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) { | |
138 | size_t len = cl->words * sizeof(long); | |
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 | } | |
33db4125 | 277 | if (!cl || cl->words * sizeof(long) < 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 | 369 | |
49e261a8 | 370 | skb_orphan(skb); |
7f8a436e | 371 | memset(IP6CB(skb), 0, sizeof(struct inet6_skb_parm)); |
daaa7d64 FW |
372 | err = nf_ct_frag6_gather(net, skb, user); |
373 | if (err) | |
374 | return err; | |
7f8a436e | 375 | |
daaa7d64 | 376 | key->ip.proto = ipv6_hdr(skb)->nexthdr; |
7f8a436e | 377 | ovs_cb.mru = IP6CB(skb)->frag_max_size; |
7f8a436e JS |
378 | #endif |
379 | } else { | |
74c16618 | 380 | kfree_skb(skb); |
7f8a436e JS |
381 | return -EPFNOSUPPORT; |
382 | } | |
383 | ||
384 | key->ip.frag = OVS_FRAG_TYPE_NONE; | |
385 | skb_clear_hash(skb); | |
386 | skb->ignore_df = 1; | |
387 | *OVS_CB(skb) = ovs_cb; | |
388 | ||
389 | return 0; | |
390 | } | |
391 | ||
392 | static struct nf_conntrack_expect * | |
393 | ovs_ct_expect_find(struct net *net, const struct nf_conntrack_zone *zone, | |
394 | u16 proto, const struct sk_buff *skb) | |
395 | { | |
396 | struct nf_conntrack_tuple tuple; | |
397 | ||
a31f1adc | 398 | if (!nf_ct_get_tuplepr(skb, skb_network_offset(skb), proto, net, &tuple)) |
7f8a436e JS |
399 | return NULL; |
400 | return __nf_ct_expect_find(net, zone, &tuple); | |
401 | } | |
402 | ||
289f2253 JR |
403 | /* This replicates logic from nf_conntrack_core.c that is not exported. */ |
404 | static enum ip_conntrack_info | |
405 | ovs_ct_get_info(const struct nf_conntrack_tuple_hash *h) | |
406 | { | |
407 | const struct nf_conn *ct = nf_ct_tuplehash_to_ctrack(h); | |
408 | ||
409 | if (NF_CT_DIRECTION(h) == IP_CT_DIR_REPLY) | |
410 | return IP_CT_ESTABLISHED_REPLY; | |
411 | /* Once we've had two way comms, always ESTABLISHED. */ | |
412 | if (test_bit(IPS_SEEN_REPLY_BIT, &ct->status)) | |
413 | return IP_CT_ESTABLISHED; | |
414 | if (test_bit(IPS_EXPECTED_BIT, &ct->status)) | |
415 | return IP_CT_RELATED; | |
416 | return IP_CT_NEW; | |
417 | } | |
418 | ||
419 | /* Find an existing connection which this packet belongs to without | |
420 | * re-attributing statistics or modifying the connection state. This allows an | |
421 | * skb->nfct lost due to an upcall to be recovered during actions execution. | |
422 | * | |
423 | * Must be called with rcu_read_lock. | |
424 | * | |
425 | * On success, populates skb->nfct and skb->nfctinfo, and returns the | |
426 | * connection. Returns NULL if there is no existing entry. | |
427 | */ | |
428 | static struct nf_conn * | |
429 | ovs_ct_find_existing(struct net *net, const struct nf_conntrack_zone *zone, | |
430 | u8 l3num, struct sk_buff *skb) | |
431 | { | |
432 | struct nf_conntrack_l3proto *l3proto; | |
433 | struct nf_conntrack_l4proto *l4proto; | |
434 | struct nf_conntrack_tuple tuple; | |
435 | struct nf_conntrack_tuple_hash *h; | |
436 | enum ip_conntrack_info ctinfo; | |
437 | struct nf_conn *ct; | |
438 | unsigned int dataoff; | |
439 | u8 protonum; | |
440 | ||
441 | l3proto = __nf_ct_l3proto_find(l3num); | |
442 | if (!l3proto) { | |
443 | pr_debug("ovs_ct_find_existing: Can't get l3proto\n"); | |
444 | return NULL; | |
445 | } | |
446 | if (l3proto->get_l4proto(skb, skb_network_offset(skb), &dataoff, | |
447 | &protonum) <= 0) { | |
448 | pr_debug("ovs_ct_find_existing: Can't get protonum\n"); | |
449 | return NULL; | |
450 | } | |
451 | l4proto = __nf_ct_l4proto_find(l3num, protonum); | |
452 | if (!l4proto) { | |
453 | pr_debug("ovs_ct_find_existing: Can't get l4proto\n"); | |
454 | return NULL; | |
455 | } | |
456 | if (!nf_ct_get_tuple(skb, skb_network_offset(skb), dataoff, l3num, | |
457 | protonum, net, &tuple, l3proto, l4proto)) { | |
458 | pr_debug("ovs_ct_find_existing: Can't get tuple\n"); | |
459 | return NULL; | |
460 | } | |
461 | ||
462 | /* look for tuple match */ | |
463 | h = nf_conntrack_find_get(net, zone, &tuple); | |
464 | if (!h) | |
465 | return NULL; /* Not found. */ | |
466 | ||
467 | ct = nf_ct_tuplehash_to_ctrack(h); | |
468 | ||
469 | ctinfo = ovs_ct_get_info(h); | |
470 | if (ctinfo == IP_CT_NEW) { | |
471 | /* This should not happen. */ | |
472 | WARN_ONCE(1, "ovs_ct_find_existing: new packet for %p\n", ct); | |
473 | } | |
474 | skb->nfct = &ct->ct_general; | |
475 | skb->nfctinfo = ctinfo; | |
476 | return ct; | |
477 | } | |
478 | ||
7f8a436e | 479 | /* Determine whether skb->nfct is equal to the result of conntrack lookup. */ |
289f2253 JR |
480 | static bool skb_nfct_cached(struct net *net, |
481 | const struct sw_flow_key *key, | |
482 | const struct ovs_conntrack_info *info, | |
