Loading...
1#include <linux/skbuff.h>
2#include <linux/export.h>
3#include <linux/ip.h>
4#include <linux/ipv6.h>
5#include <linux/if_vlan.h>
6#include <net/ip.h>
7#include <net/ipv6.h>
8#include <linux/igmp.h>
9#include <linux/icmp.h>
10#include <linux/sctp.h>
11#include <linux/dccp.h>
12#include <linux/if_tunnel.h>
13#include <linux/if_pppox.h>
14#include <linux/ppp_defs.h>
15#include <net/flow_keys.h>
16
17/* copy saddr & daddr, possibly using 64bit load/store
18 * Equivalent to : flow->src = iph->saddr;
19 * flow->dst = iph->daddr;
20 */
21static void iph_to_flow_copy_addrs(struct flow_keys *flow, const struct iphdr *iph)
22{
23 BUILD_BUG_ON(offsetof(typeof(*flow), dst) !=
24 offsetof(typeof(*flow), src) + sizeof(flow->src));
25 memcpy(&flow->src, &iph->saddr, sizeof(flow->src) + sizeof(flow->dst));
26}
27
28/**
29 * skb_flow_get_ports - extract the upper layer ports and return them
30 * @skb: buffer to extract the ports from
31 * @thoff: transport header offset
32 * @ip_proto: protocol for which to get port offset
33 *
34 * The function will try to retrieve the ports at offset thoff + poff where poff
35 * is the protocol port offset returned from proto_ports_offset
36 */
37__be32 skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto)
38{
39 int poff = proto_ports_offset(ip_proto);
40
41 if (poff >= 0) {
42 __be32 *ports, _ports;
43
44 ports = skb_header_pointer(skb, thoff + poff,
45 sizeof(_ports), &_ports);
46 if (ports)
47 return *ports;
48 }
49
50 return 0;
51}
52EXPORT_SYMBOL(skb_flow_get_ports);
53
54bool skb_flow_dissect(const struct sk_buff *skb, struct flow_keys *flow)
55{
56 int nhoff = skb_network_offset(skb);
57 u8 ip_proto;
58 __be16 proto = skb->protocol;
59
60 memset(flow, 0, sizeof(*flow));
61
62again:
63 switch (proto) {
64 case htons(ETH_P_IP): {
65 const struct iphdr *iph;
66 struct iphdr _iph;
67ip:
68 iph = skb_header_pointer(skb, nhoff, sizeof(_iph), &_iph);
69 if (!iph || iph->ihl < 5)
70 return false;
71 nhoff += iph->ihl * 4;
72
73 ip_proto = iph->protocol;
74 if (ip_is_fragment(iph))
75 ip_proto = 0;
76
77 iph_to_flow_copy_addrs(flow, iph);
78 break;
79 }
80 case htons(ETH_P_IPV6): {
81 const struct ipv6hdr *iph;
82 struct ipv6hdr _iph;
83ipv6:
84 iph = skb_header_pointer(skb, nhoff, sizeof(_iph), &_iph);
85 if (!iph)
86 return false;
87
88 ip_proto = iph->nexthdr;
89 flow->src = (__force __be32)ipv6_addr_hash(&iph->saddr);
90 flow->dst = (__force __be32)ipv6_addr_hash(&iph->daddr);
91 nhoff += sizeof(struct ipv6hdr);
92 break;
93 }
94 case htons(ETH_P_8021AD):
95 case htons(ETH_P_8021Q): {
96 const struct vlan_hdr *vlan;
97 struct vlan_hdr _vlan;
98
99 vlan = skb_header_pointer(skb, nhoff, sizeof(_vlan), &_vlan);
100 if (!vlan)
101 return false;
102
103 proto = vlan->h_vlan_encapsulated_proto;
104 nhoff += sizeof(*vlan);
105 goto again;
106 }
107 case htons(ETH_P_PPP_SES): {
108 struct {
109 struct pppoe_hdr hdr;
110 __be16 proto;
111 } *hdr, _hdr;
112 hdr = skb_header_pointer(skb, nhoff, sizeof(_hdr), &_hdr);
113 if (!hdr)
114 return false;
115 proto = hdr->proto;
116 nhoff += PPPOE_SES_HLEN;
117 switch (proto) {
118 case htons(PPP_IP):
119 goto ip;
120 case htons(PPP_IPV6):
121 goto ipv6;
122 default:
123 return false;
124 }
125 }
126 default:
127 return false;
128 }
129
130 switch (ip_proto) {
131 case IPPROTO_GRE: {
132 struct gre_hdr {
133 __be16 flags;
134 __be16 proto;
135 } *hdr, _hdr;
136
137 hdr = skb_header_pointer(skb, nhoff, sizeof(_hdr), &_hdr);
138 if (!hdr)
139 return false;
140 /*
141 * Only look inside GRE if version zero and no
142 * routing
143 */
144 if (!(hdr->flags & (GRE_VERSION|GRE_ROUTING))) {
145 proto = hdr->proto;
146 nhoff += 4;
147 if (hdr->flags & GRE_CSUM)
148 nhoff += 4;
149 if (hdr->flags & GRE_KEY)
150 nhoff += 4;
151 if (hdr->flags & GRE_SEQ)
152 nhoff += 4;
153 if (proto == htons(ETH_P_TEB)) {
154 const struct ethhdr *eth;
155 struct ethhdr _eth;
156
157 eth = skb_header_pointer(skb, nhoff,
158 sizeof(_eth), &_eth);
159 if (!eth)
160 return false;
161 proto = eth->h_proto;
162 nhoff += sizeof(*eth);
163 }
164 goto again;
165 }
166 break;
167 }
168 case IPPROTO_IPIP:
169 proto = htons(ETH_P_IP);
170 goto ip;
171 case IPPROTO_IPV6:
172 proto = htons(ETH_P_IPV6);
173 goto ipv6;
174 default:
175 break;
176 }
177
178 flow->ip_proto = ip_proto;
179 flow->ports = skb_flow_get_ports(skb, nhoff, ip_proto);
180 flow->thoff = (u16) nhoff;
181
182 return true;
183}
184EXPORT_SYMBOL(skb_flow_dissect);
185
186static u32 hashrnd __read_mostly;
187static __always_inline void __flow_hash_secret_init(void)
188{
189 net_get_random_once(&hashrnd, sizeof(hashrnd));
190}
191
192static __always_inline u32 __flow_hash_3words(u32 a, u32 b, u32 c)
193{
194 __flow_hash_secret_init();
195 return jhash_3words(a, b, c, hashrnd);
196}
197
198static __always_inline u32 __flow_hash_1word(u32 a)
199{
200 __flow_hash_secret_init();
201 return jhash_1word(a, hashrnd);
202}
203
204/*
205 * __skb_get_hash: calculate a flow hash based on src/dst addresses
206 * and src/dst port numbers. Sets hash in skb to non-zero hash value
207 * on success, zero indicates no valid hash. Also, sets l4_hash in skb
208 * if hash is a canonical 4-tuple hash over transport ports.
209 */
210void __skb_get_hash(struct sk_buff *skb)
211{
212 struct flow_keys keys;
213 u32 hash;
214
215 if (!skb_flow_dissect(skb, &keys))
216 return;
217
218 if (keys.ports)
219 skb->l4_hash = 1;
220
221 /* get a consistent hash (same value on both flow directions) */
222 if (((__force u32)keys.dst < (__force u32)keys.src) ||
223 (((__force u32)keys.dst == (__force u32)keys.src) &&
224 ((__force u16)keys.port16[1] < (__force u16)keys.port16[0]))) {
225 swap(keys.dst, keys.src);
226 swap(keys.port16[0], keys.port16[1]);
227 }
228
229 hash = __flow_hash_3words((__force u32)keys.dst,
230 (__force u32)keys.src,
231 (__force u32)keys.ports);
232 if (!hash)
233 hash = 1;
234
235 skb->hash = hash;
236}
237EXPORT_SYMBOL(__skb_get_hash);
238
239/*
240 * Returns a Tx hash based on the given packet descriptor a Tx queues' number
241 * to be used as a distribution range.
