |
1014 | break; | |
1015 | ||
1016 | case OVS_NAT_ATTR_PROTO_HASH: | |
1017 | info->range.flags |= NF_NAT_RANGE_PROTO_RANDOM; | |
1018 | break; | |
1019 | ||
1020 | case OVS_NAT_ATTR_PROTO_RANDOM: | |
1021 | info->range.flags |= NF_NAT_RANGE_PROTO_RANDOM_FULLY; | |
1022 | break; | |
1023 | ||
1024 | default: | |
1025 | OVS_NLERR(log, "Unknown nat attribute (%d).\n", type); | |
1026 | return -EINVAL; | |
1027 | } | |
1028 | } | |
1029 | ||
1030 | if (rem > 0) { | |
1031 | OVS_NLERR(log, "NAT attribute has %d unknown bytes.\n", rem); | |
1032 | return -EINVAL; | |
1033 | } | |
1034 | if (!info->nat) { | |
1035 | /* Do not allow flags if no type is given. */ | |
1036 | if (info->range.flags) { | |
1037 | OVS_NLERR(log, | |
1038 | "NAT flags may be given only when NAT range (SRC or DST) is also specified.\n" | |
1039 | ); | |
1040 | return -EINVAL; | |
1041 | } | |
1042 | info->nat = OVS_CT_NAT; /* NAT existing connections. */ | |
1043 | } else if (!info->commit) { | |
1044 | OVS_NLERR(log, | |
1045 | "NAT attributes may be specified only when CT COMMIT flag is also specified.\n" | |
1046 | ); | |
1047 | return -EINVAL; | |
1048 | } | |
1049 | /* Allow missing IP_MAX. */ | |
1050 | if (info->range.flags & NF_NAT_RANGE_MAP_IPS && !have_ip_max) { | |
1051 | memcpy(&info->range.max_addr, &info->range.min_addr, | |
1052 | sizeof(info->range.max_addr)); | |
1053 | } | |
1054 | /* Allow missing PROTO_MAX. */ | |
1055 | if (info->range.flags & NF_NAT_RANGE_PROTO_SPECIFIED && | |
1056 | !have_proto_max) { | |
1057 | info->range.max_proto.all = info->range.min_proto.all; | |
1058 | } | |
1059 | return 0; | |
1060 | } | |
1061 | #endif | |
1062 | ||
7f8a436e | 1063 | static const struct ovs_ct_len_tbl ovs_ct_attr_lens[OVS_CT_ATTR_MAX + 1] = { |
ab38a7b5 | 1064 | [OVS_CT_ATTR_COMMIT] = { .minlen = 0, .maxlen = 0 }, |
7f8a436e JS |
1065 | [OVS_CT_ATTR_ZONE] = { .minlen = sizeof(u16), |
1066 | .maxlen = sizeof(u16) }, | |
182e3042 JS |
1067 | [OVS_CT_ATTR_MARK] = { .minlen = sizeof(struct md_mark), |
1068 | .maxlen = sizeof(struct md_mark) }, | |
33db4125 JS |
1069 | [OVS_CT_ATTR_LABELS] = { .minlen = sizeof(struct md_labels), |
1070 | .maxlen = sizeof(struct md_labels) }, | |
cae3a262 | 1071 | [OVS_CT_ATTR_HELPER] = { .minlen = 1, |
05752523 JR |
1072 | .maxlen = NF_CT_HELPER_NAME_LEN }, |
1073 | #ifdef CONFIG_NF_NAT_NEEDED | |
1074 | /* NAT length is checked when parsing the nested attributes. */ | |
1075 | [OVS_CT_ATTR_NAT] = { .minlen = 0, .maxlen = INT_MAX }, | |
1076 | #endif | |
7f8a436e JS |
1077 | }; |
1078 | ||
1079 | static int parse_ct(const struct nlattr *attr, struct ovs_conntrack_info *info, | |
cae3a262 | 1080 | const char **helper, bool log) |
7f8a436e JS |
1081 | { |
1082 | struct nlattr *a; | |
1083 | int rem; | |
1084 | ||