483 | struct sk_buff *skb) | |
7f8a436e JS |
484 | { |
485 | enum ip_conntrack_info ctinfo; | |
486 | struct nf_conn *ct; | |
487 | ||
488 | ct = nf_ct_get(skb, &ctinfo); | |
289f2253 JR |
489 | /* If no ct, check if we have evidence that an existing conntrack entry |
490 | * might be found for this skb. This happens when we lose a skb->nfct | |
491 | * due to an upcall. If the connection was not confirmed, it is not | |
492 | * cached and needs to be run through conntrack again. | |
493 | */ | |
494 | if (!ct && key->ct.state & OVS_CS_F_TRACKED && | |
495 | !(key->ct.state & OVS_CS_F_INVALID) && | |
496 | key->ct.zone == info->zone.id) | |
497 | ct = ovs_ct_find_existing(net, &info->zone, info->family, skb); | |
7f8a436e JS |
498 | if (!ct) |
499 | return false; | |
500 | if (!net_eq(net, read_pnet(&ct->ct_net))) | |
501 | return false; | |
502 | if (!nf_ct_zone_equal_any(info->ct, nf_ct_zone(ct))) | |
503 | return false; | |
cae3a262 JS |
504 | if (info->helper) { |
505 | struct nf_conn_help *help; | |
506 | ||
507 | help = nf_ct_ext_find(ct, NF_CT_EXT_HELPER); | |
508 | if (help && rcu_access_pointer(help->helper) != info->helper) | |
509 | return false; | |
510 | } | |
7f8a436e JS |
511 | |
512 | return true; | |
513 | } | |
514 | ||
05752523 JR |
515 | #ifdef CONFIG_NF_NAT_NEEDED |
516 | /* Modelled after nf_nat_ipv[46]_fn(). | |
517 | * range is only used for new, uninitialized NAT state. | |
518 | * Returns either NF_ACCEPT or NF_DROP. | |
519 | */ | |
520 | static int ovs_ct_nat_execute(struct sk_buff *skb, struct nf_conn *ct, | |
521 | enum ip_conntrack_info ctinfo, | |
522 | const struct nf_nat_range *range, | |
523 | enum nf_nat_manip_type maniptype) | |
524 | { | |
525 | int hooknum, nh_off, err = NF_ACCEPT; | |
526 | ||
527 | nh_off = skb_network_offset(skb); | |
528 | skb_pull(skb, nh_off); | |
529 | ||
530 | /* See HOOK2MANIP(). */ | |
531 | if (maniptype == NF_NAT_MANIP_SRC) | |
532 | hooknum = NF_INET_LOCAL_IN; /* Source NAT */ | |
533 | else | |
534 | hooknum = NF_INET_LOCAL_OUT; /* Destination NAT */ | |
535 | ||
536 | switch (ctinfo) { | |
537 | case IP_CT_RELATED: | |
538 | case IP_CT_RELATED_REPLY: | |
99b7248e AB |
539 | if (IS_ENABLED(CONFIG_NF_NAT_IPV4) && |
540 | skb->protocol == htons(ETH_P_IP) && | |
05752523 JR |
541 | ip_hdr(skb)->protocol == IPPROTO_ICMP) { |
542 | if (!nf_nat_icmp_reply_translation(skb, ct, ctinfo, | |
543 | hooknum)) | |
544 | err = NF_DROP; | |
545 | goto push; | |
99b7248e AB |
546 | } else if (IS_ENABLED(CONFIG_NF_NAT_IPV6) && |
547 | skb->protocol == htons(ETH_P_IPV6)) { | |
05752523 JR |
548 | __be16 frag_off; |
549 | u8 nexthdr = ipv6_hdr(skb)->nexthdr; | |
550 | int hdrlen = ipv6_skip_exthdr(skb, | |
551 | sizeof(struct ipv6hdr), | |
552 | &nexthdr, &frag_off); | |
553 | ||
554 | if (hdrlen >= 0 && nexthdr == IPPROTO_ICMPV6) { | |
555 | if (!nf_nat_icmpv6_reply_translation(skb, ct, | |
556 | ctinfo, | |
557 | hooknum, | |
558 | hdrlen)) | |
559 | err = NF_DROP; | |
560 | goto push; | |
561 | } | |
05752523 JR |
562 | } |
563 | /* Non-ICMP, fall thru to initialize if needed. */ | |
564 | case IP_CT_NEW: | |
565 | /* Seen it before? This can happen for loopback, retrans, | |
566 | * or local packets. | |
567 | */ | |
568 | if (!nf_nat_initialized(ct, maniptype)) { | |
569 | /* Initialize according to the NAT action. */ | |
570 | err = (range && range->flags & NF_NAT_RANGE_MAP_IPS) | |
571 | /* Action is set up to establish a new | |
572 | * mapping. | |
573 | */ | |
574 | ? nf_nat_setup_info(ct, range, maniptype) | |
575 | : nf_nat_alloc_null_binding(ct, hooknum); | |
576 | if (err != NF_ACCEPT) | |
577 | goto push; | |
578 | } | |
579 | break; | |
580 | ||
581 | case IP_CT_ESTABLISHED: | |
582 | case IP_CT_ESTABLISHED_REPLY: | |
583 | break; | |
584 | ||
585 | default: | |
586 | err = NF_DROP; | |
587 | goto push; | |
588 | } | |
589 | ||
590 | err = nf_nat_packet(ct, ctinfo, hooknum, skb); | |
591 | push: | |
592 | skb_push(skb, nh_off); | |
593 | ||
594 | return err; | |
595 | } | |
596 | ||
597 | static void ovs_nat_update_key(struct sw_flow_key *key, | |
598 | const struct sk_buff *skb, | |
599 | enum nf_nat_manip_type maniptype) | |
600 | { | |
601 | if (maniptype == NF_NAT_MANIP_SRC) { | |
602 | __be16 src; | |
603 | ||
604 | key->ct.state |= OVS_CS_F_SRC_NAT; | |
605 | if (key->eth.type == htons(ETH_P_IP)) | |
606 | key->ipv4.addr.src = ip_hdr(skb)->saddr; | |
607 | else if (key->eth.type == htons(ETH_P_IPV6)) | |
608 | memcpy(&key->ipv6.addr.src, &ipv6_hdr(skb)->saddr, | |
609 | sizeof(key->ipv6.addr.src)); | |
610 | else | |
611 | return; | |
612 | ||
613 | if (key->ip.proto == IPPROTO_UDP) | |