242 */
243u16 __skb_tx_hash(const struct net_device *dev, const struct sk_buff *skb,
244 unsigned int num_tx_queues)
245{
246 u32 hash;
247 u16 qoffset = 0;
248 u16 qcount = num_tx_queues;
249
250 if (skb_rx_queue_recorded(skb)) {
251 hash = skb_get_rx_queue(skb);
252 while (unlikely(hash >= num_tx_queues))
253 hash -= num_tx_queues;
254 return hash;
255 }
256
257 if (dev->num_tc) {
258 u8 tc = netdev_get_prio_tc_map(dev, skb->priority);
259 qoffset = dev->tc_to_txq[tc].offset;
260 qcount = dev->tc_to_txq[tc].count;
261 }
262
263 if (skb->sk && skb->sk->sk_hash)
264 hash = skb->sk->sk_hash;
265 else
266 hash = (__force u16) skb->protocol;
267 hash = __flow_hash_1word(hash);
268
269 return (u16) (((u64) hash * qcount) >> 32) + qoffset;
270}
271EXPORT_SYMBOL(__skb_tx_hash);
272
273/* __skb_get_poff() returns the offset to the payload as far as it could
274 * be dissected. The main user is currently BPF, so that we can dynamically
275 * truncate packets without needing to push actual payload to the user
276 * space and can analyze headers only, instead.
277 */
278u32 __skb_get_poff(const struct sk_buff *skb)
279{
280 struct flow_keys keys;
281 u32 poff = 0;
282
283 if (!skb_flow_dissect(skb, &keys))
284 return 0;
285
286 poff += keys.thoff;
287 switch (keys.ip_proto) {
288 case IPPROTO_TCP: {
289 const struct tcphdr *tcph;
290 struct tcphdr _tcph;
291
292 tcph = skb_header_pointer(skb, poff, sizeof(_tcph), &_tcph);
293 if (!tcph)
294 return poff;
295
296 poff += max_t(u32, sizeof(struct tcphdr), tcph->doff * 4);
297 break;
298 }
299 case IPPROTO_UDP:
300 case IPPROTO_UDPLITE:
301 poff += sizeof(struct udphdr);
302 break;
303 /* For the rest, we do not really care about header
304 * extensions at this point for now.
305 */
306 case IPPROTO_ICMP:
307 poff += sizeof(struct icmphdr);
308 break;
309 case IPPROTO_ICMPV6:
310 poff += sizeof(struct icmp6hdr);
311 break;
312 case IPPROTO_IGMP:
313 poff += sizeof(struct igmphdr);
314 break;
315 case IPPROTO_DCCP:
316 poff += sizeof(struct dccp_hdr);
317 break;
318 case IPPROTO_SCTP:
319 poff += sizeof(struct sctphdr);
320 break;
321 }
322
323 return poff;
324}
325
326static inline int get_xps_queue(struct net_device *dev, struct sk_buff *skb)
327{
328#ifdef CONFIG_XPS
329 struct xps_dev_maps *dev_maps;
330 struct xps_map *map;
331 int queue_index = -1;
332
333 rcu_read_lock();
334 dev_maps = rcu_dereference(dev->xps_maps);
335 if (dev_maps) {
336 map = rcu_dereference(
337 dev_maps->cpu_map[raw_smp_processor_id()]);
338 if (map) {
339 if (map->len == 1)
340 queue_index = map->queues[0];
341 else {
342 u32 hash;
343 if (skb->sk && skb->sk->sk_hash)
344 hash = skb->sk->sk_hash;
345 else
346 hash = (__force u16) skb->protocol ^
347 skb->hash;
348 hash = __flow_hash_1word(hash);
349 queue_index = map->queues[
350 ((u64)hash * map->len) >> 32];
351 }
352 if (unlikely(queue_index >= dev->real_num_tx_queues))
353 queue_index = -1;
354 }
355 }
356 rcu_read_unlock();
357
358 return queue_index;
359#else
360 return -1;
361#endif
362}
363
364static u16 __netdev_pick_tx(struct net_device *dev, struct sk_buff *skb)
365{
366 struct sock *sk = skb->sk;
367 int queue_index = sk_tx_queue_get(sk);
368
369 if (queue_index < 0 || skb->ooo_okay ||
370 queue_index >= dev->real_num_tx_queues) {
371 int new_index = get_xps_queue(dev, skb);
372 if (new_index < 0)
373 new_index = skb_tx_hash(dev, skb);
374
375 if (queue_index != new_index && sk &&
376 rcu_access_pointer(sk->sk_dst_cache))
377 sk_tx_queue_set(sk, new_index);
378
379 queue_index = new_index;
380 }
381
382 return queue_index;
383}
384
385struct netdev_queue *netdev_pick_tx(struct net_device *dev,
386 struct sk_buff *skb,
387 void *accel_priv)
388{
389 int queue_index = 0;
390
391 if (dev->real_num_tx_queues != 1) {
392 const struct net_device_ops *ops = dev->netdev_ops;
393 if (ops->ndo_select_queue)
394 queue_index = ops->ndo_select_queue(dev, skb, accel_priv,
395 __netdev_pick_tx);
396 else
397 queue_index = __netdev_pick_tx(dev, skb);
398
399 if (!accel_priv)
400 queue_index = netdev_cap_txqueue(dev, queue_index);
401 }
402
403 skb_set_queue_mapping(skb, queue_index);
404 return netdev_get_tx_queue(dev, queue_index);
405}
1#include <linux/kernel.h>
2#include <linux/skbuff.h>
3#include <linux/export.h>
4#include <linux/ip.h>
5#include <linux/ipv6.h>
6#include <linux/if_vlan.h>
7#include <net/dsa.h>
8#include <net/dst_metadata.h>
9#include <net/ip.h>
10#include <net/ipv6.h>
11#include <net/gre.h>
12#include <net/pptp.h>
13#include <net/tipc.h>
14#include <linux/igmp.h>
15#include <linux/icmp.h>
16#include <linux/sctp.h>
17#include <linux/dccp.h>
18#include <linux/if_tunnel.h>
19#include <linux/if_pppox.h>
20#include <linux/ppp_defs.h>
21#include <linux/stddef.h>
22#include <linux/if_ether.h>
23#include <linux/mpls.h>
24#include <linux/tcp.h>
25#include <net/flow_dissector.h>
26#include <scsi/fc/fc_fcoe.h>
27#include <uapi/linux/batadv_packet.h>
28
29static void dissector_set_key(struct flow_dissector *flow_dissector,
30 enum flow_dissector_key_id key_id)
31{
32 flow_dissector->used_keys |= (1 << key_id);
33}
34
35void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
36 const struct flow_dissector_key *key,
37 unsigned int key_count)
38{
39 unsigned int i;
40
41 memset(flow_dissector, 0, sizeof(*flow_dissector));
42
43 for (i = 0; i < key_count; i++, key++) {
44 /* User should make sure that every key target offset is withing
45 * boundaries of unsigned short.
46 */
47 BUG_ON(key->offset > USHRT_MAX);
48 BUG_ON(dissector_uses_key(flow_dissector,
49 key->key_id));
50
51 dissector_set_key(flow_dissector, key->key_id);
52 flow_dissector->offset[key->key_id] = key->offset;
53 }
54
55 /* Ensure that the dissector always includes control and basic key.
56 * That way we are able to avoid handling lack of these in fast path.