1085 | nla_for_each_nested(a, attr, rem) { | |
1086 | int type = nla_type(a); | |
1087 | int maxlen = ovs_ct_attr_lens[type].maxlen; | |
1088 | int minlen = ovs_ct_attr_lens[type].minlen; | |
1089 | ||
1090 | if (type > OVS_CT_ATTR_MAX) { | |
1091 | OVS_NLERR(log, | |
1092 | "Unknown conntrack attr (type=%d, max=%d)", | |
1093 | type, OVS_CT_ATTR_MAX); | |
1094 | return -EINVAL; | |
1095 | } | |
1096 | if (nla_len(a) < minlen || nla_len(a) > maxlen) { | |
1097 | OVS_NLERR(log, | |
1098 | "Conntrack attr type has unexpected length (type=%d, length=%d, expected=%d)", | |
1099 | type, nla_len(a), maxlen); | |
1100 | return -EINVAL; | |
1101 | } | |
1102 | ||
1103 | switch (type) { | |
ab38a7b5 JS |
1104 | case OVS_CT_ATTR_COMMIT: |
1105 | info->commit = true; | |
7f8a436e JS |
1106 | break; |
1107 | #ifdef CONFIG_NF_CONNTRACK_ZONES | |
1108 | case OVS_CT_ATTR_ZONE: | |
1109 | info->zone.id = nla_get_u16(a); | |
1110 | break; | |
182e3042 JS |
1111 | #endif |
1112 | #ifdef CONFIG_NF_CONNTRACK_MARK | |
1113 | case OVS_CT_ATTR_MARK: { | |
1114 | struct md_mark *mark = nla_data(a); | |
1115 | ||
e754ec69 JS |
1116 | if (!mark->mask) { |
1117 | OVS_NLERR(log, "ct_mark mask cannot be 0"); | |
1118 | return -EINVAL; | |
1119 | } | |
182e3042 JS |
1120 | info->mark = *mark; |
1121 | break; | |
1122 | } | |
c2ac6673 JS |
1123 | #endif |
1124 | #ifdef CONFIG_NF_CONNTRACK_LABELS | |
33db4125 JS |
1125 | case OVS_CT_ATTR_LABELS: { |
1126 | struct md_labels *labels = nla_data(a); | |
c2ac6673 | 1127 | |
e754ec69 JS |
1128 | if (!labels_nonzero(&labels->mask)) { |
1129 | OVS_NLERR(log, "ct_labels mask cannot be 0"); | |
1130 | return -EINVAL; | |
1131 | } | |
33db4125 | 1132 | info->labels = *labels; |
c2ac6673 JS |
1133 | break; |
1134 | } | |
7f8a436e | 1135 | #endif |
cae3a262 JS |
1136 | case OVS_CT_ATTR_HELPER: |
1137 | *helper = nla_data(a); | |
1138 | if (!memchr(*helper, '\0', nla_len(a))) { | |
1139 | OVS_NLERR(log, "Invalid conntrack helper"); | |
1140 | return -EINVAL; | |
1141 | } | |
1142 | break; | |
05752523 JR |
1143 | #ifdef CONFIG_NF_NAT_NEEDED |
1144 | case OVS_CT_ATTR_NAT: { | |
1145 | int err = parse_nat(a, info, log); | |
1146 | ||
1147 | if (err) | |
1148 | return err; | |
1149 | break; | |
1150 | } | |
1151 | #endif | |
7f8a436e JS |
1152 | default: |
1153 | OVS_NLERR(log, "Unknown conntrack attr (%d)", | |
1154 | type); | |
1155 | return -EINVAL; | |
1156 | } | |
1157 | } | |
1158 | ||
7d904c7b JR |
1159 | #ifdef CONFIG_NF_CONNTRACK_MARK |
1160 | if (!info->commit && info->mark.mask) { | |
1161 | OVS_NLERR(log, | |
1162 | "Setting conntrack mark requires 'commit' flag."); | |
1163 | return -EINVAL; | |
1164 | } | |
1165 | #endif | |
1166 | #ifdef CONFIG_NF_CONNTRACK_LABELS | |
1167 | if (!info->commit && labels_nonzero(&info->labels.mask)) { | |