614 | src = udp_hdr(skb)->source; | |
615 | else if (key->ip.proto == IPPROTO_TCP) | |
616 | src = tcp_hdr(skb)->source; | |
617 | else if (key->ip.proto == IPPROTO_SCTP) | |
618 | src = sctp_hdr(skb)->source; | |
619 | else | |
620 | return; | |
621 | ||
622 | key->tp.src = src; | |
623 | } else { | |
624 | __be16 dst; | |
625 | ||
626 | key->ct.state |= OVS_CS_F_DST_NAT; | |
627 | if (key->eth.type == htons(ETH_P_IP)) | |
628 | key->ipv4.addr.dst = ip_hdr(skb)->daddr; | |
629 | else if (key->eth.type == htons(ETH_P_IPV6)) | |
630 | memcpy(&key->ipv6.addr.dst, &ipv6_hdr(skb)->daddr, | |
631 | sizeof(key->ipv6.addr.dst)); | |
632 | else | |
633 | return; | |
634 | ||
635 | if (key->ip.proto == IPPROTO_UDP) | |
636 | dst = udp_hdr(skb)->dest; | |
637 | else if (key->ip.proto == IPPROTO_TCP) | |
638 | dst = tcp_hdr(skb)->dest; | |
639 | else if (key->ip.proto == IPPROTO_SCTP) | |
640 | dst = sctp_hdr(skb)->dest; | |
641 | else | |
642 | return; | |
643 | ||
644 | key->tp.dst = dst; | |
645 | } | |
646 | } | |
647 | ||
648 | /* Returns NF_DROP if the packet should be dropped, NF_ACCEPT otherwise. */ | |
649 | static int ovs_ct_nat(struct net *net, struct sw_flow_key *key, | |
650 | const struct ovs_conntrack_info *info, | |
651 | struct sk_buff *skb, struct nf_conn *ct, | |
652 | enum ip_conntrack_info ctinfo) | |
653 | { | |
654 | enum nf_nat_manip_type maniptype; | |
655 | int err; | |
656 | ||
657 | if (nf_ct_is_untracked(ct)) { | |
658 | /* A NAT action may only be performed on tracked packets. */ | |
659 | return NF_ACCEPT; | |
660 | } | |
661 | ||
662 | /* Add NAT extension if not confirmed yet. */ | |
663 | if (!nf_ct_is_confirmed(ct) && !nf_ct_nat_ext_add(ct)) | |
664 | return NF_ACCEPT; /* Can't NAT. */ | |
665 | ||
666 | /* Determine NAT type. | |
667 | * Check if the NAT type can be deduced from the tracked connection. | |
5745b0be JR |
668 | * Make sure new expected connections (IP_CT_RELATED) are NATted only |
669 | * when committing. | |
05752523 JR |
670 | */ |
671 | if (info->nat & OVS_CT_NAT && ctinfo != IP_CT_NEW && | |
672 | ct->status & IPS_NAT_MASK && | |
5745b0be | 673 | (ctinfo != IP_CT_RELATED || info->commit)) { |
05752523 JR |
674 | /* NAT an established or related connection like before. */ |
675 | if (CTINFO2DIR(ctinfo) == IP_CT_DIR_REPLY) | |
676 | /* This is the REPLY direction for a connection | |
677 | * for which NAT was applied in the forward | |
678 | * direction. Do the reverse NAT. | |
679 | */ | |
680 | maniptype = ct->status & IPS_SRC_NAT | |
681 | ? NF_NAT_MANIP_DST : NF_NAT_MANIP_SRC; | |
682 | else | |
683 | maniptype = ct->status & IPS_SRC_NAT | |
684 | ? NF_NAT_MANIP_SRC : NF_NAT_MANIP_DST; | |
685 | } else if (info->nat & OVS_CT_SRC_NAT) { | |
686 | maniptype = NF_NAT_MANIP_SRC; | |
687 | } else if (info->nat & OVS_CT_DST_NAT) { | |
688 | maniptype = NF_NAT_MANIP_DST; | |
689 | } else { | |
690 | return NF_ACCEPT; /* Connection is not NATed. */ | |
691 | } | |
692 | err = ovs_ct_nat_execute(skb, ct, ctinfo, &info->range, maniptype); | |
693 | ||
694 | /* Mark NAT done if successful and update the flow key. */ | |
695 | if (err == NF_ACCEPT) | |
696 | ovs_nat_update_key(key, skb, maniptype); | |
697 | ||
698 | return err; | |
699 | } | |
700 | #else /* !CONFIG_NF_NAT_NEEDED */ | |
701 | static int ovs_ct_nat(struct net *net, struct sw_flow_key *key, | |
702 | const struct ovs_conntrack_info *info, | |
703 | struct sk_buff *skb, struct nf_conn *ct, | |
704 | enum ip_conntrack_info ctinfo) | |
705 | { | |
706 | return NF_ACCEPT; | |
707 | } | |
708 | #endif | |
709 | ||
9f13ded8 | 710 | /* Pass 'skb' through conntrack in 'net', using zone configured in 'info', if |
394e910e JR |
711 | * not done already. Update key with new CT state after passing the packet |
712 | * through conntrack. | |
9f13ded8 JR |
713 | * Note that if the packet is deemed invalid by conntrack, skb->nfct will be |
714 | * set to NULL and 0 will be returned. | |
715 | */ | |
4f0909ee | 716 | static int __ovs_ct_lookup(struct net *net, struct sw_flow_key *key, |
7f8a436e JS |
717 | const struct ovs_conntrack_info *info, |
718 | struct sk_buff *skb) | |
719 | { | |
720 | /* If we are recirculating packets to match on conntrack fields and | |
721 | * committing with a separate conntrack action, then we don't need to | |
722 | * actually run the packet through conntrack twice unless it's for a | |
723 | * different zone. | |
724 | */ | |
28b6e0c1 JR |
725 | bool cached = skb_nfct_cached(net, key, info, skb); |
726 | enum ip_conntrack_info ctinfo; | |
727 | struct nf_conn *ct; | |
728 | ||
729 | if (!cached) { | |
7f8a436e | 730 | struct nf_conn *tmpl = info->ct; |
