2 * Copyright (c) 2009, 2010, 2011, 2012 Nicira, Inc.
4 * Licensed under the Apache License, Version 2.0 (the "License");
5 * you may not use this file except in compliance with the License.
6 * You may obtain a copy of the License at:
8 * http://www.apache.org/licenses/LICENSE-2.0
10 * Unless required by applicable law or agreed to in writing, software
11 * distributed under the License is distributed on an "AS IS" BASIS,
12 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13 * See the License for the specific language governing permissions and
14 * limitations under the License.
20 #include <arpa/inet.h>
21 #include <sys/socket.h>
22 #include <netinet/in.h>
24 #include "byte-order.h"
28 #include "dynamic-string.h"
31 const struct in6_addr in6addr_exact
= IN6ADDR_EXACT_INIT
;
33 /* Parses 's' as a 16-digit hexadecimal number representing a datapath ID. On
34 * success stores the dpid into '*dpidp' and returns true, on failure stores 0
35 * into '*dpidp' and returns false.
37 * Rejects an all-zeros dpid as invalid. */
39 dpid_from_string(const char *s
, uint64_t *dpidp
)
41 *dpidp
= (strlen(s
) == 16 && strspn(s
, "0123456789abcdefABCDEF") == 16
42 ? strtoull(s
, NULL
, 16)
47 /* Returns true if 'ea' is a reserved address, that a bridge must never
48 * forward, false otherwise.
50 * If you change this function's behavior, please update corresponding
51 * documentation in vswitch.xml at the same time. */
53 eth_addr_is_reserved(const uint8_t ea
[ETH_ADDR_LEN
])
55 struct eth_addr_node
{
56 struct hmap_node hmap_node
;
60 static struct eth_addr_node nodes
[] = {
61 /* STP, IEEE pause frames, and other reserved protocols. */
62 { HMAP_NODE_NULL_INITIALIZER
, 0x0108c2000000ULL
},
63 { HMAP_NODE_NULL_INITIALIZER
, 0x0108c2000001ULL
},
64 { HMAP_NODE_NULL_INITIALIZER
, 0x0108c2000002ULL
},
65 { HMAP_NODE_NULL_INITIALIZER
, 0x0108c2000003ULL
},
66 { HMAP_NODE_NULL_INITIALIZER
, 0x0108c2000004ULL
},
67 { HMAP_NODE_NULL_INITIALIZER
, 0x0108c2000005ULL
},
68 { HMAP_NODE_NULL_INITIALIZER
, 0x0108c2000006ULL
},
69 { HMAP_NODE_NULL_INITIALIZER
, 0x0108c2000007ULL
},
70 { HMAP_NODE_NULL_INITIALIZER
, 0x0108c2000008ULL
},
71 { HMAP_NODE_NULL_INITIALIZER
, 0x0108c2000009ULL
},
72 { HMAP_NODE_NULL_INITIALIZER
, 0x0108c200000aULL
},
73 { HMAP_NODE_NULL_INITIALIZER
, 0x0108c200000bULL
},
74 { HMAP_NODE_NULL_INITIALIZER
, 0x0108c200000cULL
},
75 { HMAP_NODE_NULL_INITIALIZER
, 0x0108c200000dULL
},
76 { HMAP_NODE_NULL_INITIALIZER
, 0x0108c200000eULL
},
77 { HMAP_NODE_NULL_INITIALIZER
, 0x0108c200000fULL
},
79 /* Extreme protocols. */
80 { HMAP_NODE_NULL_INITIALIZER
, 0x00e02b000000ULL
}, /* EDP. */
81 { HMAP_NODE_NULL_INITIALIZER
, 0x00e02b000004ULL
}, /* EAPS. */
82 { HMAP_NODE_NULL_INITIALIZER
, 0x00e02b000006ULL
}, /* EAPS. */
84 /* Cisco protocols. */
85 { HMAP_NODE_NULL_INITIALIZER
, 0x01000c000000ULL
}, /* ISL. */
86 { HMAP_NODE_NULL_INITIALIZER
, 0x01000cccccccULL
}, /* PAgP, UDLD, CDP,
88 { HMAP_NODE_NULL_INITIALIZER
, 0x01000ccccccdULL
}, /* PVST+. */
89 { HMAP_NODE_NULL_INITIALIZER
, 0x01000ccdcdcdULL
}, /* STP Uplink Fast,
93 { HMAP_NODE_NULL_INITIALIZER