1168 | OVS_NLERR(log, | |
1169 | "Setting conntrack labels requires 'commit' flag."); | |
1170 | return -EINVAL; | |
1171 | } | |
1172 | #endif | |
7f8a436e JS |
1173 | if (rem > 0) { |
1174 | OVS_NLERR(log, "Conntrack attr has %d unknown bytes", rem); | |
1175 | return -EINVAL; | |
1176 | } | |
1177 | ||
1178 | return 0; | |
1179 | } | |
1180 | ||
c2ac6673 | 1181 | bool ovs_ct_verify(struct net *net, enum ovs_key_attr attr) |
7f8a436e JS |
1182 | { |
1183 | if (attr == OVS_KEY_ATTR_CT_STATE) | |
1184 | return true; | |
1185 | if (IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES) && | |
1186 | attr == OVS_KEY_ATTR_CT_ZONE) | |
1187 | return true; | |
182e3042 JS |
1188 | if (IS_ENABLED(CONFIG_NF_CONNTRACK_MARK) && |
1189 | attr == OVS_KEY_ATTR_CT_MARK) | |
1190 | return true; | |
c2ac6673 | 1191 | if (IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS) && |
33db4125 | 1192 | attr == OVS_KEY_ATTR_CT_LABELS) { |
c2ac6673 JS |
1193 | struct ovs_net *ovs_net = net_generic(net, ovs_net_id); |
1194 | ||
1195 | return ovs_net->xt_label; | |
1196 | } | |
7f8a436e JS |
1197 | |
1198 | return false; | |
1199 | } | |
1200 | ||
1201 | int ovs_ct_copy_action(struct net *net, const struct nlattr *attr, | |
1202 | const struct sw_flow_key *key, | |
1203 | struct sw_flow_actions **sfa, bool log) | |
1204 | { | |
1205 | struct ovs_conntrack_info ct_info; | |
cae3a262 | 1206 | const char *helper = NULL; |
7f8a436e JS |
1207 | u16 family; |
1208 | int err; | |
1209 | ||
1210 | family = key_to_nfproto(key); | |
1211 | if (family == NFPROTO_UNSPEC) { | |
1212 | OVS_NLERR(log, "ct family unspecified"); | |
1213 | return -EINVAL; | |
1214 | } | |
1215 | ||
1216 | memset(&ct_info, 0, sizeof(ct_info)); | |
1217 | ct_info.family = family; | |
1218 | ||
1219 | nf_ct_zone_init(&ct_info.zone, NF_CT_DEFAULT_ZONE_ID, | |
1220 | NF_CT_DEFAULT_ZONE_DIR, 0); | |
1221 | ||
cae3a262 | 1222 | err = parse_ct(attr, &ct_info, &helper, log); |
7f8a436e JS |
1223 | if (err) |
1224 | return err; | |
1225 | ||
1226 | /* Set up template for tracking connections in specific zones. */ | |
1227 | ct_info.ct = nf_ct_tmpl_alloc(net, &ct_info.zone, GFP_KERNEL); | |
1228 | if (!ct_info.ct) { | |
1229 | OVS_NLERR(log, "Failed to allocate conntrack template"); | |
1230 | return -ENOMEM; | |
1231 | } | |
90c7afc9 JS |
1232 | |
1233 | __set_bit(IPS_CONFIRMED_BIT, &ct_info.ct->status); | |
1234 | nf_conntrack_get(&ct_info.ct->ct_general); | |
1235 | ||
cae3a262 JS |
1236 | if (helper) { |
1237 | err = ovs_ct_add_helper(&ct_info, helper, key, log); | |
1238 | if (err) | |
1239 | goto err_free_ct; | |
1240 | } | |
7f8a436e JS |
1241 | |
1242 | err = ovs_nla_add_action(sfa, OVS_ACTION_ATTR_CT, &ct_info, | |
1243 | sizeof(ct_info), log); | |
1244 | if (err) | |
1245 | goto err_free_ct; | |
1246 | ||