5b6b9293 | 731 | int err; |
7f8a436e JS |
732 | |
733 | /* Associate skb with specified zone. */ | |
734 | if (tmpl) { | |
735 | if (skb->nfct) | |
736 | nf_conntrack_put(skb->nfct); | |
737 | nf_conntrack_get(&tmpl->ct_general); | |
738 | skb->nfct = &tmpl->ct_general; | |
739 | skb->nfctinfo = IP_CT_NEW; | |
740 | } | |
741 | ||
5b6b9293 JR |
742 | /* Repeat if requested, see nf_iterate(). */ |
743 | do { | |
744 | err = nf_conntrack_in(net, info->family, | |
745 | NF_INET_PRE_ROUTING, skb); | |
746 | } while (err == NF_REPEAT); | |
747 | ||
748 | if (err != NF_ACCEPT) | |
7f8a436e | 749 | return -ENOENT; |
cae3a262 | 750 | |
05752523 JR |
751 | /* Clear CT state NAT flags to mark that we have not yet done |
752 | * NAT after the nf_conntrack_in() call. We can actually clear | |
753 | * the whole state, as it will be re-initialized below. | |
754 | */ | |
755 | key->ct.state = 0; | |
756 | ||
757 | /* Update the key, but keep the NAT flags. */ | |
758 | ovs_ct_update_key(skb, info, key, true, true); | |
28b6e0c1 | 759 | } |
394e910e | 760 | |
28b6e0c1 | 761 | ct = nf_ct_get(skb, &ctinfo); |
05752523 JR |
762 | if (ct) { |
763 | /* Packets starting a new connection must be NATted before the | |
764 | * helper, so that the helper knows about the NAT. We enforce | |
765 | * this by delaying both NAT and helper calls for unconfirmed | |
766 | * connections until the committing CT action. For later | |
767 | * packets NAT and Helper may be called in either order. | |
768 | * | |
769 | * NAT will be done only if the CT action has NAT, and only | |
770 | * once per packet (per zone), as guarded by the NAT bits in | |
771 | * the key->ct.state. | |
772 | */ | |
773 | if (info->nat && !(key->ct.state & OVS_CS_F_NAT_MASK) && | |
774 | (nf_ct_is_confirmed(ct) || info->commit) && | |
775 | ovs_ct_nat(net, key, info, skb, ct, ctinfo) != NF_ACCEPT) { | |
776 | return -EINVAL; | |
777 | } | |
778 | ||
779 | /* Call the helper only if: | |
780 | * - nf_conntrack_in() was executed above ("!cached") for a | |
781 | * confirmed connection, or | |
782 | * - When committing an unconfirmed connection. | |
783 | */ | |
784 | if ((nf_ct_is_confirmed(ct) ? !cached : info->commit) && | |
785 | ovs_ct_helper(skb, info->family) != NF_ACCEPT) { | |
786 | return -EINVAL; | |
787 | } | |
7f8a436e JS |
788 | } |
789 | ||
790 | return 0; | |
791 | } | |
792 | ||
793 | /* Lookup connection and read fields into key. */ | |
794 | static int ovs_ct_lookup(struct net *net, struct sw_flow_key *key, | |
795 | const struct ovs_conntrack_info *info, | |
796 | struct sk_buff *skb) | |
797 | { | |
798 | struct nf_conntrack_expect *exp; | |
799 | ||
9f13ded8 JR |
800 | /* If we pass an expected packet through nf_conntrack_in() the |
801 | * expectation is typically removed, but the packet could still be | |
802 | * lost in upcall processing. To prevent this from happening we | |
803 | * perform an explicit expectation lookup. Expected connections are | |
804 | * always new, and will be passed through conntrack only when they are | |
805 | * committed, as it is OK to remove the expectation at that time. | |
806 | */ | |
7f8a436e JS |
807 | exp = ovs_ct_expect_find(net, &info->zone, info->family, skb); |
808 | if (exp) { | |
809 | u8 state; | |
810 | ||
05752523 JR |
811 | /* NOTE: New connections are NATted and Helped only when |
812 | * committed, so we are not calling into NAT here. | |
813 | */ | |
7f8a436e | 814 | state = OVS_CS_F_TRACKED | OVS_CS_F_NEW | OVS_CS_F_RELATED; |
182e3042 | 815 | __ovs_ct_update_key(key, state, &info->zone, exp->master); |
05752523 JR |
816 | } else |
817 | return __ovs_ct_lookup(net, key, info, skb); | |
7f8a436e JS |
818 | |
819 | return 0; | |
820 | } | |
821 | ||
822 | /* Lookup connection and confirm if unconfirmed. */ | |
823 | static int ovs_ct_commit(struct net *net, struct sw_flow_key *key, | |
824 | const struct ovs_conntrack_info *info, | |
825 | struct sk_buff *skb) | |
826 | { | |
7f8a436e JS |
827 | int err; |
828 | ||
7f8a436e JS |
829 | err = __ovs_ct_lookup(net, key, info, skb); |
830 | if (err) | |
831 | return err; | |
9f13ded8 | 832 | /* This is a no-op if the connection has already been confirmed. */ |
7f8a436e JS |
833 | if (nf_conntrack_confirm(skb) != NF_ACCEPT) |
834 | return -EINVAL; | |
835 | ||
7f8a436e JS |
836 | return 0; |
837 | } | |
838 | ||
33db4125 | 839 | static bool labels_nonzero(const struct ovs_key_ct_labels *labels) |
c2ac6673 JS |
840 | { |
841 | size_t i; | |
842 | ||
33db4125 JS |
843 | for (i = 0; i < sizeof(*labels); i++) |
844 | if (labels->ct_labels[i]) | |
c2ac6673 JS |
845 | return true; |
846 | ||
847 | return false; | |