57 */
58 BUG_ON(!dissector_uses_key(flow_dissector,
59 FLOW_DISSECTOR_KEY_CONTROL));
60 BUG_ON(!dissector_uses_key(flow_dissector,
61 FLOW_DISSECTOR_KEY_BASIC));
62}
63EXPORT_SYMBOL(skb_flow_dissector_init);
64
65/**
66 * skb_flow_get_be16 - extract be16 entity
67 * @skb: sk_buff to extract from
68 * @poff: offset to extract at
69 * @data: raw buffer pointer to the packet
70 * @hlen: packet header length
71 *
72 * The function will try to retrieve a be32 entity at
73 * offset poff
74 */
75static __be16 skb_flow_get_be16(const struct sk_buff *skb, int poff,
76 void *data, int hlen)
77{
78 __be16 *u, _u;
79
80 u = __skb_header_pointer(skb, poff, sizeof(_u), data, hlen, &_u);
81 if (u)
82 return *u;
83
84 return 0;
85}
86
87/**
88 * __skb_flow_get_ports - extract the upper layer ports and return them
89 * @skb: sk_buff to extract the ports from
90 * @thoff: transport header offset
91 * @ip_proto: protocol for which to get port offset
92 * @data: raw buffer pointer to the packet, if NULL use skb->data
93 * @hlen: packet header length, if @data is NULL use skb_headlen(skb)
94 *
95 * The function will try to retrieve the ports at offset thoff + poff where poff
96 * is the protocol port offset returned from proto_ports_offset
97 */
98__be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
99 void *data, int hlen)
100{
101 int poff = proto_ports_offset(ip_proto);
102
103 if (!data) {
104 data = skb->data;
105 hlen = skb_headlen(skb);
106 }
107
108 if (poff >= 0) {
109 __be32 *ports, _ports;
110
111 ports = __skb_header_pointer(skb, thoff + poff,
112 sizeof(_ports), data, hlen, &_ports);
113 if (ports)
114 return *ports;
115 }
116
117 return 0;
118}
119EXPORT_SYMBOL(__skb_flow_get_ports);
120
121static void
122skb_flow_dissect_set_enc_addr_type(enum flow_dissector_key_id type,
123 struct flow_dissector *flow_dissector,
124 void *target_container)
125{
126 struct flow_dissector_key_control *ctrl;
127
128 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_CONTROL))
129 return;
130
131 ctrl = skb_flow_dissector_target(flow_dissector,
132 FLOW_DISSECTOR_KEY_ENC_CONTROL,
133 target_container);
134 ctrl->addr_type = type;
135}
136
137void
138skb_flow_dissect_tunnel_info(const struct sk_buff *skb,
139 struct flow_dissector *flow_dissector,
140 void *target_container)
141{
142 struct ip_tunnel_info *info;
143 struct ip_tunnel_key *key;
144
145 /* A quick check to see if there might be something to do. */
146 if (!dissector_uses_key(flow_dissector,
147 FLOW_DISSECTOR_KEY_ENC_KEYID) &&
148 !dissector_uses_key(flow_dissector,
149 FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS) &&
150 !dissector_uses_key(flow_dissector,
151 FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS) &&
152 !dissector_uses_key(flow_dissector,
153 FLOW_DISSECTOR_KEY_ENC_CONTROL) &&
154 !dissector_uses_key(flow_dissector,
155 FLOW_DISSECTOR_KEY_ENC_PORTS))
156 return;
157
158 info = skb_tunnel_info(skb);
159 if (!info)
160 return;
161
162 key = &info->key;
163
164 switch (ip_tunnel_info_af(info)) {
165 case AF_INET:
166 skb_flow_dissect_set_enc_addr_type(FLOW_DISSECTOR_KEY_IPV4_ADDRS,
167 flow_dissector,
168 target_container);
169 if (dissector_uses_key(flow_dissector,
170 FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS)) {
171 struct flow_dissector_key_ipv4_addrs *ipv4;
172
173 ipv4 = skb_flow_dissector_target(flow_dissector,
174 FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS,
175 target_container);
176 ipv4->src = key->u.ipv4.src;
177 ipv4->dst = key->u.ipv4.dst;
178 }
179 break;
180 case AF_INET6:
181 skb_flow_dissect_set_enc_addr_type(FLOW_DISSECTOR_KEY_IPV6_ADDRS,
182 flow_dissector,
183 target_container);
184 if (dissector_uses_key(flow_dissector,
185 FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS)) {
186 struct flow_dissector_key_ipv6_addrs *ipv6;
187
188 ipv6 = skb_flow_dissector_target(flow_dissector,
189 FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS,
190 target_container);
191 ipv6->src = key->u.ipv6.src;
192 ipv6->dst = key->u.ipv6.dst;
193 }
194 break;
195 }
196
197 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_KEYID)) {
198 struct flow_dissector_key_keyid *keyid;
199
200 keyid = skb_flow_dissector_target(flow_dissector,
201 FLOW_DISSECTOR_KEY_ENC_KEYID,
202 target_container);
203 keyid->keyid = tunnel_id_to_key32(key->tun_id);
204 }
205
206 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_PORTS)) {
207 struct flow_dissector_key_ports *tp;
208
209 tp = skb_flow_dissector_target(flow_dissector,
210 FLOW_DISSECTOR_KEY_ENC_PORTS,
211 target_container);
212 tp->src = key->tp_src;
213 tp->dst = key->tp_dst;
214 }
215}
216EXPORT_SYMBOL(skb_flow_dissect_tunnel_info);
217
218static enum flow_dissect_ret
219__skb_flow_dissect_mpls(const struct sk_buff *skb,
220 struct flow_dissector *flow_dissector,
221 void *target_container, void *data, int nhoff, int hlen)
222{
223 struct flow_dissector_key_keyid *key_keyid;
224 struct mpls_label *hdr, _hdr[2];
225 u32 entry, label;
226
227 if (!dissector_uses_key(flow_dissector,
228 FLOW_DISSECTOR_KEY_MPLS_ENTROPY) &&
229 !dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_MPLS))
230 return FLOW_DISSECT_RET_OUT_GOOD;
231
232 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data,
233 hlen, &_hdr);
234 if (!hdr)
235 return FLOW_DISSECT_RET_OUT_BAD;
236
237 entry = ntohl(hdr[0].entry);
238 label = (entry & MPLS_LS_LABEL_MASK) >> MPLS_LS_LABEL_SHIFT;
239
240 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_MPLS)) {
241 struct flow_dissector_key_mpls *key_mpls;
242
243 key_mpls = skb_flow_dissector_target(flow_dissector,
244 FLOW_DISSECTOR_KEY_MPLS,
245 target_container);
246 key_mpls->mpls_label = label;
247 key_mpls->mpls_ttl = (entry & MPLS_LS_TTL_MASK)
248 >> MPLS_LS_TTL_SHIFT;
249 key_mpls->mpls_tc = (entry & MPLS_LS_TC_MASK)
250 >> MPLS_LS_TC_SHIFT;
251 key_mpls->mpls_bos = (entry & MPLS_LS_S_MASK)
252 >> MPLS_LS_S_SHIFT;
253 }
254
255 if (label == MPLS_LABEL_ENTROPY) {
256 key_keyid = skb_flow_dissector_target(flow_dissector,
257 FLOW_DISSECTOR_KEY_MPLS_ENTROPY,
258 target_container);
259 key_keyid->keyid = hdr[1].entry & htonl(MPLS_LS_LABEL_MASK);
260 }
261 return FLOW_DISSECT_RET_OUT_GOOD;
262}
263
264static enum flow_dissect_ret
265__skb_flow_dissect_arp(const struct sk_buff *skb,
266 struct flow_dissector *flow_dissector,
267 void *target_container, void *data, int nhoff, int hlen)
268{
269 struct flow_dissector_key_arp *key_arp;
270 struct {
271 unsigned char ar_sha[ETH_ALEN];
272 unsigned char ar_sip[4];
273 unsigned char ar_tha[ETH_ALEN];
274 unsigned char ar_tip[4];
275 } *arp_eth, _arp_eth;
276 const struct arphdr *arp;
277 struct arphdr _arp;
278
279 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ARP))
280 return FLOW_DISSECT_RET_OUT_GOOD;
281
282 arp = __skb_header_pointer(skb, nhoff, sizeof(_arp), data,
283 hlen, &_arp);
284 if (!arp)
285 return FLOW_DISSECT_RET_OUT_BAD;
286
287 if (arp->ar_hrd != htons(ARPHRD_ETHER) ||
288 arp->ar_pro != htons(ETH_P_IP) ||
289 arp->ar_hln != ETH_ALEN ||
290 arp->ar_pln != 4 ||
291 (arp->ar_op != htons(ARPOP_REPLY) &&
292 arp->ar_op != htons(ARPOP_REQUEST)))
293 return FLOW_DISSECT_RET_OUT_BAD;
294
295 arp_eth = __skb_header_pointer(skb, nhoff + sizeof(_arp),
296 sizeof(_arp_eth), data,
297 hlen, &_arp_eth);
298 if (!arp_eth)
299 return FLOW_DISSECT_RET_OUT_BAD;
300
301 key_arp = skb_flow_dissector_target(flow_dissector,
302 FLOW_DISSECTOR_KEY_ARP,
303 target_container);
304
305 memcpy(&key_arp->sip, arp_eth->ar_sip, sizeof(key_arp->sip));
306 memcpy(&key_arp->tip, arp_eth->ar_tip, sizeof(key_arp->tip));
307
308 /* Only store the lower byte of the opcode;
309 * this covers ARPOP_REPLY and ARPOP_REQUEST.