7f8a436e JS |
1247 | return 0; |
1248 | err_free_ct: | |
2f3ab9f9 | 1249 | __ovs_ct_free_action(&ct_info); |
7f8a436e JS |
1250 | return err; |
1251 | } | |
1252 | ||
05752523 JR |
1253 | #ifdef CONFIG_NF_NAT_NEEDED |
1254 | static bool ovs_ct_nat_to_attr(const struct ovs_conntrack_info *info, | |
1255 | struct sk_buff *skb) | |
1256 | { | |
1257 | struct nlattr *start; | |
1258 | ||
1259 | start = nla_nest_start(skb, OVS_CT_ATTR_NAT); | |
1260 | if (!start) | |
1261 | return false; | |
1262 | ||
1263 | if (info->nat & OVS_CT_SRC_NAT) { | |
1264 | if (nla_put_flag(skb, OVS_NAT_ATTR_SRC)) | |
1265 | return false; | |
1266 | } else if (info->nat & OVS_CT_DST_NAT) { | |
1267 | if (nla_put_flag(skb, OVS_NAT_ATTR_DST)) | |
1268 | return false; | |
1269 | } else { | |
1270 | goto out; | |
1271 | } | |
1272 | ||
1273 | if (info->range.flags & NF_NAT_RANGE_MAP_IPS) { | |
99b7248e AB |
1274 | if (IS_ENABLED(CONFIG_NF_NAT_IPV4) && |
1275 | info->family == NFPROTO_IPV4) { | |
05752523 JR |
1276 | if (nla_put_in_addr(skb, OVS_NAT_ATTR_IP_MIN, |
1277 | info->range.min_addr.ip) || | |
1278 | (info->range.max_addr.ip | |
1279 | != info->range.min_addr.ip && | |
1280 | (nla_put_in_addr(skb, OVS_NAT_ATTR_IP_MAX, | |
1281 | info->range.max_addr.ip)))) | |
1282 | return false; | |
99b7248e AB |
1283 | } else if (IS_ENABLED(CONFIG_NF_NAT_IPV6) && |
1284 | info->family == NFPROTO_IPV6) { | |
05752523 JR |
1285 | if (nla_put_in6_addr(skb, OVS_NAT_ATTR_IP_MIN, |
1286 | &info->range.min_addr.in6) || | |
1287 | (memcmp(&info->range.max_addr.in6, | |
1288 | &info->range.min_addr.in6, | |
1289 | sizeof(info->range.max_addr.in6)) && | |
1290 | (nla_put_in6_addr(skb, OVS_NAT_ATTR_IP_MAX, | |
1291 | &info->range.max_addr.in6)))) | |
1292 | return false; | |
05752523 JR |
1293 | } else { |
1294 | return false; | |
1295 | } | |
1296 | } | |
1297 | if (info->range.flags & NF_NAT_RANGE_PROTO_SPECIFIED && | |
1298 | (nla_put_u16(skb, OVS_NAT_ATTR_PROTO_MIN, | |
1299 | ntohs(info->range.min_proto.all)) || | |
1300 | (info->range.max_proto.all != info->range.min_proto.all && | |
1301 | nla_put_u16(skb, OVS_NAT_ATTR_PROTO_MAX, | |
1302 | ntohs(info->range.max_proto.all))))) | |
1303 | return false; | |
1304 | ||
1305 | if (info->range.flags & NF_NAT_RANGE_PERSISTENT && | |
1306 | nla_put_flag(skb, OVS_NAT_ATTR_PERSISTENT)) | |
1307 | return false; | |
1308 | if (info->range.flags & NF_NAT_RANGE_PROTO_RANDOM && | |
1309 | nla_put_flag(skb, OVS_NAT_ATTR_PROTO_HASH)) | |
1310 | return false; | |
1311 | if (info->range.flags & NF_NAT_RANGE_PROTO_RANDOM_FULLY && | |
1312 | nla_put_flag(skb, OVS_NAT_ATTR_PROTO_RANDOM)) | |
1313 | return false; | |
1314 | out: | |
1315 | nla_nest_end(skb, start); | |
1316 | ||
1317 | return true; | |
1318 | } | |
1319 | #endif | |
1320 | ||
7f8a436e JS |