848 | } | |
849 | ||
74c16618 JS |
850 | /* Returns 0 on success, -EINPROGRESS if 'skb' is stolen, or other nonzero |
851 | * value if 'skb' is freed. | |
852 | */ | |
7f8a436e JS |
853 | int ovs_ct_execute(struct net *net, struct sk_buff *skb, |
854 | struct sw_flow_key *key, | |
855 | const struct ovs_conntrack_info *info) | |
856 | { | |
857 | int nh_ofs; | |
858 | int err; | |
859 | ||
860 | /* The conntrack module expects to be working at L3. */ | |
861 | nh_ofs = skb_network_offset(skb); | |
862 | skb_pull(skb, nh_ofs); | |
863 | ||
864 | if (key->ip.frag != OVS_FRAG_TYPE_NONE) { | |
865 | err = handle_fragments(net, key, info->zone.id, skb); | |
866 | if (err) | |
867 | return err; | |
868 | } | |
869 | ||
ab38a7b5 | 870 | if (info->commit) |
7f8a436e JS |
871 | err = ovs_ct_commit(net, key, info, skb); |
872 | else | |
873 | err = ovs_ct_lookup(net, key, info, skb); | |
182e3042 JS |
874 | if (err) |
875 | goto err; | |
7f8a436e | 876 | |
c2ac6673 | 877 | if (info->mark.mask) { |
182e3042 JS |
878 | err = ovs_ct_set_mark(skb, key, info->mark.value, |
879 | info->mark.mask); | |
c2ac6673 JS |
880 | if (err) |
881 | goto err; | |
882 | } | |
33db4125 JS |
883 | if (labels_nonzero(&info->labels.mask)) |
884 | err = ovs_ct_set_labels(skb, key, &info->labels.value, | |
885 | &info->labels.mask); | |
182e3042 | 886 | err: |
7f8a436e | 887 | skb_push(skb, nh_ofs); |
74c16618 JS |
888 | if (err) |
889 | kfree_skb(skb); | |
7f8a436e JS |
890 | return err; |
891 | } | |
892 | ||
cae3a262 JS |
893 | static int ovs_ct_add_helper(struct ovs_conntrack_info *info, const char *name, |
894 | const struct sw_flow_key *key, bool log) | |
895 | { | |
896 | struct nf_conntrack_helper *helper; | |
897 | struct nf_conn_help *help; | |
898 | ||
899 | helper = nf_conntrack_helper_try_module_get(name, info->family, | |
900 | key->ip.proto); | |
901 | if (!helper) { | |
902 | OVS_NLERR(log, "Unknown helper \"%s\"", name); | |
903 | return -EINVAL; | |
904 | } | |
905 | ||
906 | help = nf_ct_helper_ext_add(info->ct, helper, GFP_KERNEL); | |
907 | if (!help) { | |
908 | module_put(helper->me); | |
909 | return -ENOMEM; | |
910 | } | |
911 | ||
912 | rcu_assign_pointer(help->helper, helper); | |
913 | info->helper = helper; | |
914 | return 0; | |
915 | } | |
916 | ||
05752523 JR |
917 | #ifdef CONFIG_NF_NAT_NEEDED |
918 | static int parse_nat(const struct nlattr *attr, | |
919 | struct ovs_conntrack_info *info, bool log) | |
920 | { | |
921 | struct nlattr *a; | |
922 | int rem; | |
923 | bool have_ip_max = false; | |
924 | bool have_proto_max = false; | |
925 | bool ip_vers = (info->family == NFPROTO_IPV6); | |
926 | ||
927 | nla_for_each_nested(a, attr, rem) { | |
928 | static const int ovs_nat_attr_lens[OVS_NAT_ATTR_MAX + 1][2] = { | |
929 | [OVS_NAT_ATTR_SRC] = {0, 0}, | |
930 | [OVS_NAT_ATTR_DST] = {0, 0}, | |
931 | [OVS_NAT_ATTR_IP_MIN] = {sizeof(struct in_addr), | |
932 | sizeof(struct in6_addr)}, | |
933 | [OVS_NAT_ATTR_IP_MAX] = {sizeof(struct in_addr), | |
934 | sizeof(struct in6_addr)}, | |
935 | [OVS_NAT_ATTR_PROTO_MIN] = {sizeof(u16), sizeof(u16)}, | |
936 | [OVS_NAT_ATTR_PROTO_MAX] = {sizeof(u16), sizeof(u16)}, | |
937 | [OVS_NAT_ATTR_PERSISTENT] = {0, 0}, | |
938 | [OVS_NAT_ATTR_PROTO_HASH] = {0, 0}, | |
939 | [OVS_NAT_ATTR_PROTO_RANDOM] = {0, 0}, | |
940 | }; | |
941 | int type = nla_type(a); | |
942 | ||
943 | if (type > OVS_NAT_ATTR_MAX) { | |
944 | OVS_NLERR(log, | |
945 | "Unknown NAT attribute (type=%d, max=%d).\n", | |
946 | type, OVS_NAT_ATTR_MAX); | |
947 | return -EINVAL; | |
948 | } | |
949 | ||
950 | if (nla_len(a) != ovs_nat_attr_lens[type][ip_vers]) { | |
951 | OVS_NLERR(log, | |
952 | "NAT attribute type %d has unexpected length (%d != %d).\n", | |
953 | type, nla_len(a), | |
954 | ovs_nat_attr_lens[type][ip_vers]); | |
955 | return -EINVAL; | |
956 | } | |
957 | ||
958 | switch (type) { | |
959 | case OVS_NAT_ATTR_SRC: | |
960 | case OVS_NAT_ATTR_DST: | |
961 | if (info->nat) { | |
962 | OVS_NLERR(log, | |
963 | "Only one type of NAT may be specified.\n" | |
964 | ); | |
965 | return -ERANGE; | |
966 | } | |
967 | info->nat |= OVS_CT_NAT; | |
968 | info->nat |= ((type == OVS_NAT_ATTR_SRC) | |
969 | ? OVS_CT_SRC_NAT : OVS_CT_DST_NAT); | |
970 | break; | |
971 | ||
972 | case OVS_NAT_ATTR_IP_MIN: | |
ac71b46e HY |
973 | nla_memcpy(&info->range.min_addr, a, |
974 | sizeof(info->range.min_addr)); | |
05752523 JR |
975 | info->range.flags |= NF_NAT_RANGE_MAP_IPS; |
976 | break; | |
977 | ||
978 | case OVS_NAT_ATTR_IP_MAX: | |
979 | have_ip_max = true; | |
980 | nla_memcpy(&info->range.max_addr, a, | |
981 | sizeof(info->range.max_addr)); | |