, 0x01000cccccc0ULL
},
94 { HMAP_NODE_NULL_INITIALIZER
, 0x01000cccccc1ULL
},
95 { HMAP_NODE_NULL_INITIALIZER
, 0x01000cccccc2ULL
},
96 { HMAP_NODE_NULL_INITIALIZER
, 0x01000cccccc3ULL
},
97 { HMAP_NODE_NULL_INITIALIZER
, 0x01000cccccc4ULL
},
98 { HMAP_NODE_NULL_INITIALIZER
, 0x01000cccccc5ULL
},
99 { HMAP_NODE_NULL_INITIALIZER
, 0x01000cccccc6ULL
},
100 { HMAP_NODE_NULL_INITIALIZER
, 0x01000cccccc7ULL
},
103 static struct hmap addrs
= HMAP_INITIALIZER(&addrs
);
104 struct eth_addr_node
*node
;
107 if (hmap_is_empty(&addrs
)) {
108 for (node
= nodes
; node
< &nodes
[ARRAY_SIZE(nodes
)]; node
++) {
109 hmap_insert(&addrs
, &node
->hmap_node
,
110 hash_2words(node
->ea64
, node
->ea64
>> 32));
114 ea64
= eth_addr_to_uint64(ea
);
115 HMAP_FOR_EACH_IN_BUCKET (node
, hmap_node
, hash_2words(ea64
, ea64
>> 32),
117 if (node
->ea64
== ea64
) {
125 eth_addr_from_string(const char *s
, uint8_t ea
[ETH_ADDR_LEN
])
127 if (sscanf(s
, ETH_ADDR_SCAN_FMT
, ETH_ADDR_SCAN_ARGS(ea
))
128 == ETH_ADDR_SCAN_COUNT
) {
131 memset(ea
, 0, ETH_ADDR_LEN
);
136 /* Fills 'b' with a Reverse ARP packet with Ethernet source address 'eth_src'.
137 * This function is used by Open vSwitch to compose packets in cases where
138 * context is important but content doesn't (or shouldn't) matter.
140 * The returned packet has enough headroom to insert an 802.1Q VLAN header if
143 compose_rarp(struct ofpbuf
*b
, const uint8_t eth_src
[ETH_ADDR_LEN
])
145 struct eth_header
*eth
;
146 struct arp_eth_header
*arp
;
149 ofpbuf_prealloc_tailroom(b
, ETH_HEADER_LEN
+ VLAN_HEADER_LEN
150 + ARP_ETH_HEADER_LEN
);
151 ofpbuf_reserve(b
, VLAN_HEADER_LEN
);
152 eth
= ofpbuf_put_uninit(b
, sizeof *eth
);
153 memcpy(eth
->eth_dst
, eth_addr_broadcast
, ETH_ADDR_LEN
);
154 memcpy(eth
->eth_src
, eth_src
, ETH_ADDR_LEN
);
155 eth
->eth_type
= htons(ETH_TYPE_RARP
);
157 arp
= ofpbuf_put_uninit(b
, sizeof *arp
);
158 arp
->ar_hrd
= htons(ARP_HRD_ETHERNET
);
159 arp
->ar_pro
= htons(ARP_PRO_IP
);
160 arp
->ar_hln
= sizeof arp
->ar_sha
;
161 arp
->ar_pln
= sizeof arp
->ar_spa
;
162 arp
->ar_op
= htons(ARP_OP_RARP
);
163 memcpy(arp
->ar_sha
, eth_src
, ETH_ADDR_LEN
);
164 arp
->ar_spa
= htonl(0);
165 memcpy(arp
->ar_tha
, eth_src
, ETH_ADDR_LEN
);
166 arp
->ar_tpa
= htonl(0);
169 /* Insert VLAN header according to given TCI. Packet passed must be Ethernet
170 * packet. Ignores the CFI bit of 'tci' using 0 instead.
172 * Also sets 'packet->l2' to point to the new Ethernet header. */
174 eth_push_vlan(struct ofpbuf
*packet
, ovs_be16 tci
)
176 struct eth_header
*eh
= packet
->data
;
177 struct vlan_eth_header
*veh
;
179 /* Insert new 802.1Q header. */
180 struct vlan_eth_header tmp
;
181 memcpy(tmp
.veth_dst
, eh
->eth_dst
, ETH_ADDR_LEN
);
182 memcpy(tmp
.veth_src
, eh
->eth_src
, ETH_ADDR_LEN
);
183 tmp
.veth_type
= htons(ETH_TYPE_VLAN
);
184 tmp
.veth_tci
= tci
& htons(~VLAN_CFI
);
185 tmp
.veth_next_type
= eh
->eth_type
;
187 veh
= ofpbuf_push_uninit(packet
, VLAN_HEADER_LEN
);
188 memcpy(veh
, &tmp
, sizeof tmp
);
190 packet
->l2
= packet
->data
;
193 /* Removes outermost VLAN header (if any is present) from 'packet'.