310 */
311 key_arp->op = ntohs(arp->ar_op) & 0xff;
312
313 ether_addr_copy(key_arp->sha, arp_eth->ar_sha);
314 ether_addr_copy(key_arp->tha, arp_eth->ar_tha);
315
316 return FLOW_DISSECT_RET_OUT_GOOD;
317}
318
319static enum flow_dissect_ret
320__skb_flow_dissect_gre(const struct sk_buff *skb,
321 struct flow_dissector_key_control *key_control,
322 struct flow_dissector *flow_dissector,
323 void *target_container, void *data,
324 __be16 *p_proto, int *p_nhoff, int *p_hlen,
325 unsigned int flags)
326{
327 struct flow_dissector_key_keyid *key_keyid;
328 struct gre_base_hdr *hdr, _hdr;
329 int offset = 0;
330 u16 gre_ver;
331
332 hdr = __skb_header_pointer(skb, *p_nhoff, sizeof(_hdr),
333 data, *p_hlen, &_hdr);
334 if (!hdr)
335 return FLOW_DISSECT_RET_OUT_BAD;
336
337 /* Only look inside GRE without routing */
338 if (hdr->flags & GRE_ROUTING)
339 return FLOW_DISSECT_RET_OUT_GOOD;
340
341 /* Only look inside GRE for version 0 and 1 */
342 gre_ver = ntohs(hdr->flags & GRE_VERSION);
343 if (gre_ver > 1)
344 return FLOW_DISSECT_RET_OUT_GOOD;
345
346 *p_proto = hdr->protocol;
347 if (gre_ver) {
348 /* Version1 must be PPTP, and check the flags */
349 if (!(*p_proto == GRE_PROTO_PPP && (hdr->flags & GRE_KEY)))
350 return FLOW_DISSECT_RET_OUT_GOOD;
351 }
352
353 offset += sizeof(struct gre_base_hdr);
354
355 if (hdr->flags & GRE_CSUM)
356 offset += sizeof(((struct gre_full_hdr *) 0)->csum) +
357 sizeof(((struct gre_full_hdr *) 0)->reserved1);
358
359 if (hdr->flags & GRE_KEY) {
360 const __be32 *keyid;
361 __be32 _keyid;
362
363 keyid = __skb_header_pointer(skb, *p_nhoff + offset,
364 sizeof(_keyid),
365 data, *p_hlen, &_keyid);
366 if (!keyid)
367 return FLOW_DISSECT_RET_OUT_BAD;
368
369 if (dissector_uses_key(flow_dissector,
370 FLOW_DISSECTOR_KEY_GRE_KEYID)) {
371 key_keyid = skb_flow_dissector_target(flow_dissector,
372 FLOW_DISSECTOR_KEY_GRE_KEYID,
373 target_container);
374 if (gre_ver == 0)
375 key_keyid->keyid = *keyid;
376 else
377 key_keyid->keyid = *keyid & GRE_PPTP_KEY_MASK;
378 }
379 offset += sizeof(((struct gre_full_hdr *) 0)->key);
380 }
381
382 if (hdr->flags & GRE_SEQ)
383 offset += sizeof(((struct pptp_gre_header *) 0)->seq);
384
385 if (gre_ver == 0) {
386 if (*p_proto == htons(ETH_P_TEB)) {
387 const struct ethhdr *eth;
388 struct ethhdr _eth;
389
390 eth = __skb_header_pointer(skb, *p_nhoff + offset,
391 sizeof(_eth),
392 data, *p_hlen, &_eth);
393 if (!eth)
394 return FLOW_DISSECT_RET_OUT_BAD;
395 *p_proto = eth->h_proto;
396 offset += sizeof(*eth);
397
398 /* Cap headers that we access via pointers at the
399 * end of the Ethernet header as our maximum alignment
400 * at that point is only 2 bytes.
401 */
402 if (NET_IP_ALIGN)
403 *p_hlen = *p_nhoff + offset;
404 }
405 } else { /* version 1, must be PPTP */
406 u8 _ppp_hdr[PPP_HDRLEN];
407 u8 *ppp_hdr;
408
409 if (hdr->flags & GRE_ACK)
410 offset += sizeof(((struct pptp_gre_header *) 0)->ack);
411
412 ppp_hdr = __skb_header_pointer(skb, *p_nhoff + offset,
413 sizeof(_ppp_hdr),
414 data, *p_hlen, _ppp_hdr);
415 if (!ppp_hdr)
416 return FLOW_DISSECT_RET_OUT_BAD;
417
418 switch (PPP_PROTOCOL(ppp_hdr)) {
419 case PPP_IP:
420 *p_proto = htons(ETH_P_IP);
421 break;
422 case PPP_IPV6:
423 *p_proto = htons(ETH_P_IPV6);
424 break;
425 default:
426 /* Could probably catch some more like MPLS */
427 break;
428 }
429
430 offset += PPP_HDRLEN;
431 }
432
433 *p_nhoff += offset;
434 key_control->flags |= FLOW_DIS_ENCAPSULATION;
435 if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP)
436 return FLOW_DISSECT_RET_OUT_GOOD;
437
438 return FLOW_DISSECT_RET_PROTO_AGAIN;
439}
440
441/**
442 * __skb_flow_dissect_batadv() - dissect batman-adv header
443 * @skb: sk_buff to with the batman-adv header
444 * @key_control: flow dissectors control key
445 * @data: raw buffer pointer to the packet, if NULL use skb->data
446 * @p_proto: pointer used to update the protocol to process next
447 * @p_nhoff: pointer used to update inner network header offset
448 * @hlen: packet header length
449 * @flags: any combination of FLOW_DISSECTOR_F_*
450 *
451 * ETH_P_BATMAN packets are tried to be dissected. Only
452 * &struct batadv_unicast packets are actually processed because they contain an
453 * inner ethernet header and are usually followed by actual network header. This
454 * allows the flow dissector to continue processing the packet.
455 *
456 * Return: FLOW_DISSECT_RET_PROTO_AGAIN when &struct batadv_unicast was found,
457 * FLOW_DISSECT_RET_OUT_GOOD when dissector should stop after encapsulation,
458 * otherwise FLOW_DISSECT_RET_OUT_BAD
459 */
460static enum flow_dissect_ret
461__skb_flow_dissect_batadv(const struct sk_buff *skb,
462 struct flow_dissector_key_control *key_control,
463 void *data, __be16 *p_proto, int *p_nhoff, int hlen,
464 unsigned int flags)
465{
466 struct {
467 struct batadv_unicast_packet batadv_unicast;
468 struct ethhdr eth;
469 } *hdr, _hdr;
470
471 hdr = __skb_header_pointer(skb, *p_nhoff, sizeof(_hdr), data, hlen,
472 &_hdr);
473 if (!hdr)
474 return FLOW_DISSECT_RET_OUT_BAD;
475
476 if (hdr->batadv_unicast.version != BATADV_COMPAT_VERSION)
477 return FLOW_DISSECT_RET_OUT_BAD;
478
479 if (hdr->batadv_unicast.packet_type != BATADV_UNICAST)
480 return FLOW_DISSECT_RET_OUT_BAD;
481
482 *p_proto = hdr->eth.h_proto;
483 *p_nhoff += sizeof(*hdr);
484
485 key_control->flags |= FLOW_DIS_ENCAPSULATION;
486 if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP)
487 return FLOW_DISSECT_RET_OUT_GOOD;
488
489 return FLOW_DISSECT_RET_PROTO_AGAIN;
490}
491
492static void
493__skb_flow_dissect_tcp(const struct sk_buff *skb,
494 struct flow_dissector *flow_dissector,
495 void *target_container, void *data, int thoff, int hlen)
496{
497 struct flow_dissector_key_tcp *key_tcp;
498 struct tcphdr *th, _th;
499
500 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_TCP))
501 return;
502
503 th = __skb_header_pointer(skb, thoff, sizeof(_th), data, hlen, &_th);
504 if (!th)
505 return;
506
507 if (unlikely(__tcp_hdrlen(th) < sizeof(_th)))
508 return;
509
510 key_tcp = skb_flow_dissector_target(flow_dissector,
511 FLOW_DISSECTOR_KEY_TCP,
512 target_container);
513 key_tcp->flags = (*(__be16 *) &tcp_flag_word(th) & htons(0x0FFF));
514}
515
516static void
517__skb_flow_dissect_ipv4(const struct sk_buff *skb,
518 struct flow_dissector *flow_dissector,
519 void *target_container, void *data, const struct iphdr *iph)
520{
521 struct flow_dissector_key_ip *key_ip;
522
523 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IP))
524 return;
525
526 key_ip = skb_flow_dissector_target(flow_dissector,
527 FLOW_DISSECTOR_KEY_IP,
528 target_container);
529 key_ip->tos = iph->tos;
530 key_ip->ttl = iph->ttl;
531}
532
533static void
534__skb_flow_dissect_ipv6(const struct sk_buff *skb,
535 struct flow_dissector *flow_dissector,
536 void *target_container, void *data, const struct ipv6hdr *iph)
537{
538 struct flow_dissector_key_ip *key_ip;
539
540 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IP))
541 return;
542
543 key_ip = skb_flow_dissector_target(flow_dissector,
544 FLOW_DISSECTOR_KEY_IP,
545 target_container);
546 key_ip->tos = ipv6_get_dsfield(iph);
547 key_ip->ttl = iph->hop_limit;
548}
549
550/* Maximum number of protocol headers that can be parsed in
551 * __skb_flow_dissect
552 */
553#define MAX_FLOW_DISSECT_HDRS 15
554
555static bool skb_flow_dissect_allowed(int *num_hdrs)
556{
557 ++*num_hdrs;
558
559 return (*num_hdrs <= MAX_FLOW_DISSECT_HDRS);
560}
561
562/**
563 * __skb_flow_dissect - extract the flow_keys struct and return it
564 * @skb: sk_buff to extract the flow from, can be NULL if the rest are specified
565 * @flow_dissector: list of keys to dissect
566 * @target_container: target structure to put dissected values into
567 * @data: raw buffer pointer to the packet, if NULL use skb->data
568 * @proto: protocol for which to get the flow, if @data is NULL use skb->protocol
569 * @nhoff: network header offset, if @data is NULL use skb_network_offset(skb)
570 * @hlen: packet header length, if @data is NULL use skb_headlen(skb)
571 *
572 * The function will try to retrieve individual keys into target specified
573 * by flow_dissector from either the skbuff or a raw buffer specified by the
574 * rest parameters.