982 | info->range.flags |= NF_NAT_RANGE_MAP_IPS; | |
983 | break; | |
984 | ||
985 | case OVS_NAT_ATTR_PROTO_MIN: | |
986 | info->range.min_proto.all = htons(nla_get_u16(a)); | |
987 | info->range.flags |= NF_NAT_RANGE_PROTO_SPECIFIED; | |
988 | break; | |
989 | ||
990 | case OVS_NAT_ATTR_PROTO_MAX: | |
991 | have_proto_max = true; | |
992 | info->range.max_proto.all = htons(nla_get_u16(a)); | |
993 | info->range.flags |= NF_NAT_RANGE_PROTO_SPECIFIED; | |
994 | break; | |
995 | ||
996 | case OVS_NAT_ATTR_PERSISTENT: | |
997 | info->range.flags |= NF_NAT_RANGE_PERSISTENT; | |
998 | break; | |
999 | ||
1000 | case OVS_NAT_ATTR_PROTO_HASH: | |
1001 | info->range.flags |= NF_NAT_RANGE_PROTO_RANDOM; | |
1002 | break; | |
1003 | ||
1004 | case OVS_NAT_ATTR_PROTO_RANDOM: | |
1005 | info->range.flags |= NF_NAT_RANGE_PROTO_RANDOM_FULLY; | |
1006 | break; | |
1007 | ||
1008 | default: | |
1009 | OVS_NLERR(log, "Unknown nat attribute (%d).\n", type); | |
1010 | return -EINVAL; | |
1011 | } | |
1012 | } | |
1013 | ||
1014 | if (rem > 0) { | |
1015 | OVS_NLERR(log, "NAT attribute has %d unknown bytes.\n", rem); | |
1016 | return -EINVAL; | |
1017 | } | |
1018 | if (!info->nat) { | |
1019 | /* Do not allow flags if no type is given. */ | |
1020 | if (info->range.flags) { | |
1021 | OVS_NLERR(log, | |
1022 | "NAT flags may be given only when NAT range (SRC or DST) is also specified.\n" | |
1023 | ); | |
1024 | return -EINVAL; | |
1025 | } | |
1026 | info->nat = OVS_CT_NAT; /* NAT existing connections. */ | |
1027 | } else if (!info->commit) { | |
1028 | OVS_NLERR(log, | |
1029 | "NAT attributes may be specified only when CT COMMIT flag is also specified.\n" | |
1030 | ); | |
1031 | return -EINVAL; | |
1032 | } | |
1033 | /* Allow missing IP_MAX. */ | |
1034 | if (info->range.flags & NF_NAT_RANGE_MAP_IPS && !have_ip_max) { | |
1035 | memcpy(&info->range.max_addr, &info->range.min_addr, | |
1036 | sizeof(info->range.max_addr)); | |
1037 | } | |
1038 | /* Allow missing PROTO_MAX. */ | |
1039 | if (info->range.flags & NF_NAT_RANGE_PROTO_SPECIFIED && | |
1040 | !have_proto_max) { | |
1041 | info->range.max_proto.all = info->range.min_proto.all; | |
1042 | } | |
1043 | return 0; | |
1044 | } | |
1045 | #endif | |
1046 | ||
7f8a436e | 1047 | static const struct ovs_ct_len_tbl ovs_ct_attr_lens[OVS_CT_ATTR_MAX + 1] = { |
ab38a7b5 | 1048 | [OVS_CT_ATTR_COMMIT] = { .minlen = 0, .maxlen = 0 }, |
7f8a436e JS |
1049 | [OVS_CT_ATTR_ZONE] = { .minlen = sizeof(u16), |
1050 | .maxlen = sizeof(u16) }, | |
182e3042 JS |
1051 | [OVS_CT_ATTR_MARK] = { .minlen = sizeof(struct md_mark), |
1052 | .maxlen = sizeof(struct md_mark) }, | |
33db4125 JS |
1053 | [OVS_CT_ATTR_LABELS] = { .minlen = sizeof(struct md_labels), |
1054 | .maxlen = sizeof(struct md_labels) }, | |
cae3a262 | 1055 | [OVS_CT_ATTR_HELPER] = { .minlen = 1, |
05752523 JR |
1056 | .maxlen = NF_CT_HELPER_NAME_LEN }, |
1057 | #ifdef CONFIG_NF_NAT_NEEDED | |
1058 | /* NAT length is checked when parsing the nested attributes. */ | |
1059 | [OVS_CT_ATTR_NAT] = { .minlen = 0, .maxlen = INT_MAX }, | |
1060 | #endif | |
7f8a436e JS |
1061 | }; |
1062 | ||
1063 | static int parse_ct(const struct nlattr *attr, struct ovs_conntrack_info *info, | |
cae3a262 | 1064 | const char **helper, bool log) |
7f8a436e JS |
1065 | { |
1066 | struct nlattr *a; | |
1067 | int rem; | |
1068 | ||
1069 | nla_for_each_nested(a, attr, rem) { | |
1070 | int type = nla_type(a); | |
1071 | int maxlen = ovs_ct_attr_lens[type].maxlen; | |
1072 | int minlen = ovs_ct_attr_lens[type].minlen; | |
1073 | ||
1074 | if (type > OVS_CT_ATTR_MAX) { | |
1075 | OVS_NLERR(log, | |
1076 | "Unknown conntrack attr (type=%d, max=%d)", | |
1077 | type, OVS_CT_ATTR_MAX); | |
1078 | return -EINVAL; | |
1079 | } | |
1080 | if (nla_len(a) < minlen || nla_len(a) > maxlen) { | |
1081 | OVS_NLERR(log, | |
1082 | "Conntrack attr type has unexpected length (type=%d, length=%d, expected=%d)", | |
1083 | type, nla_len(a), maxlen); | |
1084 | return -EINVAL; | |
1085 | } | |
1086 | ||
1087 | switch (type) { | |
ab38a7b5 JS |
1088 | case OVS_CT_ATTR_COMMIT: |
1089 | info->commit = true; | |
7f8a436e JS |
1090 | break; |
1091 | #ifdef CONFIG_NF_CONNTRACK_ZONES | |
1092 | case OVS_CT_ATTR_ZONE: | |
1093 | info->zone.id = nla_get_u16(a); | |
1094 | break; | |
182e3042 JS |
1095 | #endif |
1096 | #ifdef CONFIG_NF_CONNTRACK_MARK | |
1097 | case OVS_CT_ATTR_MARK: { | |
1098 | struct md_mark *mark = nla_data(a); | |
1099 | ||
e754ec69 JS |
1100 | if (!mark->mask) { |
1101 | OVS_NLERR(log, "ct_mark mask cannot be 0"); | |
1102 | return -EINVAL; | |
1103 | } | |
182e3042 JS |
1104 | info->mark = *mark; |
1105 | break; | |
1106 | } | |