195 * 'packet->l2' must initially point to 'packet''s Ethernet header. */
197 eth_pop_vlan(struct ofpbuf
*packet
)
199 struct vlan_eth_header
*veh
= packet
->l2
;
200 if (packet
->size
>= sizeof *veh
201 && veh
->veth_type
== htons(ETH_TYPE_VLAN
)) {
202 struct eth_header tmp
;
204 memcpy(tmp
.eth_dst
, veh
->veth_dst
, ETH_ADDR_LEN
);
205 memcpy(tmp
.eth_src
, veh
->veth_src
, ETH_ADDR_LEN
);
206 tmp
.eth_type
= veh
->veth_next_type
;
208 ofpbuf_pull(packet
, VLAN_HEADER_LEN
);
209 packet
->l2
= (char*)packet
->l2
+ VLAN_HEADER_LEN
;
210 memcpy(packet
->data
, &tmp
, sizeof tmp
);
214 /* Converts hex digits in 'hex' to an Ethernet packet in '*packetp'. The
215 * caller must free '*packetp'. On success, returns NULL. On failure, returns
216 * an error message and stores NULL in '*packetp'. */
218 eth_from_hex(const char *hex
, struct ofpbuf
**packetp
)
220 struct ofpbuf
*packet
;
222 packet
= *packetp
= ofpbuf_new(strlen(hex
) / 2);
224 if (ofpbuf_put_hex(packet
, hex
, NULL
)[0] != '\0') {
225 ofpbuf_delete(packet
);
227 return "Trailing garbage in packet data";
230 if (packet
->size
< ETH_HEADER_LEN
) {
231 ofpbuf_delete(packet
);
233 return "Packet data too short for Ethernet";
240 eth_format_masked(const uint8_t eth
[ETH_ADDR_LEN
],
241 const uint8_t mask
[ETH_ADDR_LEN
], struct ds
*s
)
243 ds_put_format(s
, ETH_ADDR_FMT
, ETH_ADDR_ARGS(eth
));
244 if (mask
&& !eth_mask_is_exact(mask
)) {
245 ds_put_format(s
, "/"ETH_ADDR_FMT
, ETH_ADDR_ARGS(mask
));
250 eth_addr_bitand(const uint8_t src
[ETH_ADDR_LEN
],
251 const uint8_t mask
[ETH_ADDR_LEN
],
252 uint8_t dst
[ETH_ADDR_LEN
])
256 for (i
= 0; i
< ETH_ADDR_LEN
; i
++) {
257 dst
[i
] = src
[i
] & mask
[i
];
261 /* Given the IP netmask 'netmask', returns the number of bits of the IP address
262 * that it specifies, that is, the number of 1-bits in 'netmask'.
264 * If 'netmask' is not a CIDR netmask (see ip_is_cidr()), the return value will
265 * still be in the valid range but isn't otherwise meaningful. */
267 ip_count_cidr_bits(ovs_be32 netmask
)
269 return 32 - ctz(ntohl(netmask
));
273 ip_format_masked(ovs_be32 ip
, ovs_be32 mask
, struct ds
*s
)
275 ds_put_format(s
, IP_FMT
, IP_ARGS(&ip
));
276 if (mask
!= htonl(UINT32_MAX
)) {
277 if (ip_is_cidr(mask
)) {
278 ds_put_format(s
, "/%d", ip_count_cidr_bits(mask
));
280 ds_put_format(s
, "/"IP_FMT
, IP_ARGS(&mask
));
286 /* Stores the string representation of the IPv6 address 'addr' into the
287 * character array 'addr_str', which must be at least INET6_ADDRSTRLEN
290 format_ipv6_addr(char *addr_str
, const struct in6_addr
*addr
)
292 inet_ntop(AF_INET6
, addr
, addr_str
, INET6_ADDRSTRLEN
);
296 print_ipv6_addr(struct ds
*string
, const struct in6_addr
*addr
)
300 ds_reserve(string
, string
->length
+ INET6_ADDRSTRLEN
);
302 dst
= string
->string
+ string
->length
;
303 format_ipv6_addr(dst
, addr
);
304 string
->length
+= strlen(dst
);
308 print_ipv6_masked(struct ds
*s
, const struct in6_addr
*addr
,
309 const struct in6_addr
*mask
)
311 print_ipv6_addr(s
, addr
);
312 if (mask
&& !ipv6_mask_is_exact(mask
)) {
313 if (ipv6_is_cidr(mask
)) {
314 int cidr_bits
= ipv6_count_cidr_bits(mask
);
315 ds_put_format(s
, "/%d", cidr_bits
);
318 print_ipv6_addr(s