575 *
576 * Caller must take care of zeroing target container memory.
577 */
578bool __skb_flow_dissect(const struct sk_buff *skb,
579 struct flow_dissector *flow_dissector,
580 void *target_container,
581 void *data, __be16 proto, int nhoff, int hlen,
582 unsigned int flags)
583{
584 struct flow_dissector_key_control *key_control;
585 struct flow_dissector_key_basic *key_basic;
586 struct flow_dissector_key_addrs *key_addrs;
587 struct flow_dissector_key_ports *key_ports;
588 struct flow_dissector_key_icmp *key_icmp;
589 struct flow_dissector_key_tags *key_tags;
590 struct flow_dissector_key_vlan *key_vlan;
591 enum flow_dissect_ret fdret;
592 bool skip_vlan = false;
593 int num_hdrs = 0;
594 u8 ip_proto = 0;
595 bool ret;
596
597 if (!data) {
598 data = skb->data;
599 proto = skb_vlan_tag_present(skb) ?
600 skb->vlan_proto : skb->protocol;
601 nhoff = skb_network_offset(skb);
602 hlen = skb_headlen(skb);
603#if IS_ENABLED(CONFIG_NET_DSA)
604 if (unlikely(skb->dev && netdev_uses_dsa(skb->dev))) {
605 const struct dsa_device_ops *ops;
606 int offset;
607
608 ops = skb->dev->dsa_ptr->tag_ops;
609 if (ops->flow_dissect &&
610 !ops->flow_dissect(skb, &proto, &offset)) {
611 hlen -= offset;
612 nhoff += offset;
613 }
614 }
615#endif
616 }
617
618 /* It is ensured by skb_flow_dissector_init() that control key will
619 * be always present.
620 */
621 key_control = skb_flow_dissector_target(flow_dissector,
622 FLOW_DISSECTOR_KEY_CONTROL,
623 target_container);
624
625 /* It is ensured by skb_flow_dissector_init() that basic key will
626 * be always present.
627 */
628 key_basic = skb_flow_dissector_target(flow_dissector,
629 FLOW_DISSECTOR_KEY_BASIC,
630 target_container);
631
632 if (dissector_uses_key(flow_dissector,
633 FLOW_DISSECTOR_KEY_ETH_ADDRS)) {
634 struct ethhdr *eth = eth_hdr(skb);
635 struct flow_dissector_key_eth_addrs *key_eth_addrs;
636
637 key_eth_addrs = skb_flow_dissector_target(flow_dissector,
638 FLOW_DISSECTOR_KEY_ETH_ADDRS,
639 target_container);
640 memcpy(key_eth_addrs, ð->h_dest, sizeof(*key_eth_addrs));
641 }
642
643proto_again:
644 fdret = FLOW_DISSECT_RET_CONTINUE;
645
646 switch (proto) {
647 case htons(ETH_P_IP): {
648 const struct iphdr *iph;
649 struct iphdr _iph;
650
651 iph = __skb_header_pointer(skb, nhoff, sizeof(_iph), data, hlen, &_iph);
652 if (!iph || iph->ihl < 5) {
653 fdret = FLOW_DISSECT_RET_OUT_BAD;
654 break;
655 }
656
657 nhoff += iph->ihl * 4;
658
659 ip_proto = iph->protocol;
660
661 if (dissector_uses_key(flow_dissector,
662 FLOW_DISSECTOR_KEY_IPV4_ADDRS)) {
663 key_addrs = skb_flow_dissector_target(flow_dissector,
664 FLOW_DISSECTOR_KEY_IPV4_ADDRS,
665 target_container);
666
667 memcpy(&key_addrs->v4addrs, &iph->saddr,
668 sizeof(key_addrs->v4addrs));
669 key_control->addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
670 }
671
672 if (ip_is_fragment(iph)) {
673 key_control->flags |= FLOW_DIS_IS_FRAGMENT;
674
675 if (iph->frag_off & htons(IP_OFFSET)) {
676 fdret = FLOW_DISSECT_RET_OUT_GOOD;
677 break;
678 } else {
679 key_control->flags |= FLOW_DIS_FIRST_FRAG;
680 if (!(flags &
681 FLOW_DISSECTOR_F_PARSE_1ST_FRAG)) {
682 fdret = FLOW_DISSECT_RET_OUT_GOOD;
683 break;
684 }
685 }
686 }
687
688 __skb_flow_dissect_ipv4(skb, flow_dissector,
689 target_container, data, iph);
690
691 if (flags & FLOW_DISSECTOR_F_STOP_AT_L3) {
692 fdret = FLOW_DISSECT_RET_OUT_GOOD;
693 break;
694 }
695
696 break;
697 }
698 case htons(ETH_P_IPV6): {
699 const struct ipv6hdr *iph;
700 struct ipv6hdr _iph;
701
702 iph = __skb_header_pointer(skb, nhoff, sizeof(_iph), data, hlen, &_iph);
703 if (!iph) {
704 fdret = FLOW_DISSECT_RET_OUT_BAD;
705 break;
706 }
707
708 ip_proto = iph->nexthdr;
709 nhoff += sizeof(struct ipv6hdr);
710
711 if (dissector_uses_key(flow_dissector,
712 FLOW_DISSECTOR_KEY_IPV6_ADDRS)) {
713 key_addrs = skb_flow_dissector_target(flow_dissector,
714 FLOW_DISSECTOR_KEY_IPV6_ADDRS,
715 target_container);
716
717 memcpy(&key_addrs->v6addrs, &iph->saddr,
718 sizeof(key_addrs->v6addrs));
719 key_control->addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
720 }
721
722 if ((dissector_uses_key(flow_dissector,
723 FLOW_DISSECTOR_KEY_FLOW_LABEL) ||
724 (flags & FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL)) &&
725 ip6_flowlabel(iph)) {
726 __be32 flow_label = ip6_flowlabel(iph);
727
728 if (dissector_uses_key(flow_dissector,
729 FLOW_DISSECTOR_KEY_FLOW_LABEL)) {
730 key_tags = skb_flow_dissector_target(flow_dissector,
731 FLOW_DISSECTOR_KEY_FLOW_LABEL,
732 target_container);
733 key_tags->flow_label = ntohl(flow_label);
734 }
735 if (flags & FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL) {
736 fdret = FLOW_DISSECT_RET_OUT_GOOD;
737 break;
738 }
739 }
740
741 __skb_flow_dissect_ipv6(skb, flow_dissector,
742 target_container, data, iph);
743
744 if (flags & FLOW_DISSECTOR_F_STOP_AT_L3)
745 fdret = FLOW_DISSECT_RET_OUT_GOOD;
746
747 break;
748 }
749 case htons(ETH_P_8021AD):
750 case htons(ETH_P_8021Q): {
751 const struct vlan_hdr *vlan;
752 struct vlan_hdr _vlan;
753 bool vlan_tag_present = skb && skb_vlan_tag_present(skb);
754
755 if (vlan_tag_present)
756 proto = skb->protocol;
757
758 if (!vlan_tag_present || eth_type_vlan(skb->protocol)) {
759 vlan = __skb_header_pointer(skb, nhoff, sizeof(_vlan),
760 data, hlen, &_vlan);
761 if (!vlan) {
762 fdret = FLOW_DISSECT_RET_OUT_BAD;
763 break;
764 }
765
766 proto = vlan->h_vlan_encapsulated_proto;
767 nhoff += sizeof(*vlan);
768 if (skip_vlan) {
769 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
770 break;
771 }
772 }
773
774 skip_vlan = true;
775 if (dissector_uses_key(flow_dissector,
776 FLOW_DISSECTOR_KEY_VLAN)) {
777 key_vlan = skb_flow_dissector_target(flow_dissector,
778 FLOW_DISSECTOR_KEY_VLAN,
779 target_container);
780
781 if (vlan_tag_present) {
782 key_vlan->vlan_id = skb_vlan_tag_get_id(skb);
783 key_vlan->vlan_priority =
784 (skb_vlan_tag_get_prio(skb) >> VLAN_PRIO_SHIFT);
785 } else {
786 key_vlan->vlan_id = ntohs(vlan->h_vlan_TCI) &
787 VLAN_VID_MASK;