c2ac6673 JS |
1107 | #endif |
1108 | #ifdef CONFIG_NF_CONNTRACK_LABELS | |
33db4125 JS |
1109 | case OVS_CT_ATTR_LABELS: { |
1110 | struct md_labels *labels = nla_data(a); | |
c2ac6673 | 1111 | |
e754ec69 JS |
1112 | if (!labels_nonzero(&labels->mask)) { |
1113 | OVS_NLERR(log, "ct_labels mask cannot be 0"); | |
1114 | return -EINVAL; | |
1115 | } | |
33db4125 | 1116 | info->labels = *labels; |
c2ac6673 JS |
1117 | break; |
1118 | } | |
7f8a436e | 1119 | #endif |
cae3a262 JS |
1120 | case OVS_CT_ATTR_HELPER: |
1121 | *helper = nla_data(a); | |
1122 | if (!memchr(*helper, '\0', nla_len(a))) { | |
1123 | OVS_NLERR(log, "Invalid conntrack helper"); | |
1124 | return -EINVAL; | |
1125 | } | |
1126 | break; | |
05752523 JR |
1127 | #ifdef CONFIG_NF_NAT_NEEDED |
1128 | case OVS_CT_ATTR_NAT: { | |
1129 | int err = parse_nat(a, info, log); | |
1130 | ||
1131 | if (err) | |
1132 | return err; | |
1133 | break; | |
1134 | } | |
1135 | #endif | |
7f8a436e JS |
1136 | default: |
1137 | OVS_NLERR(log, "Unknown conntrack attr (%d)", | |
1138 | type); | |
1139 | return -EINVAL; | |
1140 | } | |
1141 | } | |
1142 | ||
1143 | if (rem > 0) { | |
1144 | OVS_NLERR(log, "Conntrack attr has %d unknown bytes", rem); | |
1145 | return -EINVAL; | |
1146 | } | |
1147 | ||
1148 | return 0; | |
1149 | } | |
1150 | ||
c2ac6673 | 1151 | bool ovs_ct_verify(struct net *net, enum ovs_key_attr attr) |
7f8a436e JS |
1152 | { |
1153 | if (attr == OVS_KEY_ATTR_CT_STATE) | |
1154 | return true; | |
1155 | if (IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES) && | |
1156 | attr == OVS_KEY_ATTR_CT_ZONE) | |
1157 | return true; | |
182e3042 JS |
1158 | if (IS_ENABLED(CONFIG_NF_CONNTRACK_MARK) && |
1159 | attr == OVS_KEY_ATTR_CT_MARK) | |
1160 | return true; | |
c2ac6673 | 1161 | if (IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS) && |
33db4125 | 1162 | attr == OVS_KEY_ATTR_CT_LABELS) { |
c2ac6673 JS |
1163 | struct ovs_net *ovs_net = net_generic(net, ovs_net_id); |
1164 | ||
1165 | return ovs_net->xt_label; | |
1166 | } | |
7f8a436e JS |
1167 | |
1168 | return false; | |
1169 | } | |
1170 | ||
1171 | int ovs_ct_copy_action(struct net *net, const struct nlattr *attr, | |
1172 | const struct sw_flow_key *key, | |
1173 | struct sw_flow_actions **sfa, bool log) | |
1174 | { | |
1175 | struct ovs_conntrack_info ct_info; | |
cae3a262 | 1176 | const char *helper = NULL; |
7f8a436e JS |
1177 | u16 family; |
1178 | int err; | |
1179 | ||
1180 | family = key_to_nfproto(key); | |
1181 | if (family == NFPROTO_UNSPEC) { | |
1182 | OVS_NLERR(log, "ct family unspecified"); | |
1183 | return -EINVAL; | |
1184 | } | |
1185 | ||
1186 | memset(&ct_info, 0, sizeof(ct_info)); | |
1187 | ct_info.family = family; | |
1188 | ||
1189 | nf_ct_zone_init(&ct_info.zone, NF_CT_DEFAULT_ZONE_ID, | |
1190 | NF_CT_DEFAULT_ZONE_DIR, 0); | |
1191 | ||
cae3a262 | 1192 | err = parse_ct(attr, &ct_info, &helper, log); |
7f8a436e JS |
1193 | if (err) |
1194 | return err; | |
1195 | ||
1196 | /* Set up template for tracking connections in specific zones. */ | |
1197 | ct_info.ct = nf_ct_tmpl_alloc(net, &ct_info.zone, GFP_KERNEL); | |
1198 | if (!ct_info.ct) { | |
1199 | OVS_NLERR(log, "Failed to allocate conntrack template"); | |
1200 | return -ENOMEM; | |
1201 | } | |
90c7afc9 JS |
1202 | |
1203 | __set_bit(IPS_CONFIRMED_BIT, &ct_info.ct->status); | |
1204 | nf_conntrack_get(&ct_info.ct->ct_general); | |
1205 | ||
cae3a262 JS |
1206 | if (helper) { |
1207 | err = ovs_ct_add_helper(&ct_info, helper, key, log); | |
1208 | if (err) | |
1209 | goto err_free_ct; | |
1210 | } | |
7f8a436e JS |
1211 | |
1212 | err = ovs_nla_add_action(sfa, OVS_ACTION_ATTR_CT, &ct_info, | |
1213 | sizeof(ct_info), log); | |
1214 | if (err) | |
1215 | goto err_free_ct; | |
1216 | ||
7f8a436e JS |
1217 | return 0; |
1218 | err_free_ct: | |
2f3ab9f9 | 1219 | __ovs_ct_free_action(&ct_info); |
7f8a436e JS |
1220 | return err; |
1221 | } | |
1222 | ||
05752523 JR |
1223 | #ifdef CONFIG_NF_NAT_NEEDED |
1224 | static bool ovs_ct_nat_to_attr(const struct ovs_conntrack_info *info, | |
1225 | struct sk_buff *skb) | |
1226 | { | |
1227 | struct nlattr *start; | |
1228 | ||
1229 | start = nla_nest_start(skb, OVS_CT_ATTR_NAT); | |
1230 | if (!start) | |
1231 | return false; | |
1232 | ||
1233 | if (info->nat & OVS_CT_SRC_NAT) { | |
1234 | if (nla_put_flag(skb, OVS_NAT_ATTR_SRC)) | |
1235 | return false; | |
1236 | } else if (info->nat & OVS_CT_DST_NAT) { | |
1237 | if (nla_put_flag(skb, OVS_NAT_ATTR_DST)) | |
1238 | return false; | |
1239 | } else { | |
1240 | goto out; | |
1241 | } | |
1242 | ||
1243 | if (info->range.flags & NF_NAT_RANGE_MAP_IPS) { | |
99b7248e AB |