, mask
);
323 struct in6_addr
ipv6_addr_bitand(const struct in6_addr
*a
,
324 const struct in6_addr
*b
)
330 for (i
=0; i
<4; i
++) {
331 dst
.s6_addr32
[i
] = a
->s6_addr32
[i
] & b
->s6_addr32
[i
];
334 for (i
=0; i
<16; i
++) {
335 dst
.s6_addr
[i
] = a
->s6_addr
[i
] & b
->s6_addr
[i
];
342 /* Returns an in6_addr consisting of 'mask' high-order 1-bits and 128-N
343 * low-order 0-bits. */
345 ipv6_create_mask(int mask
)
347 struct in6_addr netmask
;
348 uint8_t *netmaskp
= &netmask
.s6_addr
[0];
350 memset(&netmask
, 0, sizeof netmask
);
358 *netmaskp
= 0xff << (8 - mask
);
364 /* Given the IPv6 netmask 'netmask', returns the number of bits of the IPv6
365 * address that it specifies, that is, the number of 1-bits in 'netmask'.
366 * 'netmask' must be a CIDR netmask (see ipv6_is_cidr()).
368 * If 'netmask' is not a CIDR netmask (see ipv6_is_cidr()), the return value
369 * will still be in the valid range but isn't otherwise meaningful. */
371 ipv6_count_cidr_bits(const struct in6_addr
*netmask
)
375 const uint8_t *netmaskp
= &netmask
->s6_addr
[0];
377 for (i
=0; i
<16; i
++) {
378 if (netmaskp
[i
] == 0xff) {
383 for(nm
= netmaskp
[i
]; nm
; nm
<<= 1) {
394 /* Returns true if 'netmask' is a CIDR netmask, that is, if it consists of N
395 * high-order 1-bits and 128-N low-order 0-bits. */
397 ipv6_is_cidr(const struct in6_addr
*netmask
)
399 const uint8_t *netmaskp
= &netmask
->s6_addr
[0];
402 for (i
=0; i
<16; i
++) {
403 if (netmaskp
[i
] != 0xff) {
404 uint8_t x
= ~netmaskp
[i
];
419 /* Populates 'b' with an Ethernet II packet headed with the given 'eth_dst',
420 * 'eth_src' and 'eth_type' parameters. A payload of 'size' bytes is allocated
421 * in 'b' and returned. This payload may be populated with appropriate
422 * information by the caller. Sets 'b''s 'l2' and 'l3' pointers to the
423 * Ethernet header and payload respectively.
425 * The returned packet has enough headroom to insert an 802.1Q VLAN header if
428 eth_compose(struct ofpbuf
*b
, const uint8_t eth_dst
[ETH_ADDR_LEN
],
429 const uint8_t eth_src
[ETH_ADDR_LEN
], uint16_t eth_type
,
433 struct eth_header
*eth
;
437 ofpbuf_prealloc_tailroom(b
, ETH_HEADER_LEN
+ VLAN_HEADER_LEN
+ size
);
438 ofpbuf_reserve(b
, VLAN_HEADER_LEN
);
439 eth
= ofpbuf_put_uninit(b
, ETH_HEADER_LEN
);
440 data
= ofpbuf_put_uninit(b
, size
);
442 memcpy(eth
->eth_dst
, eth_dst
, ETH_ADDR_LEN
);
443 memcpy(eth
->eth_src
, eth_src
, ETH_ADDR_LEN
);
444 eth
->eth_type
= htons(eth_type
);
453 packet_set_ipv4_addr(struct ofpbuf
*packet
, ovs_be32
*addr
, ovs_be32 new_addr
)
455 struct ip_header
*nh
= packet
->l3
;
457 if (nh
->ip_proto
== IPPROTO_TCP
&& packet
->l7
) {
458 struct tcp_header
*th
= packet
->l4
;
460 th
->tcp_csum
= recalc_csum32(th
->tcp_csum
, *addr
, new_addr
);
461 } else if (nh
->ip_proto
== IPPROTO_UDP
&& packet
->l7
) {
462 struct udp_header
*uh
= packet
->l4
;
465 uh
->udp_csum
= recalc_csum32(uh
->udp_csum
, *addr
, new_addr
);
467 uh
->udp_csum
= htons(0xffff);
471 nh
->ip_csum
= recalc_csum32(nh
->ip_csum
, *addr
, new_addr
);
475 /* Modifies the IPv4 header fields of 'packet' to be consistent with 'src',
476 * 'dst', 'tos', and 'ttl'. Updates 'packet''s L4 checksums as appropriate.