1321 | int ovs_ct_action_to_attr(const struct ovs_conntrack_info *ct_info, |
1322 | struct sk_buff *skb) | |
1323 | { | |
1324 | struct nlattr *start; | |
1325 | ||
1326 | start = nla_nest_start(skb, OVS_ACTION_ATTR_CT); | |
1327 | if (!start) | |
1328 | return -EMSGSIZE; | |
1329 | ||
ab38a7b5 | 1330 | if (ct_info->commit && nla_put_flag(skb, OVS_CT_ATTR_COMMIT)) |
7f8a436e JS |
1331 | return -EMSGSIZE; |
1332 | if (IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES) && | |
1333 | nla_put_u16(skb, OVS_CT_ATTR_ZONE, ct_info->zone.id)) | |
1334 | return -EMSGSIZE; | |
e754ec69 | 1335 | if (IS_ENABLED(CONFIG_NF_CONNTRACK_MARK) && ct_info->mark.mask && |
182e3042 JS |
1336 | nla_put(skb, OVS_CT_ATTR_MARK, sizeof(ct_info->mark), |
1337 | &ct_info->mark)) | |
1338 | return -EMSGSIZE; | |
c2ac6673 | 1339 | if (IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS) && |
e754ec69 | 1340 | labels_nonzero(&ct_info->labels.mask) && |
33db4125 JS |
1341 | nla_put(skb, OVS_CT_ATTR_LABELS, sizeof(ct_info->labels), |
1342 | &ct_info->labels)) | |
c2ac6673 | 1343 | return -EMSGSIZE; |
cae3a262 JS |
1344 | if (ct_info->helper) { |
1345 | if (nla_put_string(skb, OVS_CT_ATTR_HELPER, | |
1346 | ct_info->helper->name)) | |
1347 | return -EMSGSIZE; | |
1348 | } | |
05752523 JR |
1349 | #ifdef CONFIG_NF_NAT_NEEDED |
1350 | if (ct_info->nat && !ovs_ct_nat_to_attr(ct_info, skb)) | |
1351 | return -EMSGSIZE; | |
1352 | #endif | |
7f8a436e JS |
1353 | nla_nest_end(skb, start); |
1354 | ||
1355 | return 0; | |
1356 | } | |
1357 | ||
1358 | void ovs_ct_free_action(const struct nlattr *a) | |
1359 | { | |
1360 | struct ovs_conntrack_info *ct_info = nla_data(a); | |
1361 | ||
2f3ab9f9 JS |
1362 | __ovs_ct_free_action(ct_info); |
1363 | } | |
1364 | ||
1365 | static void __ovs_ct_free_action(struct ovs_conntrack_info *ct_info) | |
1366 | { | |
cae3a262 JS |
1367 | if (ct_info->helper) |
1368 | module_put(ct_info->helper->me); | |
7f8a436e | 1369 | if (ct_info->ct) |
76644232 | 1370 | nf_ct_tmpl_free(ct_info->ct); |
7f8a436e | 1371 | } |
c2ac6673 JS |
1372 | |
1373 | void ovs_ct_init(struct net *net) | |
1374 | { | |
33db4125 | 1375 | unsigned int n_bits = sizeof(struct ovs_key_ct_labels) * BITS_PER_BYTE; |
c2ac6673 JS |
1376 | struct ovs_net *ovs_net = net_generic(net, ovs_net_id); |
1377 | ||
adff6c65 | 1378 | if (nf_connlabels_get(net, n_bits - 1)) { |
c2ac6673 JS |
1379 | ovs_net->xt_label = false; |
1380 | OVS_NLERR(true, "Failed to set connlabel length"); | |
1381 | } else { | |
1382 | ovs_net->xt_label = true; | |
1383 | } | |
1384 | } | |
1385 | ||
1386 | void ovs_ct_exit(struct net *net) | |
1387 | { | |
1388 | struct ovs_net *ovs_net = net_generic(net, ovs_net_id); | |
1389 | ||
1390 | if (ovs_net->xt_label) | |
1391 | nf_connlabels_put(net); | |
1392 | } |