1244 | if (IS_ENABLED(CONFIG_NF_NAT_IPV4) && |
1245 | info->family == NFPROTO_IPV4) { | |
05752523 JR |
1246 | if (nla_put_in_addr(skb, OVS_NAT_ATTR_IP_MIN, |
1247 | info->range.min_addr.ip) || | |
1248 | (info->range.max_addr.ip | |
1249 | != info->range.min_addr.ip && | |
1250 | (nla_put_in_addr(skb, OVS_NAT_ATTR_IP_MAX, | |
1251 | info->range.max_addr.ip)))) | |
1252 | return false; | |
99b7248e AB |
1253 | } else if (IS_ENABLED(CONFIG_NF_NAT_IPV6) && |
1254 | info->family == NFPROTO_IPV6) { | |
05752523 JR |
1255 | if (nla_put_in6_addr(skb, OVS_NAT_ATTR_IP_MIN, |
1256 | &info->range.min_addr.in6) || | |
1257 | (memcmp(&info->range.max_addr.in6, | |
1258 | &info->range.min_addr.in6, | |
1259 | sizeof(info->range.max_addr.in6)) && | |
1260 | (nla_put_in6_addr(skb, OVS_NAT_ATTR_IP_MAX, | |
1261 | &info->range.max_addr.in6)))) | |
1262 | return false; | |
05752523 JR |
1263 | } else { |
1264 | return false; | |
1265 | } | |
1266 | } | |
1267 | if (info->range.flags & NF_NAT_RANGE_PROTO_SPECIFIED && | |
1268 | (nla_put_u16(skb, OVS_NAT_ATTR_PROTO_MIN, | |
1269 | ntohs(info->range.min_proto.all)) || | |
1270 | (info->range.max_proto.all != info->range.min_proto.all && | |
1271 | nla_put_u16(skb, OVS_NAT_ATTR_PROTO_MAX, | |
1272 | ntohs(info->range.max_proto.all))))) | |
1273 | return false; | |
1274 | ||
1275 | if (info->range.flags & NF_NAT_RANGE_PERSISTENT && | |
1276 | nla_put_flag(skb, OVS_NAT_ATTR_PERSISTENT)) | |
1277 | return false; | |
1278 | if (info->range.flags & NF_NAT_RANGE_PROTO_RANDOM && | |
1279 | nla_put_flag(skb, OVS_NAT_ATTR_PROTO_HASH)) | |
1280 | return false; | |
1281 | if (info->range.flags & NF_NAT_RANGE_PROTO_RANDOM_FULLY && | |
1282 | nla_put_flag(skb, OVS_NAT_ATTR_PROTO_RANDOM)) | |
1283 | return false; | |
1284 | out: | |
1285 | nla_nest_end(skb, start); | |
1286 | ||
1287 | return true; | |
1288 | } | |
1289 | #endif | |
1290 | ||
7f8a436e JS |
1291 | int ovs_ct_action_to_attr(const struct ovs_conntrack_info *ct_info, |
1292 | struct sk_buff *skb) | |
1293 | { | |
1294 | struct nlattr *start; | |
1295 | ||
1296 | start = nla_nest_start(skb, OVS_ACTION_ATTR_CT); | |
1297 | if (!start) | |
1298 | return -EMSGSIZE; | |
1299 | ||
ab38a7b5 | 1300 | if (ct_info->commit && nla_put_flag(skb, OVS_CT_ATTR_COMMIT)) |
7f8a436e JS |
1301 | return -EMSGSIZE; |
1302 | if (IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES) && | |
1303 | nla_put_u16(skb, OVS_CT_ATTR_ZONE, ct_info->zone.id)) | |
1304 | return -EMSGSIZE; | |
e754ec69 | 1305 | if (IS_ENABLED(CONFIG_NF_CONNTRACK_MARK) && ct_info->mark.mask && |
182e3042 JS |
1306 | nla_put(skb, OVS_CT_ATTR_MARK, sizeof(ct_info->mark), |
1307 | &ct_info->mark)) | |
1308 | return -EMSGSIZE; | |
c2ac6673 | 1309 | if (IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS) && |
e754ec69 | 1310 | labels_nonzero(&ct_info->labels.mask) && |
33db4125 JS |
1311 | nla_put(skb, OVS_CT_ATTR_LABELS, sizeof(ct_info->labels), |
1312 | &ct_info->labels)) | |
c2ac6673 | 1313 | return -EMSGSIZE; |
cae3a262 JS |
1314 | if (ct_info->helper) { |
1315 | if (nla_put_string(skb, OVS_CT_ATTR_HELPER, | |
1316 | ct_info->helper->name)) | |
1317 | return -EMSGSIZE; | |
1318 | } | |
05752523 JR |
1319 | #ifdef CONFIG_NF_NAT_NEEDED |
1320 | if (ct_info->nat && !ovs_ct_nat_to_attr(ct_info, skb)) | |
1321 | return -EMSGSIZE; | |
1322 | #endif | |
7f8a436e JS |
1323 | nla_nest_end(skb, start); |
1324 | ||
1325 | return 0; | |
1326 | } | |
1327 | ||
1328 | void ovs_ct_free_action(const struct nlattr *a) | |
1329 | { | |
1330 | struct ovs_conntrack_info *ct_info = nla_data(a); | |
1331 | ||
2f3ab9f9 JS |
1332 | __ovs_ct_free_action(ct_info); |
1333 | } | |
1334 | ||
1335 | static void __ovs_ct_free_action(struct ovs_conntrack_info *ct_info) | |
1336 | { | |
cae3a262 JS |
1337 | if (ct_info->helper) |
1338 | module_put(ct_info->helper->me); | |
7f8a436e JS |
1339 | if (ct_info->ct) |
1340 | nf_ct_put(ct_info->ct); | |
1341 | } | |
c2ac6673 JS |
1342 | |
1343 | void ovs_ct_init(struct net *net) | |
1344 | { | |
33db4125 | 1345 | unsigned int n_bits = sizeof(struct ovs_key_ct_labels) * BITS_PER_BYTE; |
c2ac6673 JS |
1346 | struct ovs_net *ovs_net = net_generic(net, ovs_net_id); |
1347 | ||
1348 | if (nf_connlabels_get(net, n_bits)) { | |
1349 | ovs_net->xt_label = false; | |
1350 | OVS_NLERR(true, "Failed to set connlabel length"); | |
1351 | } else { | |
1352 | ovs_net->xt_label = true; | |
1353 | } | |
1354 | } | |
1355 | ||
1356 | void ovs_ct_exit(struct net *net) | |
1357 | { | |
1358 | struct ovs_net *ovs_net = net_generic(net, ovs_net_id); | |
1359 | ||
1360 | if (ovs_net->xt_label) | |
1361 | nf_connlabels_put(net); | |
1362 | } |