477 * 'packet' must contain a valid IPv4 packet with correctly populated l[347]
480 packet_set_ipv4(struct ofpbuf
*packet
, ovs_be32 src
, ovs_be32 dst
,
481 uint8_t tos
, uint8_t ttl
)
483 struct ip_header
*nh
= packet
->l3
;
485 if (nh
->ip_src
!= src
) {
486 packet_set_ipv4_addr(packet
, &nh
->ip_src
, src
);
489 if (nh
->ip_dst
!= dst
) {
490 packet_set_ipv4_addr(packet
, &nh
->ip_dst
, dst
);
493 if (nh
->ip_tos
!= tos
) {
494 uint8_t *field
= &nh
->ip_tos
;
496 nh
->ip_csum
= recalc_csum16(nh
->ip_csum
, htons((uint16_t) *field
),
497 htons((uint16_t) tos
));
501 if (nh
->ip_ttl
!= ttl
) {
502 uint8_t *field
= &nh
->ip_ttl
;
504 nh
->ip_csum
= recalc_csum16(nh
->ip_csum
, htons(*field
<< 8),
511 packet_set_port(ovs_be16
*port
, ovs_be16 new_port
, ovs_be16
*csum
)
513 if (*port
!= new_port
) {
514 *csum
= recalc_csum16(*csum
, *port
, new_port
);
519 /* Sets the TCP source and destination port ('src' and 'dst' respectively) of
520 * the TCP header contained in 'packet'. 'packet' must be a valid TCP packet
521 * with its l4 marker properly populated. */
523 packet_set_tcp_port(struct ofpbuf
*packet
, ovs_be16 src
, ovs_be16 dst
)
525 struct tcp_header
*th
= packet
->l4
;
527 packet_set_port(&th
->tcp_src
, src
, &th
->tcp_csum
);
528 packet_set_port(&th
->tcp_dst
, dst
, &th
->tcp_csum
);
531 /* Sets the UDP source and destination port ('src' and 'dst' respectively) of
532 * the UDP header contained in 'packet'. 'packet' must be a valid UDP packet
533 * with its l4 marker properly populated. */
535 packet_set_udp_port(struct ofpbuf
*packet
, ovs_be16 src
, ovs_be16 dst
)
537 struct udp_header
*uh
= packet
->l4
;
540 packet_set_port(&uh
->udp_src
, src
, &uh
->udp_csum
);
541 packet_set_port(&uh
->udp_dst
, dst
, &uh
->udp_csum
);
544 uh
->udp_csum
= htons(0xffff);
552 /* If 'packet' is a TCP packet, returns the TCP flags. Otherwise, returns 0.
554 * 'flow' must be the flow corresponding to 'packet' and 'packet''s header
555 * pointers must be properly initialized (e.g. with flow_extract()). */
557 packet_get_tcp_flags(const struct ofpbuf
*packet
, const struct flow
*flow
)
559 if ((flow
->dl_type
== htons(ETH_TYPE_IP
) ||
560 flow
->dl_type
== htons(ETH_TYPE_IPV6
)) &&
561 flow
->nw_proto
== IPPROTO_TCP
&& packet
->l7
) {
562 const struct tcp_header
*tcp
= packet
->l4
;
563 return TCP_FLAGS(tcp
->tcp_ctl
);
569 /* Appends a string representation of the TCP flags value 'tcp_flags'
570 * (e.g. obtained via packet_get_tcp_flags() or TCP_FLAGS) to 's', in the
571 * format used by tcpdump. */
573 packet_format_tcp_flags(struct ds
*s
, uint8_t tcp_flags
)
576 ds_put_cstr(s
, "none");
580 if (tcp_flags
& TCP_SYN
) {
583 if (tcp_flags
& TCP_FIN
) {
586 if (tcp_flags
& TCP_PSH
) {
589 if (tcp_flags
& TCP_RST
) {
592 if (tcp_flags
& TCP_URG
) {
595 if (tcp_flags
& TCP_ACK
) {
598 if (tcp_flags
& 0x40) {
599 ds_put_cstr(s
, "[40]");
601 if (tcp_flags
& 0x80) {
602 ds_put_cstr(s
, "[80]");