788 key_vlan->vlan_priority =
789 (ntohs(vlan->h_vlan_TCI) &
790 VLAN_PRIO_MASK) >> VLAN_PRIO_SHIFT;
791 }
792 }
793
794 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
795 break;
796 }
797 case htons(ETH_P_PPP_SES): {
798 struct {
799 struct pppoe_hdr hdr;
800 __be16 proto;
801 } *hdr, _hdr;
802 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, hlen, &_hdr);
803 if (!hdr) {
804 fdret = FLOW_DISSECT_RET_OUT_BAD;
805 break;
806 }
807
808 proto = hdr->proto;
809 nhoff += PPPOE_SES_HLEN;
810 switch (proto) {
811 case htons(PPP_IP):
812 proto = htons(ETH_P_IP);
813 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
814 break;
815 case htons(PPP_IPV6):
816 proto = htons(ETH_P_IPV6);
817 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
818 break;
819 default:
820 fdret = FLOW_DISSECT_RET_OUT_BAD;
821 break;
822 }
823 break;
824 }
825 case htons(ETH_P_TIPC): {
826 struct tipc_basic_hdr *hdr, _hdr;
827
828 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr),
829 data, hlen, &_hdr);
830 if (!hdr) {
831 fdret = FLOW_DISSECT_RET_OUT_BAD;
832 break;
833 }
834
835 if (dissector_uses_key(flow_dissector,
836 FLOW_DISSECTOR_KEY_TIPC)) {
837 key_addrs = skb_flow_dissector_target(flow_dissector,
838 FLOW_DISSECTOR_KEY_TIPC,
839 target_container);
840 key_addrs->tipckey.key = tipc_hdr_rps_key(hdr);
841 key_control->addr_type = FLOW_DISSECTOR_KEY_TIPC;
842 }
843 fdret = FLOW_DISSECT_RET_OUT_GOOD;
844 break;
845 }
846
847 case htons(ETH_P_MPLS_UC):
848 case htons(ETH_P_MPLS_MC):
849 fdret = __skb_flow_dissect_mpls(skb, flow_dissector,
850 target_container, data,
851 nhoff, hlen);
852 break;
853 case htons(ETH_P_FCOE):
854 if ((hlen - nhoff) < FCOE_HEADER_LEN) {
855 fdret = FLOW_DISSECT_RET_OUT_BAD;
856 break;
857 }
858
859 nhoff += FCOE_HEADER_LEN;
860 fdret = FLOW_DISSECT_RET_OUT_GOOD;
861 break;
862
863 case htons(ETH_P_ARP):
864 case htons(ETH_P_RARP):
865 fdret = __skb_flow_dissect_arp(skb, flow_dissector,
866 target_container, data,
867 nhoff, hlen);
868 break;
869
870 case htons(ETH_P_BATMAN):
871 fdret = __skb_flow_dissect_batadv(skb, key_control, data,
872 &proto, &nhoff, hlen, flags);
873 break;
874
875 default:
876 fdret = FLOW_DISSECT_RET_OUT_BAD;
877 break;
878 }
879
880 /* Process result of proto processing */
881 switch (fdret) {
882 case FLOW_DISSECT_RET_OUT_GOOD:
883 goto out_good;
884 case FLOW_DISSECT_RET_PROTO_AGAIN:
885 if (skb_flow_dissect_allowed(&num_hdrs))
886 goto proto_again;
887 goto out_good;
888 case FLOW_DISSECT_RET_CONTINUE:
889 case FLOW_DISSECT_RET_IPPROTO_AGAIN:
890 break;
891 case FLOW_DISSECT_RET_OUT_BAD:
892 default:
893 goto out_bad;
894 }
895
896ip_proto_again:
897 fdret = FLOW_DISSECT_RET_CONTINUE;
898
899 switch (ip_proto) {
900 case IPPROTO_GRE:
901 fdret = __skb_flow_dissect_gre(skb, key_control, flow_dissector,
902 target_container, data,
903 &proto, &nhoff, &hlen, flags);
904 break;
905
906 case NEXTHDR_HOP:
907 case NEXTHDR_ROUTING:
908 case NEXTHDR_DEST: {
909 u8 _opthdr[2], *opthdr;
910
911 if (proto != htons(ETH_P_IPV6))
912 break;
913
914 opthdr = __skb_header_pointer(skb, nhoff, sizeof(_opthdr),
915 data, hlen, &_opthdr);
916 if (!opthdr) {
917 fdret = FLOW_DISSECT_RET_OUT_BAD;
918 break;
919 }
920
921 ip_proto = opthdr[0];
922 nhoff += (opthdr[1] + 1) << 3;
923
924 fdret = FLOW_DISSECT_RET_IPPROTO_AGAIN;
925 break;
926 }
927 case NEXTHDR_FRAGMENT: {
928 struct frag_hdr _fh, *fh;
929
930 if (proto != htons(ETH_P_IPV6))
931 break;
932
933 fh = __skb_header_pointer(skb, nhoff, sizeof(_fh),
934 data, hlen, &_fh);
935
936 if (!fh) {
937 fdret = FLOW_DISSECT_RET_OUT_BAD;
938 break;
939 }
940
941 key_control->flags |= FLOW_DIS_IS_FRAGMENT;
942
943 nhoff += sizeof(_fh);
944 ip_proto = fh->nexthdr;
945
946 if (!(fh->frag_off & htons(IP6_OFFSET))) {
947 key_control->flags |= FLOW_DIS_FIRST_FRAG;
948 if (flags & FLOW_DISSECTOR_F_PARSE_1ST_FRAG) {
949 fdret = FLOW_DISSECT_RET_IPPROTO_AGAIN;
950 break;
951 }
952 }
953
954 fdret = FLOW_DISSECT_RET_OUT_GOOD;
955 break;
956 }
957 case IPPROTO_IPIP:
958 proto = htons(ETH_P_IP);
959
960 key_control->flags |= FLOW_DIS_ENCAPSULATION;
961 if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP) {
962 fdret = FLOW_DISSECT_RET_OUT_GOOD;
963 break;
964 }
965
966 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
967 break;
968
969 case IPPROTO_IPV6:
970 proto = htons(ETH_P_IPV6);
971
972 key_control->flags |= FLOW_DIS_ENCAPSULATION;
973 if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP) {
974 fdret = FLOW_DISSECT_RET_OUT_GOOD;
975 break;
976 }
977
978 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
979 break;
980
981
982 case IPPROTO_MPLS:
983 proto = htons(ETH_P_MPLS_UC);
984 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
985 break;
986
987 case IPPROTO_TCP:
988 __skb_flow_dissect_tcp(skb, flow_dissector, target_container,
989 data, nhoff, hlen);
990 break;
991
992 default:
993 break;
994 }
995
996 if (dissector_uses_key(flow_dissector,
997 FLOW_DISSECTOR_KEY_PORTS)) {
998 key_ports = skb_flow_dissector_target(flow_dissector,
999 FLOW_DISSECTOR_KEY_PORTS,
1000 target_container);
1001 key_ports->ports = __skb_flow_get_ports(skb, nhoff, ip_proto,
1002 data, hlen);
1003 }
1004
1005 if (dissector_uses_key(flow_dissector,
1006 FLOW_DISSECTOR_KEY_ICMP)) {
1007 key_icmp = skb_flow_dissector_target(flow_dissector,
1008 FLOW_DISSECTOR_KEY_ICMP,
1009 target_container);
1010 key_icmp->icmp = skb_flow_get_be16(skb, nhoff, data, hlen);
1011 }
1012
1013 /* Process result of IP proto processing */
1014 switch (fdret) {
1015 case FLOW_DISSECT_RET_PROTO_AGAIN:
1016 if (skb_flow_dissect_allowed(&num_hdrs))
1017 goto proto_again;
1018 break;
1019 case FLOW_DISSECT_RET_IPPROTO_AGAIN:
1020 if (skb_flow_dissect_allowed(&num_hdrs))
1021 goto ip_proto_again;
1022 break;
1023 case FLOW_DISSECT_RET_OUT_GOOD:
1024 case FLOW_DISSECT_RET_CONTINUE:
1025 break;
1026 case FLOW_DISSECT_RET_OUT_BAD:
1027 default:
1028 goto out_bad;
1029 }
1030
1031out_good:
1032 ret = true;
1033
1034out:
1035 key_control->thoff = min_t(u16, nhoff, skb ? skb->len : hlen);
1036 key_basic->n_proto = proto;
1037 key_basic->ip_proto = ip_proto;
1038
1039 return ret;
1040
1041out_bad:
1042 ret = false;
1043 goto out;
1044}
1045EXPORT_SYMBOL(__skb_flow_dissect);
1046
1047static u32 hashrnd __read_mostly;
1048static __always_inline void __flow_hash_secret_init(void)
1049{
1050 net_get_random_once(&hashrnd, sizeof(hashrnd));
1051}
1052
1053static __always_inline u32 __flow_hash_words(const u32 *words, u32 length,
1054 u32 keyval)
1055{
1056 return jhash2(words, length, keyval);
1057}
1058
1059static inline const u32 *flow_keys_hash_start(const struct flow_keys *flow)
1060{
1061 const void *p = flow;
1062
1063 BUILD_BUG_ON(FLOW_KEYS_HASH_OFFSET % sizeof(u32));
1064 return (const u32 *)(p + FLOW_KEYS_HASH_OFFSET);
1065}
1066
1067static inline size_t flow_keys_hash_length(const struct flow_keys *flow)
1068{
1069 size_t diff = FLOW_KEYS_HASH_OFFSET + sizeof(flow->addrs);
1070 BUILD_BUG_ON((sizeof(*flow) - FLOW_KEYS_HASH_OFFSET) % sizeof(u32));
1071 BUILD_BUG_ON(offsetof(typeof(*flow), addrs) !=
1072 sizeof(*flow) - sizeof(flow->addrs));
1073
1074 switch (flow->control.addr_type) {
1075 case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1076 diff -= sizeof(flow->addrs.v4addrs);
1077 break;
1078 case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1079 diff -= sizeof(flow->addrs.v6addrs);
1080 break;
1081 case FLOW_DISSECTOR_KEY_TIPC:
1082 diff -= sizeof(flow->addrs.tipckey);
1083 break;
1084 }
1085 return (sizeof(*flow) - diff) / sizeof(u32);
1086}
1087
1088__be32 flow_get_u32_src(const struct flow_keys *flow)
1089{
1090 switch (flow->control.addr_type) {
1091 case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1092 return flow->addrs.v4addrs.src;
1093 case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1094 return (__force __be32)ipv6_addr_hash(
1095 &flow->addrs.v6addrs.src);
1096 case FLOW_DISSECTOR_KEY_TIPC:
1097 return flow->addrs.tipckey.key;
1098 default:
1099 return 0;
1100 }
1101}
1102EXPORT_SYMBOL(flow_get_u32_src);
1103
1104__be32 flow_get_u32_dst(const struct flow_keys *flow)
1105{
1106 switch (flow->control.addr_type) {
1107 case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1108 return flow->addrs.v4addrs.dst;
1109 case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1110 return (__force __be32)ipv6_addr_hash(
1111 &flow->addrs.v6addrs.dst);
1112 default:
1113 return 0;
1114 }
1115}
1116EXPORT_SYMBOL(flow_get_u32_dst);
1117
1118static inline void __flow_hash_consistentify(struct flow_keys *keys)
1119{
1120 int addr_diff, i;
1121
1122 switch (keys->control.addr_type) {
1123 case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1124 addr_diff = (__force u32)keys->addrs.v4addrs.dst -
1125 (__force u32)keys->addrs.v4addrs.src;
1126 if ((addr_diff < 0) ||
1127 (addr_diff == 0 &&
1128 ((__force u16)keys->ports.dst <
1129 (__force u16)keys->ports.src))) {
1130 swap(keys->addrs.v4addrs.src, keys->addrs.v4addrs.dst);
1131 swap(keys->ports.src, keys->ports.dst);
1132 }
1133 break;
1134 case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1135 addr_diff = memcmp(&keys->addrs.v6addrs.dst,
1136 &keys->addrs.v6addrs.src,
1137 sizeof(keys->addrs.v6addrs.dst));
1138 if ((addr_diff < 0) ||
1139 (addr_diff == 0 &&
1140 ((__force u16)keys->ports.dst <
1141 (__force u16)keys->ports.src))) {
1142 for (i = 0; i < 4; i++)
1143 swap(keys->addrs.v6addrs.src.s6_addr32[i],
1144 keys->addrs.v6addrs.dst.s6_addr32[i]);
1145 swap(keys->ports.src, keys->ports.dst);
1146 }
1147 break;
1148 }
1149}
1150
1151static inline u32 __flow_hash_from_keys(struct flow_keys *keys, u32 keyval)
1152{
1153 u32 hash;
1154
1155 __flow_hash_consistentify(keys);
1156
1157 hash = __flow_hash_words(flow_keys_hash_start(keys),
1158 flow_keys_hash_length(keys), keyval);
1159 if (!hash)
1160 hash = 1;
1161
1162 return hash;
1163}
1164
1165u32 flow_hash_from_keys(struct flow_keys *keys)
1166{
1167 __flow_hash_secret_init();
1168 return __flow_hash_from_keys(keys, hashrnd);
1169}
1170EXPORT_SYMBOL(flow_hash_from_keys);
1171
1172static inline u32 ___skb_get_hash(const struct sk_buff *skb,
1173 struct flow_keys *keys, u32 keyval)
1174{
1175 skb_flow_dissect_flow_keys(skb, keys,
1176 FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL);
1177
1178 return __flow_hash_from_keys(keys, keyval);
1179}
1180
1181struct _flow_keys_digest_data {
1182 __be16 n_proto;
1183 u8 ip_proto;
1184 u8 padding;
1185 __be32 ports;
1186 __be32 src;
1187 __be32 dst;
1188};
1189
1190void make_flow_keys_digest(struct flow_keys_digest *digest,
1191 const struct flow_keys *flow)
1192{
1193 struct _flow_keys_digest_data *data =
1194 (struct _flow_keys_digest_data *)digest;
1195
1196 BUILD_BUG_ON(sizeof(*data) > sizeof(*digest));
1197
1198 memset(digest, 0, sizeof(*digest));
1199
1200 data->n_proto = flow->basic.n_proto;
1201 data->ip_proto = flow->basic.ip_proto;
1202 data->ports = flow->ports.ports;
1203 data->src = flow->addrs.v4addrs.src;
1204 data->dst = flow->addrs.v4addrs.dst;
1205}
1206EXPORT_SYMBOL(make_flow_keys_digest);
1207
1208static struct flow_dissector flow_keys_dissector_symmetric __read_mostly;
1209
1210u32 __skb_get_hash_symmetric(const struct sk_buff *skb)
1211{
1212 struct flow_keys keys;
1213
1214 __flow_hash_secret_init();
1215
1216 memset(&keys, 0, sizeof(keys));
1217 __skb_flow_dissect(skb, &flow_keys_dissector_symmetric, &keys,
1218 NULL, 0, 0, 0,
1219 FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL);
1220
1221 return __flow_hash_from_keys(&keys, hashrnd);
1222}
1223EXPORT_SYMBOL_GPL(__skb_get_hash_symmetric);
1224
1225/**
1226 * __skb_get_hash: calculate a flow hash
1227 * @skb: sk_buff to calculate flow hash from
1228 *
1229 * This function calculates a flow hash based on src/dst addresses
1230 * and src/dst port numbers. Sets hash in skb to non-zero hash value
1231 * on success, zero indicates no valid hash. Also, sets l4_hash in skb
1232 * if hash is a canonical 4-tuple hash over transport ports.
1233 */
1234void __skb_get_hash(struct sk_buff *skb)
1235{
1236 struct flow_keys keys;
1237 u32 hash;
1238
1239 __flow_hash_secret_init();
1240
1241 hash = ___skb_get_hash(skb, &keys, hashrnd);
1242
1243 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
1244}
1245EXPORT_SYMBOL(__skb_get_hash);
1246
1247__u32 skb_get_hash_perturb(const struct sk_buff *skb, u32 perturb)
1248{
1249 struct flow_keys keys;
1250
1251 return ___skb_get_hash(skb, &keys, perturb);
1252}
1253EXPORT_SYMBOL(skb_get_hash_perturb);
1254
1255u32 __skb_get_poff(const struct sk_buff *skb, void *data,
1256 const struct flow_keys *keys, int hlen)
1257{
1258 u32 poff = keys->control.thoff;
1259
1260 /* skip L4 headers for fragments after the first */
1261 if ((keys->control.flags & FLOW_DIS_IS_FRAGMENT) &&
1262 !(keys->control.flags & FLOW_DIS_FIRST_FRAG))
1263 return poff;
1264
1265 switch (keys->basic.ip_proto) {
1266 case IPPROTO_TCP: {
1267 /* access doff as u8 to avoid unaligned access */
1268 const u8 *doff;
1269 u8 _doff;
1270
1271 doff = __skb_header_pointer(skb, poff + 12, sizeof(_doff),
1272 data, hlen, &_doff);
1273 if (!doff)
1274 return poff;
1275
1276 poff += max_t(u32, sizeof(struct tcphdr), (*doff & 0xF0) >> 2);
1277 break;
1278 }
1279 case IPPROTO_UDP:
1280 case IPPROTO_UDPLITE:
1281 poff += sizeof(struct udphdr);
1282 break;
1283 /* For the rest, we do not really care about header
1284 * extensions at this point for now.
1285 */
1286 case IPPROTO_ICMP:
1287 poff += sizeof(struct icmphdr);
1288 break;
1289 case IPPROTO_ICMPV6:
1290 poff += sizeof(struct icmp6hdr);
1291 break;
1292 case IPPROTO_IGMP:
1293 poff += sizeof(struct igmphdr);
1294 break;
1295 case IPPROTO_DCCP:
1296 poff += sizeof(struct dccp_hdr);
1297 break;
1298 case IPPROTO_SCTP:
1299 poff += sizeof(struct sctphdr);
1300 break;
1301 }
1302
1303 return poff;
1304}
1305
1306/**
1307 * skb_get_poff - get the offset to the payload
1308 * @skb: sk_buff to get the payload offset from
1309 *
1310 * The function will get the offset to the payload as far as it could
1311 * be dissected. The main user is currently BPF, so that we can dynamically
1312 * truncate packets without needing to push actual payload to the user
1313 * space and can analyze headers only, instead.
1314 */
1315u32 skb_get_poff(const struct sk_buff *skb)
1316{
1317 struct flow_keys keys;
1318
1319 if (!skb_flow_dissect_flow_keys(skb, &keys, 0))
1320 return 0;
1321
1322 return __skb_get_poff(skb, skb->data, &keys, skb_headlen(skb));
1323}
1324
1325__u32 __get_hash_from_flowi6(const struct flowi6 *fl6, struct flow_keys *keys)
1326{
1327 memset(keys, 0, sizeof(*keys));
1328
1329 memcpy(&keys->addrs.v6addrs.src, &fl6->saddr,
1330 sizeof(keys->addrs.v6addrs.src));
1331 memcpy(&keys->addrs.v6addrs.dst, &fl6->daddr,
1332 sizeof(keys->addrs.v6addrs.dst));
1333 keys->control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
1334 keys->ports.src = fl6->fl6_sport;
1335 keys->ports.dst = fl6->fl6_dport;
1336 keys->keyid.keyid = fl6->fl6_gre_key;
1337 keys->tags.flow_label = (__force u32)fl6->flowlabel;
1338 keys->basic.ip_proto = fl6->flowi6_proto;
1339
1340 return flow_hash_from_keys(keys);
1341}
1342EXPORT_SYMBOL(__get_hash_from_flowi6);
1343
1344static const struct flow_dissector_key flow_keys_dissector_keys[] = {
1345 {
1346 .key_id = FLOW_DISSECTOR_KEY_CONTROL,
1347 .offset = offsetof(struct flow_keys, control),
1348 },
1349 {
1350 .key_id = FLOW_DISSECTOR_KEY_BASIC,
1351 .offset = offsetof(struct flow_keys, basic),
1352 },
1353 {
1354 .key_id = FLOW_DISSECTOR_KEY_IPV4_ADDRS,
1355 .offset = offsetof(struct flow_keys, addrs.v4addrs),
1356 },
1357 {
1358 .key_id = FLOW_DISSECTOR_KEY_IPV6_ADDRS,
1359 .offset = offsetof(struct flow_keys, addrs.v6addrs),
1360 },
1361 {
1362 .key_id = FLOW_DISSECTOR_KEY_TIPC,
1363 .offset = offsetof(struct flow_keys, addrs.tipckey),
1364 },
1365 {
1366 .key_id = FLOW_DISSECTOR_KEY_PORTS,
1367 .offset = offsetof(struct flow_keys, ports),
1368 },
1369 {
1370 .key_id = FLOW_DISSECTOR_KEY_VLAN,
1371 .offset = offsetof(struct flow_keys, vlan),
1372 },
1373 {
1374 .key_id = FLOW_DISSECTOR_KEY_FLOW_LABEL,
1375 .offset = offsetof(struct flow_keys, tags),
1376 },
1377 {
1378 .key_id = FLOW_DISSECTOR_KEY_GRE_KEYID,
1379 .offset = offsetof(struct flow_keys, keyid),
1380 },
1381};
1382
1383static const struct flow_dissector_key flow_keys_dissector_symmetric_keys[] = {
1384 {
1385 .key_id = FLOW_DISSECTOR_KEY_CONTROL,
1386 .offset = offsetof(struct flow_keys, control),
1387 },
1388 {
1389 .key_id = FLOW_DISSECTOR_KEY_BASIC,
1390 .offset = offsetof(struct flow_keys, basic),
1391 },
1392 {
1393 .key_id = FLOW_DISSECTOR_KEY_IPV4_ADDRS,
1394 .offset = offsetof(struct flow_keys, addrs.v4addrs),
1395 },
1396 {
1397 .key_id = FLOW_DISSECTOR_KEY_IPV6_ADDRS,
1398 .offset = offsetof(struct flow_keys, addrs.v6addrs),
1399 },
1400 {
1401 .key_id = FLOW_DISSECTOR_KEY_PORTS,
1402 .offset = offsetof(struct flow_keys, ports),
1403 },
1404};
1405
1406static const struct flow_dissector_key flow_keys_buf_dissector_keys[] = {
1407 {
1408 .key_id = FLOW_DISSECTOR_KEY_CONTROL,
1409 .offset = offsetof(struct flow_keys, control),
1410 },
1411 {
1412 .key_id = FLOW_DISSECTOR_KEY_BASIC,
1413 .offset = offsetof(struct flow_keys, basic),
1414 },
1415};
1416
1417struct flow_dissector flow_keys_dissector __read_mostly;
1418EXPORT_SYMBOL(flow_keys_dissector);
1419
1420struct flow_dissector flow_keys_buf_dissector __read_mostly;
1421
1422static int __init init_default_flow_dissectors(void)
1423{
1424 skb_flow_dissector_init(&flow_keys_dissector,
1425 flow_keys_dissector_keys,
1426 ARRAY_SIZE(flow_keys_dissector_keys));
1427 skb_flow_dissector_init(&flow_keys_dissector_symmetric,
1428 flow_keys_dissector_symmetric_keys,
1429 ARRAY_SIZE(flow_keys_dissector_symmetric_keys));
1430 skb_flow_dissector_init(&flow_keys_buf_dissector,
1431 flow_keys_buf_dissector_keys,
1432 ARRAY_SIZE(flow_keys_buf_dissector_keys));
1433 return 0;
1434}
1435
1436core_initcall(init_default_flow_dissectors);